More stories

  • in

    Easy computation of the Bayes factor to fully quantify Occam’s razor in least-squares fitting and to guide actions

    How many parameters best describe data in muon spectroscopy?Here we find that the Bayes factor demands the inclusion of more physically-meaningful parameters than the BIC or significance tests. Figure 1a presents some data that might reasonably be fitted with as few as three or as many as 22 physically-meaningful parameters. We find that the Bayes factor encourages the inclusion of all these parameters until the onset of over-fitting. Even though many of them have fitted values that fail significance tests (i.e. are consistent with zero), their omission distorts the fitting results severely.Figure 1Full size imageFigure 1a shows an anti-level-crossing spectrum observed in photo-excited muon-spin spectroscopy26 from an organic molecule27. The data are presented in Fig. 2a of Ref.27 and are given in the SI. These spectra are expected to be Lorentzian peaks. Theory permits optical excitation to affect the peak position, the width and the strength (photosensitivity). In the field region over which the measurements are carried out, there is a background from detection of positrons, which has been subtracted from the data presented27. Wang et al.27 did not attempt to fit the data rigorously; they did report a model-independent integration of the data, which demonstrated a change in area and position.The model that we fit hypothesises one or more Lorentzian peaks, with optional photosensitivity on each fitting parameter and with optional linear backgrounds y = a + bx underlying the peaks, described by the full equation given in the SI, equation (S3). To do a single LS fit to all the data, we extend the data to three dimensions, (x gauss, y asymmetry, z) where z = 0 for data in the dark and z = 1 for photoexcited data. Including all the data in a single LS fit in this way, rather than fitting the dark and photoexcited data separately, simplifies both setting up the fit and doing the subsequent analysis.Figure 1b shows the evolution of the SBIC and the lnBF as the number of fitting parameters in the model is increased. Starting with a single Lorentzian peak, three parameters are required, peak position P, width W and intensity A. Three photosensitivity parameters ΔLP, ΔLW and ΔLA are then introduced successively to the fit, (open and small data points for n = 3–6). The SBIC decreases and the lnMLI scarcely increases. It is only with the inclusion of one background term (n = 7) that any figure of merit shows any substantial increase. There is no evidence here for photosensitivity. The weak peak around 7050 G does not seem worth including in a fit, as it is evidenced by only two or three data points and is scarcely outside the error bars. However, a good fit with two peaks (P1 ~ 7210 G, P2 ~ 7150 G, the subscripts 1 and 2 in accordance with the site labelling of Fig. 2a of Ref.27) can be obtained with just five parameters (P1, P2, A1, A2, W). This gives substantial increases in the SBIC and lnMLI, further increased when W1 and W2 are distinguished and then when the single background term and the three photosensitivity parameters ΔLP2, ΔLW2 and ΔLA2 are successively included (solid or large data points for n = 5–10 in Fig. 1b). The SBIC reaches its maximum here, at n = 10, and then decreases substantially when the other three photosensitivity parameters and the other three background terms are included. These additional parameters fail significance tests as well as decreasing the SBIC (Fig. 1b). Conventionally, the n = 10 fit would be accepted as best. The outcome would be reported as two peaks, with significant photo-sensitivities ΔLP2, ΔLW2 and ΔLA2 for all three of the 7150 G peak parameters, but no photosensitivity for the 7210 G peak (Table 1).Table 1 Photosensitivity results of fitting the data of Fig. 1a with 10, 16 and 19 parameters. Parameter units as implied by Fig. 1a.Full size tableThe Bayes factor gives a very different outcome. From 10 to 16 parameters, the Bayes factor between any two of these seven models is close to unity (Fig. 1b). That is, they have approximately equal probability. The Bayes factor shows that what the conventional n = 10 analysis would report is false. Specifically, it is not the case that ΔLP2, reported as − 14 ± 4 G, has a roughly ({raise0.5exhbox{$scriptstyle 2$} kern-0.1em/kern-0.15em lower0.25exhbox{$scriptstyle 3$}}) probability of lying between − 10 and − 18 G. That is not consistent with the roughly equal probability that it lies in the n = 16 range (− 24 ± 8 G). Table 1 shows that at n = 16, ΔLP2 is the only photosensitivity parameter to pass significance tests. ΔLA2, which had the highest significance level at n = 10, is now the parameter most consistent with zero. The other four are suggestively (about 1({raise0.5exhbox{$scriptstyle 1$} kern-0.1em/kern-0.15em lower0.25exhbox{$scriptstyle 2$}})σ) different from zero.Since the Bayes factor has already radically changed the outcome by encouraging more physically-meaningful parameters, it is appropriate to try the 7050 G peak parameters in the fit. With only 28 data-points, we should be alert to over-fitting. We can include P3 and A3 (n = 18), and ΔLP3 (n = 19), but W3 and ΔLA3 do cause overfitting. Figure 1b shows substantial increases of both the SBIC and the lnMLI for n = 18 to n = 20, where the twentieth parameter is in fact ΔLA3. The symptom of over-fitting that we observe here is an increase in the logarithm of the Occam Factor (lnMLI − lnL), the values of which decrease, − 26.9, − 33.5, − 34.8, and then increase, − 33.4, for n = 16, 18, 19 and 20 respectively. Just as lnL must increase with every additional parameter, so should the Occam factor decrease, as the prior parameter volume should increase more with a new parameter than the posterior parameter volume. So we stop at n = 19. The outcome, Table 1, is that the uncertainties on the n = 16 parameters have decreased markedly. This is due to the better fit, with a substantial increase in lnL corresponding to reduced residuals on all the data. The 7210 G peak 2 now has photosensitivities on all its parameters, significant to at least the 2σ or p value ~ 0.05 level. And the photosensitivities ΔLW2 and ΔLA2, both so significant at n = 10, and already dwindling in significance at n = 16, are both now taking values quite consistent with zero. In the light of Table 1, we see that stopping the fit at n = 10 results in completely incorrect results—misleading fitted values, with certainly false uncertainties.Discriminating between models for the pressure dependence of the GaAs bandgapThe main purpose of this example is to show how the Bayes factor can be used to decide between two models which have equal goodness of fit to the data (equal values of lnL and BIC, as well as p values, etc.). This illustrates the distinction it makes between physically-meaningful and physically meaningless parameters. This example also shows how ML fitting can be used together with the Bayes factor to obtain better results. For details, see SI §7.Figure 2 shows two datasets for the pressure dependence of the bandgap of GaAs (data given in the SI). The original authors published quadratic fits, ({E}_{g}(P)={E}_{0}+bP+c{P}^{2}), with b = 10.8 ± 0.3 meV kbar−1 (Goñi et al.28) and 11.6 ± 0.2 meV kbar−1 (Perlin et al.29). Other reported experimental and calculated values for b ranged from 10.02 to 12.3 meV kbar−130. These discrepancies of about ± 10% were attributed to experimental errors in high-pressure experimentation. However, from a comparison of six such datasets, Frogley et al.30 were able to show that the discrepancies arose from fitting the data with the quadratic formula. The different datasets were reconciled by using the Murnaghan equation of state and supposing the band-gap to vary linearly with the density (see SI, §7, equations (S4) and (S5)30. The curvature c of the quadratic is constant, while the curvature of the density, due to the pressure dependence Bʹ of the bulk modulus B0, decreases with pressure—and the six datasets were recorded over very different pressure ranges, as in Fig. 2. So the fitted values of c, c0, were very different, and the correlation between b and c resulted in the variations in b0.Here, using the Bayes factor, we obtain the same result from a single dataset, that of Goñi et al.28 The two fits are shown in Fig. 2. They are equally good, with values of lnL and SBIC the same to 0.01. The key curvature parameters, c and ({text{B}}^{prime }), are both returned as non-zero by 13.5σ (SI, §7, Table S1), consequently both with p-values less than 10−18. However, c is a physically-meaningless parameter. The tightest constraint we have for setting its range is the values previously reported, ranging from 0 to 60 μeV kbar−2, so we use Δc = 100 μeV kbar−2. In contrast, ({text{B}}^{prime }) is known for GaAs to be 4.4931. For many other materials and from theory the range 4–5 is expected, so we use (Delta {text{B}}^{prime } = 1). The other ranges are same for both models (see SI §7). This difference gives a lnBF of 3.8 in favour of the Murnaghan model against the quadratic, which is strong evidence for it. Moreover, the value of ({text{B}}^{prime }) returned is 4.47 ± 0.33, in excellent agreement with the literature value. Had it been far out of range, the model would have to be rejected. The quadratic model is under no such constraint; indeed, a poor fit might be handled by adding cubic and higher terms ad lib. This justifies adding about 5 to lnBF (see “Background in fitting a carbon nanotube Raman spectrum” section), giving a decisive preference to the Murnaghan model, and the value of b it returns, 11.6 ± 0.3. Note the good agreement with the value from Perlin et al.29 If additionally we fix ({mathrm{B}}^{prime}) at its literature value of 4.4931, lnBF is scarcely improved, because the Occam factor against this parameter is small, but the uncertainty on the pressure coefficient, Ξ/B0, is much improved.When we fit the Perlin data, the Murnaghan fit returns ({text{B}}^{prime }) = 6.6 ± 2.4. This is outside range, and indicates that this data cannot give a reliable value—attempting it is over-fitting. However, it is good to fit this data together with the Goñi data. The Perlin data, very precise but at low pressures only, complement the Goñi data with their lower precision but large pressure range. We notice also that the Perlin data has a proportion of outlier data points. Weighted or rescaled LS fitting can handle the different precisions, but it cannot handle the outliers satisfactorily. Maximum Likelihood fitting handles both issues. We construct lnL using different pdfs P(r) for the two datasets, and with a double-Gaussian pdf for the Perlin data (see equation (S6) in the SI §7). Fixing ({text{B}}^{prime }) at 4.49, fitting with the same Ξ/B0 returns 11.42 ± 0.04 meV kbar−1. Separate Ξ/B0 parameters for the two datasets give an increase of lnL of 4.6, with values 11.28 ± 0.06 and 11.60 ± 0.04 meV kbar−1—a difference in b of 0.32 ± 0.07 meV kbar−1, which is significant at 4½σ. This difference could be due to systematic error, e.g. in pressure calibration. Or it could be real. Goñi et al.28 used absorption spectroscopy to measure the band-gap; Perlin et al.29 used photoluminescence. The increase of the electron effective mass with pressure might give rise to the difference. In any case, it is clear that high-pressure experimentation is much more accurate than previously thought, and that ML fitting exploits the information in the data much better than LS fitting.Figure 2GaAs band-gap. Data for Eg(P) in GaAs from Goñi et al.28 (
    ) and from Perlin et al.29 (
    ) are shown after subtraction of the straight line E0 + 8.5P to make the curvature more visible. The Perlin data is expanded × 10 on both axes for clarity. Two least-squares fits to the Goñi data are shown, polynomial (dashed red line) and Murnaghan (solid blue line). (Figure prepared using Mathematica 12.0, www.wolfram.com/mathematica/).Full size imageBackground in fitting a carbon nanotube Raman spectrumThis example demonstrates how the Bayes Factor provides a quantitative answer to the problem, whether we should accept a lower quality of fit to the data if the parameter set is intuitively preferable. It also provides a simple example of a case where the MLI calculated by Eq. (1) is in error and can readily be corrected (see SI §8 Fig. S3).The dataset is a Raman spectrum of the radial breathing modes of a sample of carbon nanotubes under pressure32. The whole spectrum at several pressures is shown with fits in Fig. 1 of Ref.32. The traditional fitting procedure used there was to include Lorentzian peaks for the clear peaks in the spectra, and then to add broad peaks as required to get a good fit, but without quantitative figures of merit and without any attempt to explain the origin of the broad peaks, and therefore with no constraints on their position, widths or intensities. The key issue in the fitting was to get the intensities of the peaks as accurately as possible, to help understand their evolution with pressure. Here, we take a part of the spectrum recorded at 0.23 GPa (the data is given in the SI.) and we monitor the quality of fit and the Bayes factor while parameters are added in four models. This part of the spectrum has seven sharp pseudo-Voigt peaks (Fig. 3a; the two strong peaks are clearly doublets). With seven peak positions Pi, peak widths Wi and peak intensities Ai, and a factor describing the Gaussian content in the pseudo-Voigt peak shape, there are already 22 parameters (for details, see SI §8). This gives a visibly very poor fit, with lnL = − 440, SBIC = − 510 and lnMLI = − 546. The ranges chosen for these parameters for calculating the MLI (see SI §8) are not important because they are used in all the subsequent models, and so they cancel out in the Bayes factors between the models.Figure 3Carbon nanotube Raman spectrum. In (a), the carbon nanotube Raman spectrum is plotted (black datapoints) with a fit (cyan solid line) using the Fourier model. The residuals for four good fits are shown, × 10 and displaced successively downwards (Fourier, Polynomial, Peaks and Tails; all at lnL about − 60, see text). The backgrounds are shown, × 8 (long dashed, chain-dotted, short dashed and solid, respectively. In (b), the evolution of the MLIs is shown against the number of parameters for these four models. (Figure prepared using Mathematica 12.0, www.wolfram.com/mathematica/).Full size imageTo improve the fit, in the Fourier model we add a Fourier background (y=sum {c}_{i}mathrm{cos}ix+{s}_{i}mathrm{sin}ix) (i = 0,..) and in the Polynomial model, we add (y=sum {a}_{i}{x}^{i}) (i = 0,..) for the background. In both, the variable x is centred (x = 0) at the centre of the fitted spectrum and scaled to be ± π or ± 1 at the ends. In the Peaks model we add extra broad peaks as background, invoking extra parameter triplets (Pi, Wi, Ai). These three models all gave good fits; at the stage shown in Fig. 3a they gave lnL values of − 65, − 54 and − 51 and BIC values of − 156, − 153 and − 148 respectively. Thus there is not much to choose between the three models, but it is noteworthy that they give quite different values for the intensities of the weaker peaks, with the peak at 265 cm−1 at 20.5 ± 1.1, 25.5 ± 1.3 and 27 ± 1.7 respectively (this is related to the curvature of the background function under the peak). So it is important to choose wisely.A fourth model was motivated by the observation that the three backgrounds look as if they are related to the sharp peaks, rather like heavily broadened replicas (see Fig. 3a). Accordingly, in the fourth model, we use no background apart from the zeroth term c0 or a0 to account for dark current). Instead, the peak shape is modified, giving it stronger, fatter tails than the pseudo-Voigt peaks (Tails model). This was done by adding to the Lorentzian peak function a smooth function approximating to exponential tails on both sides of the peak position (for details, see SI §8) with widths and amplitudes as fitting parameters. What is added may be considered as background and is shown in Fig. 3a. This model, at the stage of Fig. 3a, returned lnL = − 62, BIC = − 146, and yet another, much smaller value of 15.5 ± 1.0 for the intensity of the 265 cm−1 peak.The Tails model is intuitively preferable to the other three because it does not span the data space—e.g. if there was really were broad peaks at the positions identified by the Peaks model, or elsewhere, the Tails model could not fit them well. That it does fit the data is intuitively strong evidence for its correctness. The Bayes factor confirms this intuition quantitatively. At the stage of Fig. 3a, the lnMLI values are − 251, − 237 and − 223 for the Fourier, Poly and Peaks models, and − 211 for the Tails model. This gives a lnBF value of 12 for the Tails model over the Peaks model—decisive—and still larger lnBF values for these models over the Fourier and Poly models.All models can be taken further, with more fitting parameters. More Fourier or polynomial terms or more peaks can be added, and for the Tails model more parameters distinguishing the tails attached to each of the seven Lorentizian peaks. In this way, the three background models can improve to a lnL ~ − 20; the Tails model does not improve above lnL ~ − 50. However, as seen in Fig. 3b, the MLIs get worse with too many parameters, except when over-fitting occurs, as seen for the Poly model at 35 parameters. The Tails model retains its positive lnBF  > 10 over the other models.The other models can have an indefinite number of additional parameters—more coefficients or more peaks, to fit any data set. It is in this sense that they span the data space. The actual number used is therefore itself a fitting parameter, with an uncertainty perhaps of the order of ± 1, and a range from 0 to perhaps a quarter or a half of the number of data points m. We may therefore penalise their lnMLIs by ~ ln 4 m−1 or about − 5 for a few hundred data points. This takes Tails to a lnBF  > 15 over the other models—overwhelmingly decisive. This quantifies the intuition that a model that is not guaranteed to fit the data, but which does, is preferable to a model that certainly can fit the data because it spans the data space. It quantifies the question, how much worse a quality of fit should we accept for a model that is intuitively more satisfying. Here we accept a loss of − 30 on lnL for a greater gain of + 45 in the Occam factor. It quantifies the argument that the Tails model is the most worthy of further investigation because the fat tails probably have a physical interpretation worth seeking. In this context, it is interesting that in Fig. 3a fat tails have been added only to the 250, 265 and 299 cm−1 peaks; adding fat tails to the others did not improve the fit; however, a full analysis and interpretation is outside the scope of this paper. In the Peaks model it is not probable (though possible) that the extra peaks would have physical meaning. In the other two models it is certainly not the case that their Fourier or polynomial coefficients will have physical meaning. More

  • in

    Exploiting time series of Sentinel-1 and Sentinel-2 to detect grassland mowing events using deep learning with reject region

    Study area and datasetThe study area covers all Estonia located between 57.5(^circ ) N, 21.5(^circ ) E and 59.8(^circ ) N, 28.2(^circ ) E. The study area is relatively flat with no steep slopes and altitudes ranging between 0 and 200m above the sea level. Data about events were collected directly from field books that contained information about the mowing activity’s start and end date and the covered area. Considering the main agricultural areas of the country, we consider 2000 fields in which events are geographically evenly distributed across all Estonia, as shown in Fig. 1. In total, data about 1800 mowing and 200 non-mown events were collected in 2018, based on manual labelling. During manual labelling, the specific mowing days were labelled based on the following: a) information recorded by farmers in field books regarding mowing days, b) domain experts knowledge about the most probable days for mowing based on the climate, weather, and field conditions, c) rapid decrease in the Normalized Difference Vegetation Index (NDVI) and rapid increase in the coherence compared to past measurements. The average field size is 6.0ha, and around 95% of the fields were mown during the year. 90% of the fields are in the range of (0.5-10)ha. The greatest density of the fields is located in Lääne-Viru, Tartu and Jõgeva countries. Grassland parcels vector layer is provided by Estonian Agricultural Registers and Information Board (ARIB)50. The satellite imagery used in the study is from Copernicus program that provides free open Earth observation data to help service providers, public authorities, and international organizations improve European citizens’ quality of life.Figure 1Geographic distribution of events used in this study (This map was created by QGIS version 3.16, which can be accessed on https://qgis.org/en/site/).Full size imageSentinel-1 and Sentinel-2 dataFor Sentinel-1 data, in total, 400 S1A/BSLCIW products acquired between 1st of May 20017 and 30th of October 2018, were processed. 87 products were from relative orbit number (RON)160, 62 from RON131, 84 from RON87, 93 from RON58, and 60 from RON29. These were organised into S1A/S1B 6-day pairs. Sentinel-2 provides high spatial resolution optical imagery to perform terrestrial observations with global coverage of the Earth’s land surface. Sentinel-2 data is provided by the European Space Agency (ESA) together with a cloud mask, which can filter clouds on the image with moderately good accuracy. 400 Sentinel-2A and -2B L2A products acquired between 1 May 2017 and 30 October 2018 were processed. Each Sentinel-2 image is a maximum of three days off from the closest Sentinel-1 image. Only the NDVI was derived from Sentinel-2. NDVI has been widely used in the classification of grassland24,51 and that is mainly due to its ability in limiting spectral noise. The spatial resolution of the derived Sentinel-2 NDVI feature is 10 m.MethodsThe goal of the analysis is to detect mowing events from Sentinel-1 (S-1) and Sentinel-2 (S-2) data. For this, coherence time series were calculated about every field in the database about the event. Average coherence of a field, imaging geometry parameters, imaging time and average NDVI were stored in a database. The database formation process involved preprocessing many satellite images where average coherence and NDVI value was calculated for every parcel for every available date (constrained by image availability and cloud cover). The overall scheme of the proposed methodology is illustrated in Fig. 2. First, the time-series data from S-1 and S-2 images are preprocessed. Then, the most important features are used in a deep neural network to predict mowing events. The model has a reject region option that enables the model to abstain from the prediction in case of uncertainty, which increases trust in the model.We used the Sentinel Application Platform (SNAP) toolbox for processing S-1 data. More specifically, we followed the same following pre-processing steps in16: apply orbit file, back-geocoding (using Shuttle Radar Topography Mission (SRTM) data), coherence calculation, deburst, terrain correction, and reprojection to the local projection (EPSG:3301). Lastly, we resampled the data to 4m resolution to preserve the maximum spatial resolution and square-shaped pixels. Because the study areas’ terrain is relatively flat, there are few topographic distortions in the SAR data. Each swath’s coherence was calculated independently. Only pixels totally inside the parcel boundaries (including the average window used for coherence computation) were utilized to calculate results, and any interference beyond the parcel limits was discarded. Pair-wise coherence was calculated with 6-day time step. The data was stored into a database using a forward-looking convention: coherence regarding date X refers to the coherence between S-1 images over the period between date X and X + 6 days. For preprocessing S-2 data, L1C and L2A Sentinel-2 products were obtained through Copernicus Open Access Hub6. Next, a rule-based cloud mask solution was applied52. Finally, the fourth and eighth bands were extracted to compute NDVI values.Figure 2Flowchart of the proposed approach to detect mowing events.Full size imageFeature extraction from Sentinel-1 dataCoherence is a normalized measure of similarity between two consecutive (same relative orbit) S-1 images. Interferometric 6 day repeat pass coherence in VV polarization (cohvv), and coherence in VH polarization (cohvh) are chosen features as they are shown to be sensitive to changes in vegetation and agricultural events25. The shorter the time interval after the mowing event and the first interferometric acquisition, the higher the coherence value. Generally, up to 24 to 36 days after a mowing event, coherence stays relatively high. Precipitation caused the coherence to drop, which disturbs the detection of a mowing event. The spatial resolution of the S-1 6-day repeat pass interferometric coherence is 70 m. Given two S-1 images (s_{1}) and (s_{2}), coherence is calculated as follows:$$begin{aligned} wp =frac{|langle s_{1}s*_{2}rangle |}{sqrt{langle s_{1}s*_{1}rangle | langle s_{2}s*_{2}rangle |}} end{aligned}$$
    (1)
    where (|langle s_{1}s*_{2}rangle |) is the absolute value of the spatial average of the complex conjugate product.Coherence between two S-1 images (s_1) and (s_2) reaches its maximum value of 1 when both images have the same position and physical characteristics of the scatters. In contrast, the coherence value declines when the position or properties of the scatters change.Feature extraction from Sentinel-2 dataNDVI is related to the amount of live green vegetation. Generally, NDVI increases and decreases over the season, indicating the natural growth decay of vegetation, while the significant drops in the NDVI indicate an agricultural event such as mowing. NDVI is derived from S2 images and is calculated as follows:$$begin{aligned} NDVI=frac{band8 – band4}{band8 + band4} end{aligned}$$
    (2)
    Figure 3Typical signature of NDVI and coherence in VV and VH polarisation for non mown field during the year.Full size imageFigure 4Field with single mowing event during the year.Full size imageFigure 5NDVI measurement for a field example with a single mowing event during the season.Full size imageFigures 3, 4 and 5 show different samples of mown and non mown fields. NDVI measurements are green, cohvh and cohvv are blue and black, respectively. For non mown field, the typical signature of NDVI during the year is shown in Fig. 3. For non mown field, the typical signature of NDVI during the season is a half-oval curve; coherence is not stable but remains at almost the same level without apparent trend changes, as shown in Fig. 3. An example of a field with a single mowing event during the season is shown in Fig. 4. A mowing event is characterized by a rapid increase in both cohvh and cohvv and a sharp decrease in NDVI, as observed at day 150 (See Fig. 4). Forty days later, a similar signature is probably not due to a mowing event but likely caused by drought during summer.Notably, NDVI measurements are irregular and relatively sparse. Around 75% of total NDVI measurements are invalid in Estonia, and the percentage is slightly lower in Southern Sweden and Denmark due to cloud cover. The Cloud mask indicates the percentage of cloud coverage and allows the cloudy and cloud-free pixels to be identified. Using the standard cloud mask technique by the European Space Agency (ESA) leads to outliers noticed in the sudden decrease in the NDVI. Figure 5 shows an extreme value of NDVI that is supposed to be an outlier due to high differences to the precedent and subsequent values. The outlier is marked with a yellow dot (NDVI=0.38), nearest previous (NDVI=0.75), and next (NDVI=0.78) measurements are marked with a blue colour.Sentinel-1 and Sentinel-2 data preprocessingTo detect NDVI outliers effectively, a good understanding of the data is needed. NDVI outliers due to cloud mask errors rarely co-occur together, and hence, they can be treated as independent events53. NDVI outliers are usually identified with a sudden drop to almost zero and do not form a sequence. It is enough to look at neighbouring measurements (one before and one after) to detect individual outliers. If the difference between the adjacent measurements is high, this is an outlier signature. Hence, outliers can be handled by iterating through every three consecutive NDVI measurements for a given field and checking the difference between the first and second values and between third and second values. Figure 6 shows the scatter plot of all three consecutive NDVI measurements. The Y-axis shows the difference between third and second NDVI values in a triplet, while X-axis represents the difference between second and first NDVI values in a triplet. Triplets with up to 7 days difference are shown in blue, and triplets from 7 to 14 days are shown in green. The points structure forms a rhombus shape with a small cloud of possible outliers in the upper left corner. To filter outliers from the list of actual mowing events, we only consider triplets within up to 10 days interval (as the mowing event signature can recover in 10 days). Knowing rhombus equation (the centre is approximately in (0, 0), and the side length is around 0.6), the filtering rule can be easily applied as follows:$$begin{aligned} ndvi_3 – 2 cdot ndvi_2 + ndvi_1 ge 0.6 end{aligned}$$
    (3)
    where ndvi_1, ndvi_2, and ndvi_3 are consecutive NDVI measurements within 10 days interval.All outliers are removed, which represent around 0.1% of NDVI measurements.Figure 6Scatter plot of NDVI triplets.Full size imageSmoothing is an essential pre-processing step for noisy features. In this work, cohvh and cohvv features are smoothed using different techniques, including exponential moving average (EMA), moving average54, and Kalman filter55. Smoothing using moving average is done by taking the averages of raw data sequences. The length of the sequence over which we take the average is called the filter width. Table 1 shows the performance of moving average smoothing technique using different values for the filter width. The results show that the best AUC-ROC of 0.9671 is achieved at a filter size of 7. The Kalman filter produces estimates of the current state variables and their uncertainties. Once the outcome of the subsequent measurement is observed, these estimates are updated using a weighted average, giving more weight to estimates with higher certainty. The AUC-ROC achieved using Kalman filter is 0.962. The EMA is done by taking averages of sequences of data, in addition to assigning weights to every data point. More specifically, as values get older, they are given exponentially decreasing weights. The smoothed cohvh and cohvv EMA for cohvh and cohvv are calculated using a recursive definition (i.e., from its previous value) as follows:$$begin{aligned}&cohvh_sm(cohvh_n, alpha ) = alpha cdot (cohvh_n) + (1 – alpha ) cdot cohvh_sm(cohvh_{n-1}, alpha ) end{aligned}$$
    (4)
    $$begin{aligned}&cohvv_sm(cohvv_n, alpha ) = alpha cdot (cohvv_n) + (1 – alpha ) cdot cohvv_sm(cohvv_{n-1}, alpha ) end{aligned}$$
    (5)
    where (cohvh_sm(cohvh_{n-1}, alpha )): exponential moving average for end of (cohvh_{n-1}). (cohvv_sm(cohvv_{n-1}, alpha )): exponential moving average for end of (cohvv_{n-1}). (alpha ): a smoothing parameter.The higher the smoothing parameter, the more it reacts to fluctuations in the original signal. The lower the smoothing parameter, the more the signal is smoothed. Experimentally, we found that the best value for (alpha ) to achieve the best AUC-ROC of 0.968 is (frac{1}{3}) as shown in Table 2. The different smoothing techniques achieve comparable performance. EMA technique was selected as it achieves slightly higher performance.Table 1 Performance of moving average smoothing using different filter width.Full size tableTable 2 Performance of EMA smoothing using different values of (alpha ).Full size tableDerived featuresNew derived features from S-1 and S-2 are extracted to improve the performance of the machine learning model. The features were derived based on the following knowledge about mowing events: coherence tends to increase. In contrast, ndvi tends to decrease after mowing events and, many farmers perform mowing during the same time of the year due to the good weather conditions. Such knowledge was elaborated with the derived features. In the following, we will go through the list of derived features considered in this study. Mixed coherence is derived from S-1 features to capture the overall coherence trend. Mixed coherence is a non-linear combination of cohvh and cohvv and is calculated as follows:$$begin{aligned} Mixed_coh = sqrt{cohvh cdot cohvv} end{aligned}$$
    (6)
    The date is an important feature for the model to adapt, as it is more likely to have mowing events in the summer rather than in early spring, especially in Estonia. The normalized day of the year is calculated as normalization improves the training process of the neural network. Some methods normalize features during the training process, such as Batch Normalization used in this study56. However, neighbouring batches could have entirely different normalization variables (batch mean and variance). At the same time, DOY is a feature susceptible to small changes, e.g., mowing prediction on day 108 or 109 could have drastically different meaning (weekend or working day, day with sunny weather or day with heavy rain). It implies that unified normalization of the DOY feature before training could help avoid the unwanted impact of Batch normalization and possible gradient computation issues. The normalized day of the year is calculated as follows:$$begin{aligned} t = frac{day_of_year}{365} end{aligned}$$
    (7)
    where (day_of_year) is the year’s day, which is a number between 1 and 365, January 1st is day 1.In addition, we use another time feature dt to capture the gaps in time series. dt is defined to be the normalized difference in days between the current measurement and the previous one. Normalization was performed with min-max scaling. dt is calculated as follows:$$begin{aligned} dt = frac{diff – min_diff}{max_diff – min_diff} end{aligned}$$
    (8)
    where (min_diff): the minimum difference in days between two previous consecutive measurements obtained from training data. (max_diff): the maximum difference in days between two previous consecutive measurements obtained from training data.Since mowing is characterized by an increase in the coherence and decline in the NDVI, it is important to capture the difference in the values of features and/or slopes of the features’ curves. In the following, we summarize the list of original and derived features extracted from Sentinel-1 and Sentinel-2 included in this study.

    ndvi Normalized difference vegetation index, obtained from Sentinel-2.

    cohvv Coherence in VV polarization, Sentinel-1 feature.

    cohvh Coherence in VH polarization, Sentinel-1 feature.

    t Normalized day of the year when the measurement is obtained.

    dt Normalized difference in days between current and previous measurement. The data was interpolated with a daily grid, this feature differentiated between interpolated data and real data by capturing the difference between valid (not interpolated) measurements.

    cohvv_sm Smoothed cohvv with exponential mowing average (with parameter (frac{1}{3})).

    cohvh_sm Smoothed cohvh with exponential moving average (with parameter (frac{1}{3})).

    mixed_coh Harmonic mean of cohvv and cohvh. The harmonic mean is chosen as one of the simplest options of non-linear combination.

    ndvi_diff Difference between current and previous NDVI measurements. This feature captures the decrease in the ndvi, which is highly related to mowing detection.

    cohvv_sm_diff difference between current and previous (cohvv_sm) measurements. This feature captures the increase in the (cohvv_sm), which is highly related to mowing detection.

    cohvh_sm_diff difference between current and previous (cohvh_sm) measurements. This feature captures the increase in the (cohvh_sm), which is highly related to mowing detection.

    ndvi_der The slope of the line between previous and current NDVI values.

    cohvh_sm_der The slope of the line between previous and current (cohvh_sm) values. This feature captures the change in the smoothed cohvh.

    cohvv_sm_der The slope of the line between previous and current (cohvv_sm) values. This feature captures the change in the smoothed cohvv.

    Feature selectionThe permutation feature importance measurement was introduced by Breiman57. The importance of a particular feature is measured by the increase in the model’s prediction error after we permuted the values of this feature, which breaks the relationship between the feature and the outcome. A feature is important if shuffling its values increases the model error and is less important otherwise. The importance of features considered in this study is ranked in Table 3. It is notable from Table 3 that the ordinal features are significantly more important than the derived ones. We used backwards elimination to select the optimal subset of features to be used by the machine learning model. More specifically, we start with all the features and then remove the least significant feature at each iteration, which improves the model’s overall performance. We repeat this until no improvement is observed on the removal of features. Figures 7 and 8 show that the end of season accuracy(EOS) and event accuracy, respectively, for training using a different subsets of the most important features. We refer to (F_{x}-F_{y}) to be the set of important features from feature x to feature y in Table 3. Figure 7 shows that using only ndvi and (mixed_{coh}) achieves EOS of 93%. Increasing the number of the most important features to 3 achieves a comparable performance to the best one, as shown in Fig. 7. The results show that using the ndvi and (mixed_{coh}) achieve around 73% event accuracy while increasing the number of features, the performance declines as shown in Fig. 8. As an outcome of the feature selection process, the developed machine learning model used all the 14 features, shown in Table 3, that achieve the highest combined performance.Table 3 Ranking features based on their performance.Full size tableFigure 7End of season accuracy for different number of features.Full size image
    Figure 8Event-based accuracy for different number of features.Full size image
    Machine learning modelEach record in our dataset represents specific features about a field during one season at a particular time, in addition to the target variable (mown or non mown). In this work, we use a neural network to predict mowing events. We are interested only in observations during the vegetative season, so winter measurements are not included. More specifically, we only include the data in the vegetative season, which is almost the same across all Estonia from April till October (215 days). The dataset is partitioned into 64% for training, 20% for testing and 16% for validation. All training and testing were performed using TensorFlow58 deep learning framework with default parameters. The architecture of the neural network used is shown in Fig. 9. To guarantee a fixed time interval of 1-day, all the missing values in S-1 and S-2 features are interpolated, as shown in Fig. 10. The data is processed in batches of size (64 times 215) (times )14, where 64 is the number of fields considered per patch, 215 is the number of days in the vegetation season in Estonia, 14 is the number of selected features.Figure 9Architecture of the proposed model.Full size imageThe network’s output is a vector of size 215, representing the probability of a mowing event on each day in the vegetation season. The network consists of three one dimension convolution layers. The first and second convolution layers are followed by the Softmax activation function and batch normalization layer, while the third convolution is followed by Sigmoid activation function. The NN hyperparameters required to achieve the model learning process can significantly affect model performance. These hyperparameters include the following56:

    Number of epochs represents how many times you want your algorithm to train on your whole dataset.

    Loss function represents the prediction error of Neural Network.

    Optimizer represents algorithm or method used to change the attributes of the neural network such as weights and learning rate to reduce the loss.

    Activation function is the function through which we pass our weighted suown to have a significant output, namely as a vector of probability or a 0–1 output.

    Learning rate refers to the step of backpropagation, when parameters are updated according to an optimization function.

    Figure 10Time series mowing events before and after linear interpolation.Full size imageA good model uses the optimal combination of these hyperparameters and achieves good generalization capability. The training was performed with the conjugate gradient descent method and the binary cross-entropy loss function. The neural network was trained during 300 epochs; an early stopping was used59. The optimizer used in our model is Nadam optimizer60 with the following parameters: (beta_1=0.9), (beta_2=0.999), (epsilon=None), (schedule_{decay}=0.004), and learning (rate=0.001). Different activation functions such as ReLU, Sigmoid, Linear, and Tanh have been experimentally evaluated on the testing dataset as shown in Fig. 11. The results show that the Softmax activation function achieves the highest combined performance (event accuracy of 72.6% and EOS of 94.5%), as shown in Fig. 11.Figure 11Performance of different activation functions.Full size imageUsing 1D convolution layer acts as a filter that slides on the time dimension allowing the model to predict future mowing events from past events. However, this approach is not suitable for real-time detection of mowing events, but we use it to predict mowing events within a fixed time frame (window). Such a time frame should be greater than half the (1-D) convolution window length.Model evaluationTo evaluate our model, we used two metrics, EOS accuracy and Event-based accuracy. EOS is the accuracy of detecting a mowing event at least once during the season. If the probability of detecting a mowing event at least once during the season is more than 50%, then the field is considered mown, otherwise not mown. Event-based accuracy is used to evaluate how well our model correctly predicts mowing events. The formula for quantifying the binary accuracy is defined as follows:$$begin{aligned} acc = frac{TP + TN}{TP + TN + FP + FN} end{aligned}$$
    (9)
    where TP is the number of times that the model correctly predicted mowing events, given that the start day of the predicted mowing event is not more than 3 days earlier and not more than 6 days later than the actual start day of the mowing event. Within these 9 days, several mowing events may be predicted. To handle this case, only the first predicted mowing event is considered TP, and every next one is considered an FP. TN is the number of times that the model correctly predicted the absence of mowing events. FP is the number of times that the model incorrectly predicted mowing events. It also includes the number of times that the model correctly predicted mowing events, but the start of the event does not fit into a 9-days time frame with the actual start of some mowing event. FN is the number of times where the model missed actual mowing events.Reject region
    Figure 12Calibration plot for proposed model.Full size image
    Sometimes the model is not confident enough to give a reliable decision about the state of the field. We cannot expect reliable and confident predictions from inaccurate, incomplete or uncertain data. So, it is better in the cases of uncertainty about the prediction to allow the model to abstain from prediction. In this way, the obtained predictions are more accurate, while human experts could check rejected fields. Given the true positive rate and the true negative rate on the validation set, the reject region technique outputs a probability interval ((t_{low}), (t_{upper})) in which the model abstain prediction, where (t_{low}) and (t_{upper}) are the minimum and maximum probabilities that the model is uncertain about its prediction. Out of this interval, the model is confident about its prediction and predicts afield as mowed if the probability is higher than (t_{upper}) and not mown if the probability is less than (t_{low}). We select (t_{upper}), such that the desired true positive rate is reached. To find (t_{upper}), we sort all positives descending by their predicted probabilities and select the top percentage equal to the true positive rate. We choose (t_{low}) such that the desired true negative rate on validation data is reached. To find (t_{low}), we sort all negatives ascending by their predicted probabilities and select the top percentage equal to the true negative rate.Figure 12 shows the calibration plot for our proposed model. Notably, the predicted probabilities are close to the diagonal, which implies that the model is well-calibrated. More

  • in

    Competition and resource depletion shape the thermal response of population fitness in Aedes aegypti

    1.Mordecai, E. A., Ryan, S. J., Caldwell, J. M., Shah, M. M. & LaBeaud, A. D. Climate change could shift disease burden from malaria to arboviruses in Africa. Lancet Planet. Health 4, e416–e423 (2020).PubMed 
    PubMed Central 

    Google Scholar 
    2.W. H. O. Multisectoral approach to the prevention and control of vector-borne diseases (2020).3.Ryan, S. J. et al. Warming temperatures could expose more than 1.3 billion new people to Zika virus risk by 2050. Glob. Change Biol. 27, 84–93 (2021).
    Google Scholar 
    4.Iwamura, T., Guzman-Holst, A. & Murray, K. A. Accelerating invasion potential of disease vector Aedes aegypti under climate change. Nat. Commun. 11, 2130 (2020).CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    5.Savage, V. M., Gillooly, J. F., Brown, J. H., West, G. B. & Charnov, E. L. Effects of body size and temperature on population growth. Am. Nat. 163, 429–441 (2004).PubMed 

    Google Scholar 
    6.Shocket, M. S. et al. Transmission of West Nile and five other temperate mosquito-borne viruses peaks at temperatures between 23 °C and 26 °C. eLife 9, 1–67 (2020).
    Google Scholar 
    7.Couret, J., Dotson, E. & Benedict, M. Q. Temperature, larval diet, and density effects on development rate and survival of Aedes aegypti (Diptera: Culicidae). PLoS ONE 9, 1–9 (2014).
    Google Scholar 
    8.Barreaux, A. M. G., Stone, C. M., Barreaux, P. & Koella, J. C. The relationship between size and longevity of the malaria vector Anopheles gambiae (s.s.) depends on the larval environment. Parasites Vectors 11, 485 (2018).PubMed 
    PubMed Central 

    Google Scholar 
    9.Huxley, P. J., Murray, K. A., Pawar, S. & Cator, L. J. The effect of resource limitation on the temperature dependence of mosquito population fitness. Proc. R. Soc. B: Biol. Sci. 288, rspb.2020.3217 (2021).10.Ostfeld, R. S. & Keesing, F. Pulsed resources and community dynamics of consumers in terrestrial ecosystems. Trends Ecol. Evol. 15, 232–237 (2000).CAS 
    PubMed 

    Google Scholar 
    11.Beltran, R. S. et al. Seasonal resource pulses and the foraging depth of a Southern Ocean top predator. Proc. R. Soc. B: Biol. Sci. 288, rspb.2020.2817 (2021).12.Yang, L. H., Bastow, J. L., Spence, K. O. & Wright, A. N. What can we learn from resource pulses? Ecology 89, 621–634 (2008).PubMed 

    Google Scholar 
    13.Dye, C. Models for the population dynamics of the yellow fever mosquito, Aedes aegypti. J. Animal Ecol. 53, 247 (1984).
    Google Scholar 
    14.Southwood, T. R., Murdie, G., Yasuno, M., Tonn, R. J. & Reader, P. M. Studies on the life budget of Aedes aegypti in Wat Samphaya, Bangkok, Thailand. Bull. World Health Organ. 46, 211–226 (1972).CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    15.Arrivillaga, J. & Barrera, R. Food as a limiting factor for Aedes aegypti in water-storage containers. J. Vector Ecol. 29, 11–20 (2004).PubMed 

    Google Scholar 
    16.Barrera, R., Amador, M. & Clark, G. G. Ecological factors influencing Aedes aegypti (Diptera: Culicidae) productivity in artificial containers in Salinas, Puerto Rico. J. Med. Entomol. 43, 484–492 (2006).PubMed 

    Google Scholar 
    17.Yee, D. A. & Juliano, S. A. Concurrent effects of resource pulse amount, type, and frequency on community and population properties of consumers in detritus-based systems. Oecologia 169, 511–522 (2012).PubMed 

    Google Scholar 
    18.Subra, R. & Mouchet, J. The regulation of preimaginal populations of Aedes aegypti (L.) (Diptera: Culicidae) on the Kenya coast. Ann. Trop. Med. Parasitol. 78, 63–70 (1984).CAS 
    PubMed 

    Google Scholar 
    19.Amarasekare, P. & Savage, V. A framework for elucidating the temperature dependence of fitness. Am. Nat. 179, 178–191 (2012).PubMed 

    Google Scholar 
    20.Huey, R. B. & Kingsolver, J. G. Climate warming, resource availability, and the metabolic meltdown of ectotherms. Am. Nat. 194, 6 (2019).21.García-Carreras, B. et al. Role of carbon allocation efficiency in the temperature dependence of autotroph growth rates. Proc. Natl Acad. Sci. USA 115, E7361–E7368 (2018).PubMed 
    PubMed Central 

    Google Scholar 
    22.Smith, T. P., Clegg, T., Bell, T. & Pawar, S. Systematic variation in the temperature dependence of bacterial carbon use efficiency. Ecol. Lett. 24, 2123–2133 (2021).PubMed 

    Google Scholar 
    23.Lehmann, P. et al. Complex responses of global insect pests to climate warming. Front. Ecol. Environ. 18, 141–150 (2020).
    Google Scholar 
    24.Amarasekare, P. Effects of climate warming on consumer-resource interactions: a latitudinal perspective. Front. Ecol. Evol. 7, 1–15 (2019).25.Amarasekare, P. & Simon, M. W. Latitudinal directionality in ectotherm invasion success. Proc. R. Soc. B: Biol. Sci. 287, 20191411 (2020).
    Google Scholar 
    26.Diagne, C. et al. High and rising economic costs of biological invasions worldwide. Nature 592, 571–576 (2021).CAS 
    PubMed 

    Google Scholar 
    27.Cross, W. F., Hood, J. M., Benstead, J. P., Huryn, A. D. & Nelson, D. Interactions between temperature and nutrients across levels of ecological organization. Glob. Change Biol. 21, 1025–1040 (2015).
    Google Scholar 
    28.Mordecai, E. A. et al. Thermal biology of mosquito‐borne disease. Ecol. Lett. 22, 1690–1708 (2019).PubMed 
    PubMed Central 

    Google Scholar 
    29.Thomas, M. K. et al. Temperature-nutrient interactions exacerbate sensitivity to warming in phytoplankton. Glob. Change Biol. 23, 3269–3280 (2017).
    Google Scholar 
    30.Siegel, P., Baker, K. G., Low‐Décarie, E. & Geider, R. J. High predictability of direct competition between marine diatoms under different temperatures and nutrient states. Ecol. Evol. 10, 7276–7290 (2020).PubMed 
    PubMed Central 

    Google Scholar 
    31.Bestion, E., García-Carreras, B., Schaum, C.-E., Pawar, S. & Yvon-Durocher, G. Metabolic traits predict the effects of warming on phytoplankton competition. Ecol. Lett. 21, 655–664 (2018).PubMed 
    PubMed Central 

    Google Scholar 
    32.Jackson, C. flexsurv: A Platform for Parametric Survival Modeling in R. J. Stat. Softw. 70, 1–33 (2016).
    Google Scholar 
    33.Bellows, T. S. The descriptive properties of some models for density dependence. J. Animal Ecol. 50, 139–156 (1981).
    Google Scholar 
    34.Orcutt, J. D. & Porter, K. G. The synergistic effects of temperature and food concentration of life history parameters of Daphnia. Oecologia 63, 300–306 (1984).PubMed 

    Google Scholar 
    35.Huey, R. B. & Berrigan, D. Temperature, demography, and ectotherm fitness. Am. Nat. 158, 204–210 (2001).CAS 
    PubMed 

    Google Scholar 
    36.Caswell, H. A general formula for the sensitivity of population growth rate to changes in life history parameters. Theor. Popul. Biol. 14, 215–230 (1978).CAS 
    PubMed 

    Google Scholar 
    37.Kammenga, J. E., Busschers, M., Straalen, N. M., Van, Jepson, P. C. & Bakker, J. Stress induced fitness reduction is not determined by the most sensitive life-cycle trait. Funct. Ecol. 10, 106 (1996).
    Google Scholar 
    38.Cator, L. J. et al. The role of vector trait variation in vector-borne disease dynamics. Front. Ecol. Evol. 8, 1–25 (2020).
    Google Scholar 
    39.Juliano, S. A. Species introduction and replacement among mosquitoes: interspecific resource competition or apparent competition? Ecology 79, 255 (1998).
    Google Scholar 
    40.Shapiro, L. L. M., Murdock, C. C., Jacobs, G. R., Thomas, R. J. & Thomas, M. B. Larval food quantity affects the capacity of adult mosquitoes to transmit human malaria. Proc. R. Soc. B: Biol. Sci. 283, 20160298 (2016).
    Google Scholar 
    41.Carvajal-Lago, L., Ruiz-López, M. J., Figuerola, J. & Martínez-de la Puente, J. Implications of diet on mosquito life history traits and pathogen transmission. Environ. Res. 195, 110893 (2021).CAS 
    PubMed 

    Google Scholar 
    42.Reiner, R. C. et al. A systematic review of mathematical models of mosquito-borne pathogen transmission: 1970-2010. J. R. Soc. Interface 10, 20120921–20120921 (2013).PubMed 
    PubMed Central 

    Google Scholar 
    43.Farjana, T., Tuno, N. & Higa, Y. Effects of temperature and diet on development and interspecies competition in Aedes aegypti and Aedes albopictus. Med.Vet. Entomol. 26, 210–217 (2012).CAS 
    PubMed 

    Google Scholar 
    44.Kooijman, S. A. L. M. Dynamic energy and mass budgets in biological systems. (Cambridge University Press, 2000).45.Merritt, R. W., Dadd, R. H. & Walker, E. D. Feeding behaviour, natural food, and nutritional relationships and larval mosquitoes. Annu. Rev. Entomol. 37, 349–376 (1992).46.Craine, J. M., Fierer, N. & McLauchlan, K. K. Widespread coupling between the rate and temperature sensitivity of organic matter decay. Nat. Geosci. 3, 854–857 (2010).CAS 

    Google Scholar 
    47.Smith, T. P. et al. Community-level respiration of prokaryotic microbes may rise with global warming. Nat. Commun. 10, 5124 (2019).PubMed 
    PubMed Central 

    Google Scholar 
    48.Yee, D. A., Kaufman, M. G. & Juliano, S. A. The significance of ratios of detritus types and micro-organism productivity to competitive interactions between aquatic insect detritivores. J. Animal Ecol. 76, 1105–1115 (2007).
    Google Scholar 
    49.Chouaia, B. et al. Delayed larval development in Anopheles mosquitoes deprived of Asaia bacterial symbionts. BMC Microbiol. 12, S2 (2012).CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    50.Souza, R. S. et al. Microorganism-based larval diets affect mosquito development, size and nutritional reserves in the yellow fever mosquito Aedes aegypti (Diptera: Culicidae). Front. Physiol. 10, 1–24 (2019).
    Google Scholar 
    51.Dickson, L. B. et al. Carryover effects of larval exposure to different environmental bacteria drive adult trait variation in a mosquito vector. Sci. Adv. 3, e1700585 (2017).PubMed 
    PubMed Central 

    Google Scholar 
    52.Hery, L. et al. Natural variation in physicochemical profiles and bacterial communities associated with Aedes aegypti breeding sites and larvae on Guadeloupe and French Guiana. Microbial Ecol. 81, 93–109 (2021).CAS 

    Google Scholar 
    53.Liikanen, A., Murtoniemi, T., Tanskanen, H., Väisänen, T. & Martikainen, P. J. Effects of temperature and oxygen availability on greenhouse gas and nutrient dynamics in sediment of a eutrophic mid-boreal lake. Biogeochemistry 59, 269–286 (2002).CAS 

    Google Scholar 
    54.Lister, B. C. & Garcia, A. Climate-driven declines in arthropod abundance restructure a rainforest food web. Proc. Natl Acad. Sci. USA 115, E10397–E10406 (2018).CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    55.Du, E. et al. Global patterns of terrestrial nitrogen and phosphorus limitation. Nat. Geosci. 13, 221–226 (2020).CAS 

    Google Scholar 
    56.Briegel, H. Metabolic relationship between female body size, reserves, and fecundity of Aedes aegypti. J. Insect Physiol. 36, 165–172 (1990).
    Google Scholar 
    57.Steinwascher, K. Relationship between pupal mass and adult survivorship and fecundity for Aedes aegypti. Environ. Entomol. 11, 150–153 (1982).
    Google Scholar 
    58.Trisos, C. H., Merow, C. & Pigot, A. L. The projected timing of abrupt ecological disruption from climate change. Nature 580, 496–501 (2020).CAS 
    PubMed 

    Google Scholar 
    59.Parmesan, C. Ecological and evolutionary responses to recent climate change. Annu. Rev. Ecol., Evol. Syst. 37, 637–669 (2006).
    Google Scholar 
    60.Taheri, S., Naimi, B., Rahbek, C. & Araújo, M. B. Improvements in reports of species redistribution under climate change are required. Sci. Adv. 7, eabe1110 (2021).PubMed 
    PubMed Central 

    Google Scholar 
    61.Bargielowski, I. E., Lounibos, L. P. & Carrasquilla, M. C. Evolution of resistance to satyrization through reproductive character displacement in populations of invasive dengue vectors. Proc. Natl Acad. Sci. USA 110, 2888–2892 (2013).CAS 
    PubMed 
    PubMed Central 

    Google Scholar 
    62.Arguez, A. et al. NOAA’s 1981–2010 U.S. climate normals: an overview. Bull. Am. Meteorol. Soc. 93, 1687–1697 (2012).
    Google Scholar 
    63.Caswell, H. Matrix population models construction, analysis, and interpretation. Nat. Resource Model. (Sinauer Associates, 1989).64.Birch, L. C. The intrinsic rate of natural increase of an insect population. J. Animal Ecol. 17, 15 (1948).
    Google Scholar 
    65.Cole, L. C. The population consequences of life history phenomena. Q. Rev. Biol. 29, 103–137 (1954).CAS 
    PubMed 

    Google Scholar 
    66.R. Core Team. R: A language and environment for statistical computing. (2018).67.Stubben, C. & Milligan, B. Estimating and analyzing demographic models using the popbio Package in R. J. Stat. Softw. 22, 1–23 (2007).
    Google Scholar 
    68.Therneau, T. A Package for Survival Analysis in R. (2021).69.Agnew, P., Hide, M., Sidobre, C. & Michalakis, Y. A minimalist approach to the effects of density-dependent competition on insect life-history traits. Ecol. Entomol. 27, 396–402 (2002).
    Google Scholar 
    70.Honěk, A. Intraspecific variation in body size and fecundity in insects: a general relationship. Oikos 66, 483 (1993).
    Google Scholar 
    71.Livdahl, T. P. & Sugihara, G. Non-linear interactions of populations and the importance of estimating per capita rates of change. J. Animal Ecol. 53, 573 (1984).
    Google Scholar 
    72.Juliano, S. A. & Lounibos, L. P. Ecology of invasive mosquitoes: effects on resident species and on human health. Ecol. Lett. 8, 558–574 (2005).PubMed 
    PubMed Central 

    Google Scholar 
    73.van den Heuvel, M. J. The effect of rearing temperature on the wing length, thorax length, leg length and ovariole number of the adult mosquito, Aedes aegypti (L.). Trans. R. Entomol. Soc. Lond. 115, 197–216 (1963).
    Google Scholar 
    74.Farjana, T. & Tuno, N. Effect of body size on multiple blood feeding and egg retention of Aedes aegypti (L.) and Aedes albopictus (Skuse) (Diptera: Culicidae). Med. Entomol. Zool. 63, 123–131 (2012).
    Google Scholar 
    75.Skalski, J. R., Millspaugh, J. J., Dillingham, P. & Buchanan, R. A. Calculating the variance of the finite rate of population change from a matrix model in Mathematica. Environ. Model. Softw. 22, 359–364 (2007).
    Google Scholar 
    76.Hope, R. M. Rmisc: Rmisc: Ryan Miscellaneous. (2013).77.Caswell, H., Naiman, R. J. & Morin, R. Evaluating the consequences of reproduction in complex salmonid life cycles. Aquaculture 43, 123–134 (1984).
    Google Scholar 
    78.de Kroon, H., Plaisier, A., van Groenendael, J. & Caswell, H. Elasticity: the relative contribution of demographic parameters to population growth rate. Ecology 67, 1427–1431 (1986).
    Google Scholar 
    79.Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. J. Statist. Softw. 67, (2015).80.Padfield, D., O’Sullivan, H. & Pawar, S. rTPC and nls.multstart: a new pipeline to fit thermal performance curves in R. Methods Ecol. Evol. 12, 1138–1143 (2021).
    Google Scholar 
    81.Lactin, D. J., Holliday, N. J., Johnson, D. L. & Craigen, R. Improved rate model of temperature-dependent development by arthropods. Environ. Entomol. 24, 68–75 (1995).
    Google Scholar 
    82.Kamykowski, D. & McCollum, S. A. The temperature acclimatized swimming speed of selected marine dinoflagellates. J. Plankton Res. 8, 275–287 (1986).
    Google Scholar  More

  • in

    Insect visual sensitivity to long wavelengths enhances colour contrast of insects against vegetation

    1.Cummings, M. E., Rosenthal, G. G. & Ryan, M. J. A private ultraviolet channel in visual communication. Proc. R. Soc. B-Biol. Sci. 270, 897–904. https://doi.org/10.1098/rspb.2003.2334 (2003).Article 

    Google Scholar 
    2.Tedore, C. & Nilsson, D. E. Avian UV vision enhances leaf surface contrasts in forest environments. Nat. Commun. https://doi.org/10.1038/s41467-018-08142-5 (2019).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    3.Qi, Y. D., Bai, S. J. & Heisler, G. M. Changes in ultraviolet-B and visible optical properties and absorbing pigment concentrations in pecan leaves during a growing season. Agric. For. Meteorol. 120, 229–240. https://doi.org/10.1016/j.agrformet.2003.08.018 (2003).ADS 
    Article 

    Google Scholar 
    4.Mollon, J. D. “Tho’ she kneel’d in that place where they grew…” The uses and origins of primate colour vision. J. Exp. Biol. 146, 21–38. https://doi.org/10.1242/jeb.146.1.21 (1989).ADS 
    CAS 
    Article 
    PubMed 

    Google Scholar 
    5.Osorio, D. & Vorobyev, M. Photoreceptor sectral sensitivities in terrestrial animals: Adaptations for luminance and colour vision. Proc. R. Soc. B Biol. Sci. 272, 1745–1752. https://doi.org/10.1098/rspb.2005.3156 (2005).CAS 
    Article 

    Google Scholar 
    6.Osorio, D. & Vorobyev, M. A review of the evolution of animal colour vision and visual communication signals. Vis. Res. 48, 2042–2051. https://doi.org/10.1016/j.visres.2008.06.018 (2008).CAS 
    Article 
    PubMed 

    Google Scholar 
    7.Bowmaker, J. K. & Dartnall, H. J. A. Visual pigments of rods and cones in a human retina. J. Physiol. Lond. 298, 501–511. https://doi.org/10.1113/jphysiol.1980.sp013097 (1980).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    8.Bowmaker, J. K. & Hunt, D. M. Evolution of vertebrate visual pigments. Curr. Biol. 16, R484–R489. https://doi.org/10.1016/j.cub.2006.06.016 (2006).CAS 
    Article 
    PubMed 

    Google Scholar 
    9.van der Kooi, C. J., Stavenga, D. G., Arikawa, K., Belušič, G. & Kelber, A. Evolution of insect color vision: From spectral sensitivity to visual ecology. Annu. Rev. Entomol. 66, 435–461 (2021).Article 

    Google Scholar 
    10.Briscoe, A. D. & Chittka, L. The evolution of color vision in insects. Annu. Rev. Entomol. 46, 471–510. https://doi.org/10.1146/annurev.ento.46.1.471 (2001).CAS 
    Article 
    PubMed 

    Google Scholar 
    11.Ogawa, Y., Kinoshita, M., Stavenga, D. G. & Arikawa, K. Sex-specific retinal pigmentation results in sexually dimorphic long-wavelength-sensitive photoreceptors in the eastern pale clouded yellow butterfly, Colias erate. J. Exp. Biol. 216, 1916–1923. https://doi.org/10.1242/jeb.083485 (2013).CAS 
    Article 
    PubMed 

    Google Scholar 
    12.Kelber, A. Ovipositing butterflies use a red receptor to see green. J. Exp. Biol. 202, 2619–2630 (1999).Article 

    Google Scholar 
    13.Osorio, D. & Vorobyev, M. Colour vision as an adaptation to frugivory in primates. Proc. R. Soc. B Biol. Sci. 263, 593–599. https://doi.org/10.1098/rspb.1996.0089 (1996).ADS 
    CAS 
    Article 

    Google Scholar 
    14.Zaccardi, G., Kelber, A., Sison-Mangus, M. P. & Briscoe, A. D. Color discrimination in the red range with only one long-wavelength sensitive opsin. J. Exp. Biol. 209, 1944–1955. https://doi.org/10.1242/jeb.02207 (2006).Article 
    PubMed 

    Google Scholar 
    15.Wakakuwa, M., Stavenga, D. G., Kurasawa, M. & Arikawa, K. A unique visual pigment expressed in green, red and deep-red receptors in the eye of the small white butterfly, Pieris rapae crucivora. J. Exp. Biol. 207, 2803–2810. https://doi.org/10.1242/jeb.01078 (2004).CAS 
    Article 
    PubMed 

    Google Scholar 
    16.Satoh, A. et al. Red-shift of spectral sensitivity due to screening pigment migration in the eyes of a moth, Adoxophyes orana. Zool. Lett. https://doi.org/10.1186/s40851-017-0075-6 (2017).Article 

    Google Scholar 
    17.Pirih, P. et al. The giant butterfly-moth Paysandisia archon has spectrally rich apposition eyes with unique light-dependent photoreceptor dynamics. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 204, 639–651. https://doi.org/10.1007/s00359-018-1267-z (2018).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    18.Cronin, T. W., Jarvilehto, M., Weckstrom, M. & Lall, A. B. Tuning of photoreceptor spectral sensitivity in fireflies (Coleoptera: Lampyridae). J. Comp. Physiol. A Sens. Neural Behav. Physiol. 186, 1–12. https://doi.org/10.1007/s003590050001 (2000).CAS 
    Article 

    Google Scholar 
    19.Lall, A. B. et al. Vision in click beetles (Coleoptera: Elateridae): pigments and spectral correspondence between visual sensitivity and species bioluminescence emission. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 196, 629–638. https://doi.org/10.1007/s00359-010-0549-x (2010).Article 
    PubMed 

    Google Scholar 
    20.Frentiu, F. D. et al. Adaptive evolution of color vision as seen through the eyes of butterflies. Proc. Natl. Acad. Sci. U.S.A. 104, 8634–8640. https://doi.org/10.1073/pnas.0701447104 (2007).ADS 
    CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    21.Liénard, M. A. et al. The evolution of red color vision is linked to coordinated rhodopsin tuning in lycaenid butterflies. Proc. Natl. Acad. Sci. U. S. A. https://doi.org/10.1073/pnas.2008986118 (2021).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    22.Saito, T. et al. Spectral tuning mediated by helix III in butterfly long wavelength-sensitive visual opsins revealed by heterologous action spectroscopy. Zool. Lett. https://doi.org/10.1186/s40851-019-0150-2 (2019).Article 

    Google Scholar 
    23.Enright, J. M. et al. Cyp27c1 red-shifts the spectral sensitivity of photoreceptors by converting vitamin A1 into A2. Curr. Biol. 25, 3048–3057. https://doi.org/10.1016/j.cub.2015.10.018 (2015).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    24.Martin, M., Le Galliard, J. F., Meylan, S. & Loew, E. R. The importance of ultraviolet and near-infrared sensitivity for visual discrimination in two species of lacertid lizards. J. Exp. Biol. 218, 458–465. https://doi.org/10.1242/jeb.115923 (2015).Article 
    PubMed 

    Google Scholar 
    25.Ala-Laurila, P., Donner, K. & Koskelainen, A. Thermal activation and photoactivation of visual pigments. Biophys. J. 86, 3653–3662. https://doi.org/10.1529/biophysj.103.035626 (2004).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    26.Ala-Laurila, P., Pahlberg, J., Koskelainen, A. & Donner, K. On the relation between the photoactivation energy and the absorbance spectrum of visual pigments. Vis. Res. 44, 2153–2158. https://doi.org/10.1016/j.visres.2004.03.031 (2004).Article 
    PubMed 

    Google Scholar 
    27.Barlow, H. B. Purkinje shift and retinal noise. Nature 179, 255–256. https://doi.org/10.1038/179255b0 (1957).ADS 
    CAS 
    Article 
    PubMed 

    Google Scholar 
    28.Koskelainen, A., Ala-Laurila, P., Fyhrquist, N. & Donner, K. Measurement of thermal contribution to photoreceptor sensitivity. Nature 403, 220–223. https://doi.org/10.1038/35003242 (2000).ADS 
    CAS 
    Article 
    PubMed 

    Google Scholar 
    29.Luo, D. G., Yue, W. W. S., Ala-Laurila, P. & Yau, K. W. Activation of visual pigments by light and heat. Science 332, 1307–1312. https://doi.org/10.1126/science.1200172 (2011).ADS 
    CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    30.Rieke, F. & Baylor, D. A. Origin and functional impact of dark noise in retinal cones. Neuron 26, 181–186. https://doi.org/10.1016/s0896-6273(00)81148-4 (2000).CAS 
    Article 
    PubMed 

    Google Scholar 
    31.Cronin, T. W., Johnsen, S., Marshall, N. J. & Warrant, E. J. Visual pigments and photoreceptors. In Visual Ecology, pp. 37–65: Princeton University Press.32.Kelber, A., Yovanovich, C. & Olsson, P. Thresholds and noise limitations of colour vision in dim light. Philos. Trans. R. Soc. B-Biol. Sci. https://doi.org/10.1098/rstb.2016.0065 (2017).Article 

    Google Scholar 
    33.Kemp, D. J. et al. An integrative framework for the appraisal of coloration in nature. Am. Nat. 185, 705–724. https://doi.org/10.1086/681021 (2015).Article 
    PubMed 

    Google Scholar 
    34.Hawkeswood, T. Observations on some Buprestidae (Coleoptera) from the Blue Mountains, N.S.W.. Aust. Zool. 19, 257–275 (1978).
    Google Scholar 
    35.Hawkeswood, T. Observations on two sympatric species of Buprestidae (Coleoptera) from sand dunes on the north coast of New South Wales. Victorian Naturalist 98, 146–151 (1981).
    Google Scholar 
    36.Hawkeswood, T. Observations on some jewel beetles (Coleoptera Buprestidae) from the Armidale district, North-eastern New South Wales. Vic. Nat. 98, 152–155 (1981).
    Google Scholar 
    37.Poland, T. M., Chen, Y. G., Koch, J. & Pureswaran, D. Review of the emerald ash borer (Coleoptera: Buprestidae), life history, mating behaviours, host plant selection, and host resistance. Can. Entomol. 147, 252–262. https://doi.org/10.4039/tce.2015.4 (2015).Article 

    Google Scholar 
    38.Bellamy, C. L., Williams, G., Hasenpusch, J. & Sundholm, A. A summary of the published data on host plants and morphology of immature stages of Australian jewel beetles (Coleoptera: Buprestidae), with additional new records. Insecta Mundi, 1–172 (2013).39.Domingue, M. J. et al. Field observations of visual attraction of three European oak buprestid beetles toward conspecific and heterospecific models. Entomol. Exp. Appl. 140, 112–121. https://doi.org/10.1111/j.1570-7458.2011.01139.x (2011).Article 

    Google Scholar 
    40.Domingue, M. J. et al. Differences in spectral selectivity between stages of visually guided mating approaches in a buprestid beetle. J. Exp. Biol. 219, 2837–2843 (2016).PubMed 

    Google Scholar 
    41.Pureswaran, D. S. & Poland, T. M. Effects of visual silhouette, leaf size and host species on feeding preference by adult emerald ash borer, Agrilus planipennis Fairmaire (Coleoptera: Buprestidae). Great Lakes Entomol. 42, 4 (2018).
    Google Scholar 
    42.Crook, D. J. et al. Laboratory and field response of the emerald ash borer (Coleoptera: Buprestidae), to selected regions of the electromagnetic spectrum. J. Econ. Entomol. 102, 2160–2169 (2009).Article 

    Google Scholar 
    43.Lord, N. P. et al. A cure for the blues: Opsin duplication and subfunctionalization for short-wavelength sensitivity in jewel beetles (Coleoptera: Buprestidae). BMC Evol. Biol. 16, 107 (2016).Article 

    Google Scholar 
    44.Meglič, A., Ilić, M., Quero, C., Arikawa, K. & Belušič, G. Two chiral types of randomly rotated ommatidia are distributed across the retina of the flathead oak borer Coraebus undatus (Coleoptera: Buprestidae). J. Exp. Biol. 223, jeb225920. https://doi.org/10.1242/jeb.225920 (2020).Article 
    PubMed 

    Google Scholar 
    45.Chen, Y. G. & Poland, T. M. Biotic and abiotic factors affect green ash volatile production and emerald Ash borer adult feeding preference. Environ. Entomol. 38, 1756–1764. https://doi.org/10.1603/022.038.0629 (2009).CAS 
    Article 
    PubMed 

    Google Scholar 
    46.Govardovskii, V. I., Fyhrquist, N., Reuter, T., Kuzmin, D. G. & Donner, K. In search of the visual pigment template. Vis. Neurosci. 17, 509–528. https://doi.org/10.1017/s0952523800174036 (2000).CAS 
    Article 
    PubMed 

    Google Scholar 
    47.Dartnall, H. J. A. Visual pigment. Trans. Zool. Soc. Lond. 33, 147–152. https://doi.org/10.1111/j.1096-3642.1976.tb00047.x (1976).Article 

    Google Scholar 
    48.Arikawa, K., Scholten, D. G. W., Kinoshita, M. & Stavenga, D. G. Tuning of photoreceptor spectral sensitivities by red and yellow pigments in the butterfly Papilio xuthus. Zool. Sci. 16, 17–24. https://doi.org/10.2108/zsj.16.17 (1999).Article 

    Google Scholar 
    49.Das, D., Wilkie, S. E., Hunt, D. M. & Bowmaker, J. K. Visual pigments and oil droplets in the retina of a passerine bird, the canary Serinus canaria: microspectrophotometry and opsin sequences. Vision. Res. 39, 2801–2815. https://doi.org/10.1016/s0042-6989(99)00023-1 (1999).CAS 
    Article 
    PubMed 

    Google Scholar 
    50.Sison-Mangus, M. P., Bernard, G. D., Lampel, J. & Briscoe, A. D. Beauty in the eye of the beholder: The two blue opsins of lycaenid butterflies and the opsin gene-driven evolution of sexually dimorphic eyes. J. Exp. Biol. 209, 3079–3090. https://doi.org/10.1242/jeb.02360 (2006).CAS 
    Article 
    PubMed 

    Google Scholar 
    51.Bernard, G. D. Red-absorbing visual pigment of butterflies. Science 203, 1125. https://doi.org/10.1126/science.203.4385.1125 (1979).ADS 
    CAS 
    Article 
    PubMed 

    Google Scholar 
    52.Martínez-Harms, J. et al. Evidence of red sensitive photoreceptors in Pygopleurus israelitus (Glaphyridae: Coleoptera) and its implications for beetle pollination in the southeast Mediterranean. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 198, 451–463. https://doi.org/10.1007/s00359-012-0722-5 (2012).CAS 
    Article 
    PubMed 

    Google Scholar 
    53.Stavenga, D. G. & Arikawa, K. Photoreceptor spectral sensitivities of the Small White butterfly Pieris rapae crucivora interpreted with optical modeling. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 197, 373–385. https://doi.org/10.1007/s00359-010-0622-5 (2011).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    54.Vorobyev, M., Osorio, D., Bennett, A. T. D., Marshall, N. J. & Cuthill, I. C. Tetrachromacy, oil droplets and bird plumage colours. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 183, 621–633. https://doi.org/10.1007/s003590050286 (1998).CAS 
    Article 

    Google Scholar 
    55.Vorobyev, M. & Osorio, D. Receptor noise as a determinant of colour thresholds. Proc. R. Soc. Lond. Ser. B Biol. Sci. 265, 351–358. https://doi.org/10.1098/rspb.1998.0302 (1998).CAS 
    Article 

    Google Scholar 
    56.Vorobyev, M., Brandt, R., Peitsch, D., Laughlin, S. B. & Menzel, R. Colour thresholds and receptor noise: Behaviour and physiology compared. Vis. Res. 41, 639–653. https://doi.org/10.1016/s0042-6989(00)00288-1 (2001).CAS 
    Article 
    PubMed 

    Google Scholar 
    57.Maia, R., Gruson, H., Endler, J. A. & White, T. E. pavo 2: new tools for the spectral and spatial analysis of colour in R. Methods in Ecology and Evolution 10, 1097–1107 (2019).58.Matsushita, A., Awata, H., Wakakuwa, M., Takemura, S. Y. & Arikawa, K. Rhabdom evolution in butterflies: insights from the uniquely tiered and heterogeneous ommatidia of the Glacial Apollo butterfly, Parnassius glacialis. Proc. R. Soc. B Biol. Sci. 279, 3482–3490. https://doi.org/10.1098/rspb.2012.0475 (2012).Article 

    Google Scholar 
    59.McCulloch, K. J. et al. Sexual dimorphism and retinal mosaic diversification following the evolution of a violet receptor in butterflies. Mol. Biol. Evol. 34, 2271–2284. https://doi.org/10.1093/molbev/msx163 (2017).CAS 
    Article 
    PubMed 

    Google Scholar 
    60.Bates, D., Mächler, M., Bolker, B. M. & Walker, S. C. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48. https://doi.org/10.18637/jss.v067.i01 (2015).Article 

    Google Scholar 
    61.R: A Language and environment for statistical computing (R Foundation for Statistical Computing, 2018).62.van der Kooi, C. J., Elzenga, J. T. M., Staal, M. & Stavenga, D. G. How to colour a flower: On the optical principles of flower coloration. Proc. R. Soc. B Biol. Sci. 283, 20160429. https://doi.org/10.1098/rspb.2016.0429 (2016).CAS 
    Article 

    Google Scholar 
    63.Horler, D. N. H., Dockray, M. & Barber, J. The red edge of plant leaf reflectance. Int. J. Remote Sens. 4, 273–288. https://doi.org/10.1080/01431168308948546 (1983).Article 

    Google Scholar 
    64.Silberglied, R. E. Communication in the Ultraviolet. Annu. Rev. Ecol. Syst. 10, 373–398. https://doi.org/10.1146/annurev.es.10.110179.002105 (1979).Article 

    Google Scholar 
    65.Lind, O. Colour vision and background adaptation in a passerine bird, the zebra finch (Taeniopygia guttata). R. Soc. Open Sci. 3, 160383. https://doi.org/10.1098/rsos.160383 (2016).ADS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    66.Santiago, C. et al. Does conspicuousness scale linearly with colour distance? A test using reef fish. Proc. R. Soc. B Biol. Sci. 287, 20201456. https://doi.org/10.1098/rspb.2020.1456 (2020).Article 

    Google Scholar 
    67.Giurfa, M., Vorobyev, M., Brandt, R., Posner, B. & Menzel, R. Discrimination of coloured stimuli by honeybees: Alternative use of achromatic and chromatic signals. J. Comp. Physiol. A. 180, 235–243. https://doi.org/10.1007/s003590050044 (1997).Article 

    Google Scholar 
    68.Garcia, J. E., Spaethe, J. & Dyer, A. G. The path to colour discrimination is S-shaped: Behaviour determines the interpretation of colour models. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 203, 983–997. https://doi.org/10.1007/s00359-017-1208-2 (2017).Article 
    PubMed 

    Google Scholar 
    69.Hart, N. S., Bailes, H. J., Vorobyev, M., Marshall, N. J. & Collin, S. P. Visual ecology of the Australian lungfish (Neoceratodus forsteri). BMC Ecol. 8, 21. https://doi.org/10.1186/1472-6785-8-21 (2008).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    70.Vorobyev, M. Coloured oil droplets enhance colour discrimination. Proc. R. Soc. B Biol. Sci. 270, 1255–1261. https://doi.org/10.1098/rspb.2003.2381 (2003).Article 

    Google Scholar 
    71.Carleton, K. L. et al. Visual sensitivities tuned by heterochronic shifts in opsin gene expression. Bmc Biol. https://doi.org/10.1186/1741-7007-6-22 (2008).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    72.Seki, T. & Vogt, K. Evolutionary aspects of the diversity of visual pigment chromophores in the class Insecta. Comp. Biochem. Physiol. B Biochem. Mol. Biol. 119, 53–64. https://doi.org/10.1016/s0305-0491(97)00322-2 (1998).Article 

    Google Scholar 
    73.Stavenga, D. G., Smits, R. P. & Hoenders, B. J. Simple exponential functions describing the absorbance bands of visual pigment spectra. Vis. Res. 33, 1011–1017. https://doi.org/10.1016/0042-6989(93)90237-q (1993).CAS 
    Article 
    PubMed 

    Google Scholar 
    74.Kinoshita, M. & Arikawa, K. Color and polarization vision in foraging Papilio. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 200, 513–526. https://doi.org/10.1007/s00359-014-0903-5 (2014).Article 
    PubMed 

    Google Scholar 
    75.Vorobyev, M. & Menzel, R. Flower advertisement for insects: Bees, a case study. In Adaptive Mechanisms in the Ecology of Vision (eds S. N. Archer et al.) 537–553 (Springer Netherlands, 1999).76.Bernard, G. D. & Remington, C. L. Color vision in Lycaena butterflies: Spectral tuning of receptor arrays in relation to behavioral ecology. Proc. Natl. Acad. Sci. U.S.A. 88, 2783–2787. https://doi.org/10.1073/pnas.88.7.2783 (1991).ADS 
    CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    77.McCulloch, K. J., Osorio, D. & Briscoe, A. D. Sexual dimorphism in the compound eye of Heliconius erato: A nymphalid butterfly with at least five spectral classes of photoreceptor. J. Exp. Biol. 219, 2377–2387. https://doi.org/10.1242/jeb.136523 (2016).Article 
    PubMed 

    Google Scholar  More

  • in

    Physical simulation study on grouting water plugging of flexible isolation layer in coal seam mining

    Analysis of the roof failure characteristics of coal seamBefore mining, fracturing was conducted on a portion of gritstone in the lower section of the Naoro Formation and then entered the mining stage. Figure 9 shows the influence law of coal roof rupture under different periodic pressures. With mining of the #2 coal seam working face, the direct roof of the coal seam partially broke and collapsed, forming gangue in the goaf. There is a clear separation between the direct and basic roof. When the working face advanced to 228.2 mm, the old roof ruptured, and the working face started to enter the periodic pressure-bearing stage. As the working face advanced to 592.9 mm, the roof exhibited the fourth periodic pressure. The overlying layer roof in the excavation area was affected by the upper bearing arch pressure, leading to the collapsed rock to not completely contact the upper roof. With the increasing distance of coal seam mining, the roof developed significant subsidence, and the influence range of the bedrock boundary caused by the mining was still in the isolation layer fracturing zone. The bedrock influence boundary angle reached 73.57°, and the rock fracture angle was 56.95°. When the working face advanced to 726.5 mm, the fifth periodic pressure on the roof occurred. The bedrock layer in the upper right of the workings was near the right boundary of the first isolated coal seam rupture. Then, coal mining was suspended, and a second isolated seam fracturing process was conducted. The bedrock influence boundary angle reached 73.57°, and the rock rupture angle was 56.95°.Figure 9Influence law of coal roof rupture during different periodic pressure.Full size imageWhen the processing was advanced to 798.4 mm, the bedrock layer in the upper right of the processed area became close to the right boundary of the second isolated seam fracturing. After the third isolated layer fracturing process, the rock impact boundary angle reached 75.33°, and the rock fracture angle was 50.39°. Proceeding to 1031.6 mm, eighth periodic pressure was generated on the roof. The falling gangue in the mined-out area was in contact with the roof, with the bedrock impact boundary angle reaching 74.77° and the rock fracture angle reaching 57.06°. Thereafter, the bedrock layer of the roof gradually entered the full-scale mining stage. As the working face continues to advance, the bedrock impact boundary caused by coal seam mining should be in isolated coal seam fractures. When the bedrock layer at the working face is close to the right boundary of the isolation layer fracturing, the next isolation layer fracturing should be performed.Analysis of roof stress evolution lawFigure 10 illustrates the change law of the roof support pressure when mining of the working face, in which the roof support pressure curve is the stress change minus the initial value of the sensor before mining. After the excavation of the working face, the surrounding rock will exhibit stress redistribution. The increase in tangential stress in front of the working face or on both sides is called the support pressure. The peak value of the support pressure generally occurs on the front of the working face. As the working face advanced to 228.2 mm, the direct roof gradually broke and collapsed with mining. Due to the redistribution of surrounding rock stress, the stress fluctuation at the open cut was clear. In front of the working face, the overlying rock stress was redistributed due to mining, and the vertical pressure peak area appeared, with a stress increment of 0.03 MPa. When the working face advanced to 360.8 mm, the first cycle pressure on the roof occurred. The falling gangue in the mine-out area gradually approached its upper strata, and the peak support pressure increments reached 0.05 MPa. During the advancement of the working face to 592.9 mm, the direct roof continued to collapse. The gangue at the cuttings was gradually compacted with the roof, and the stresses gradually restored to stability. Coal seam mining led to the decompression of the floor, and the vertical stress maximum reduction at the working face was 0.045 MPa. The peak vertical pressure in front of the working face shifted to the right as mining progressed. When degradation reached 726.5 mm, the fifth periodic pressure on the roof occurred. Figure 10b shows that the fracture of the isolation layer had no apparent effect on the change in roof stress. Within 560 mm from the open excavation, the mine-out area gangue gradually compacted with the roof. Vertical pressure changes between the fourth and fifth periodic pressures are slight and practically nonsignificant.Figure 10Vertical pressure variation law with coal mining. (a) First pressure and First periodic pressure difference. (b) Fourth and First periodic pressure difference. (c) Eighth and Ninth periodic pressure difference. (d) Eleventh and Twelfth periodic pressure difference. (e) Variation laws of vertical pressure with mining.Full size imageWhen the mining reached 1031.6 mm, the directly caving gangue completely filled the goaf and was compacted with the roof. The upper roof of the caving rock was supported again, and the compaction range of the mining area extended to 821 mm. As the working face advanced to 1338.9 mm, the peak vertical pressure appeared at 1400 mm, with a maximum increment of 0.375 MPa. The compaction range of the mining area extends to 1200 mm. Then, the fractured isolation layer can be grouted. The subsequent working face advances until the end of mining, and the rock movement above the mine-out zone will exhibit a periodic “falling-filling-cutting-compaction” process. Fracture grouting of the flexible isolation layer has no significant effect on the vertical stress changes, and the stress unloading area and the peak vertical pressure will continue to change with mining. Nevertheless, consideration needs to be given to the adequacy of the gangue falling from the roof for isolation layer grouting.Roof displacement and development pattern of water-conducting fracture zoneFigure 11 shows the development law of the roof water-conducting fissures in the roof of the coal seam during different pressure periods, where the illustration shows the von Mises equivalent strain. Figure 12 shows the development trend of the water-conducting fracture zone height. From the whole observation, although the isolation layer is treated by fracturing before back mining, it has less influence on the displacement and deformation of the overlying rock layer because it is restricted by the surrounding rock of the model. When the working face was mined to 228.2 mm, the upper roof of the mining face collapsed, and the first periodic pressure occurred on the roof. The roof displacement reached the Yan’an Group mudstone layer, and the roof collapse height was only 104.3 mm. As the mining advanced, the roof fractures in the mining-out area continued to develop upwards. When the working face was mined to 360.8 mm, the first cycle pressure on the roof occurred, and the roof collapse height extended upwards to the siltstone of the Yan’an Formation, with a collapse range of 117.6 mm. At this point, only a small displacement change occurred around the direct roof, and the flexible isolation layer was basically not affected by any impact.Figure 11Development regularity of roof water-conducting fissures during different period pressure.Full size imageFigure 12Development height curve of water-conducting fracture zone.Full size imageFrom the second cycle pressure onwards, the development trend accelerated significantly, and the collapsed height rose rapidly to 210.9 mm. When the working face advanced to 537.1 mm, the third cycle pressure occurred on the roof. The collapsed Yan’an Formation mudstone layer was further pressurized by its upper layers and collapsed to a height of 344.7 mm. The roof displacement had spread to the coarse sandstone of the Naoro Formation, but the height of the water-conducting fracture zone had not reached the bottom of the isolation layer. When the workings reached 592.9 mm, the roof collapsed again, showing the fourth periodic pressure. The water-conducting fissure zone continues to develop upwards to 355.3 mm, which passes through the fissure isolation layer and reaches the gritstone at the top of the isolation layer. The fractured isolation layer is in an “activated” state.When the working face reached 1031.6 mm, fallen gangue completely filled the mining-out area and compacted with the roof, and eighth periodic pressure occurred on the roof. The height of the water-conducting fracture zone developed to 496.8 mm, which was lower than the height of the water-conducting fracture zone of 565.8 mm at the seventh periodic pressure. After that, the old roof collapsed as a cantilevered beam. The development height of the water-conducting fracture zone was allegedly less than 565.8 mm. Afterwards, the roof fracturing direction was consistent with the direction of working face advancement, from left to right. Displacement and fracture of the overlying rock layer were mainly caused by the overall downwards sliding of the upper rock seam due to the collapse of the bottom rock seam. At different heights of the coal seam roof, the degree of displacement damage decreased with increasing height.When the working face reached 1178.7 mm, the roof covering the open cut stabilized. The fractured isolation layers in the 1st ~ 13th groups were grouted, and then the coal was mined only after the slurry had completely solidified and reached a certain strength. The eleventh periodic pressure occurred on the roof, with a water-conducting fracture height of 367.6 mm at this time. When the working face was advanced to 1471.9 mm and 1645.2 mm, the roof had twelfth and fourteenth periodic pressures, and the heights of the water-conducting fracture zone were 332.0 mm and 416.0 mm, respectively. Then, the 14th ~ 15th and 16th ~ 17th group isolation layers of the upper coal seam were grouted while fracturing the right isolation layer. However, the disruption of displacement towards the extent of the development had a relatively small impact, mainly on the roof rock layer above the mining face. Table 2 indicates the development height of the water-conducting fracture zone and the fracture and grouting sequence of the isolated layer.Table 2 Development pattern of water-conducting fracture zone and fracture and grouting sequence of isolated layer.Full size tableDuring the mining process, damage to the water-conducting fissure zone was always a major factor in the displacement of the roof slab. Nonetheless, after fracturing and grouting measures, the effects of the damage were significantly reduced such that the damage to the roof rock was contained within the flexible isolation layer. After grouting, the enhanced strength of the isolation layer ensured that mining was carried out normally. During the mining period, four grouting reforms were made, and the isolation layer was fractured six times, with the maximum development height of the water-conducting fracture zone located at the seventh periodic pressure, reaching 565.8 mm.Analysis of water flow evolution law of overburden roofTo analyse the seepage law of the overburden roof, seven water flow monitoring lines were arranged from the top of the flexible isolation layer to the direct roof of the coal seam. The No. 1 water flow monitoring line was placed in the position of the third group of the isolation layer, which is initially located outside the deformation range of bedrock disturbed by mining and outside the stop line. The flow line was mainly used to monitor the influence of the rock disturbance boundary above the open cut on isolated seam fracturing and grouting. No 2–3 water flow monitoring lines were placed at the isolation layer positions of Group 12 and Group 14, which were initially located near the maximum height of the water-conducting fracture zone and were mainly used to monitor the change laws of the water-conducting fracture zone with mining impact. Monitoring Lines 4–6 were placed in isolation layers No. 17, No. 22 and No. 26 to study the impact of water flow changes with mining disturbance and the advanced influence scope. Water flow monitoring line No. 7 was placed in the thirtieth group of isolated layers, which was originally outside the cut-off line. As shown in Fig. 13, white arrows are water flow vectors in mL/min. Fracturing the 1–18 isolation layers before mining, the water tank hot water was injected into the flexible isolation layer such that the iodized salt in the flexible isolation layer was completely dissolved, and the infrared monitor showed the yellow area in the image. At this point, the water flow monitoring Lines 1–3 and 5–7 show yellow status, indicating that after the fracturing of the isolation layer, the aquifer water flows downwards along the fracture. The lower part of monitoring Line 4 was compacted at the top of the coal seam, indicating that the cracks between the roof and the aquifer had not been communicated. Therefore, the water flow rate was 0 mL/min until the sixth periodic pressure. Mining was then undertaken on the working face. The No. 1 monitoring line was therefore less affected by mining due to its layout outside the stop line, and there was no significant change in water flow before the first grouting.Figure 13Water flow evolution of the overburden roof with coal mining.Full size imageAs shown in Fig. 13, when the working face progressed to second periodic pressure, with the collapse of the coal seam, the stress of the surrounding rock was redistributed, the height of the water flowing fractured zone of the roof increased, and the water flow of the No. 2 monitoring line increased from the initial 9.1 mL/min to 14.0 mL/min. As the working face was advanced above the No. 2 monitoring line, the fifth periodic of pressure were generated in the roof. The development height of the roof water flowing fractured zone reached 504.4 mm. The roof was separated and collapsed, the cracks in the monitoring line communicated with each other, and the rock stress was released. The water flow in the No. 2 monitoring line increased significantly. Monitoring line No. 3 was affected by advanced mining, resulting in the coal seam roof’s increased rock fissures, the water flow path and resistance were reduced, and the water flow reached 48.3 mL/min. At the same time, the influence range of working face bedrock was close to the boundary of the first fracturing of the flexible isolation layer, and Groups 20–22 of isolation lays had been fractured.When mining started at the sixth periodic pressure, the roof water-conducting fracture zone gradually reached the maximum height and penetrated the fractured isolation layer, and the fracture of the roof rock increased. Lines No. 2 and No. 3 reached 44.4 mL/min and 85.6 mL/min, respectively. In fact, the encounter may indicate that the confined water of the gritstone aquifer was released, and the water flow of the working face increased. Then, the working face progressed, and the collapsed gangue above the mining-out area was compacted into the bedrock roof. The stress in the goaf did not change significantly, and the cracks in the strata decreased. The No. 2 and No. 3 water flows of the monitoring line gradually dropped. During this period, the change law of monitoring Lines 4–7 was similar to that of No. 2 and No. 3. During coal seam mining, the roof underwent a process of fracture, collapse, compaction and full mining, and the water flow monitoring line also went through a process of rising and then falling.When the working face was advanced to the eleventh periodic pressure, the grouting transformation of isolation layers 1–12 was conducted. The slurry was injected into the flexible isolation layer by hand pressure pump along the grouting pipe. After the slurry solidified, the colour of the No. 1 and No. 2 monitoring lines gradually became shallow, and the water flow gradually decreased under infrared observation. As the extraction of the coal seam progressed and the flexible insulation layer was broken and grouted, the colour of observation Lines 1–4 turned black in the infrared observation until the fourth grouting of insulation layer 18–19, and the water flow rate all showed 0 mL/min. However, the lower strata of the flexible isolation layer were not yet stabilized, so monitoring Lines 5–7 did not undergo any grouting transformation and still had a large water flow until the end of mining. Flow metre and infrared observations show that the destruction and grouting of the flexible isolation layer had a noticeable effect on the seepage characteristics of the overburden. In particular, after the grouting of the isolation layer, the slurry filled and solidified rapidly, the water flow decreased rapidly, and the water plugging effect of flexible isolation layer grouting was remarkable.Discussion and analysisDuring coal seam mining, the fracturing of the flexible isolation layer should be based on the premining overtopping influence range; that is, when the boundary line of bedrock influence extends to the range of the flexible isolation layer reached by the fracturing area of the flexible isolation layer, the next fracturing should continue. The average boundary angle range of the bedrock was 76.7°, and the field angle should not be less than 73.57°. The grouting of the flexible isolation layer considers the full mining degree of the coal seam. When there is no significant change in stress in the mined area, grouting of the flexible isolation layer at the top of the goaf is conducted. According to the simulation experiment in this paper, the full mining distance of the working face is 1338.9 mm, and the actual distance on site is 187.446 m. It is calculated that the distance between the fracture of the flexible isolation layer should be no less than 854.8 mm away from the working face, and the actual distance on site is 119.672 m. After the working face enters full mining, the shortest distance between the fracturing grouting range of the flexible isolation layer and the working face is not less than 242.6 mm, and the actual distance on site is 33.964 m.As seen from the previous analysis, with the advancement of the working face, the bedrock influence boundary angle of the coal seam does not change significantly, which only plays a guiding role in the fracturing sequence of the flexible isolation layer. The fracturing of the flexible isolation layer had an clear influence on the seepage of water-rich bedrock at the bottom of the Zhiluo Formation. The water-flowing fractured zone formed in the process of coal seam mining promoted the release of fractured water in the water-rich bedrock at the bottom of the Zhiluo Formation. The higher the height of the water-flowing fractured zone is, the greater the seepage of the water-rich bedrock. Coal seam mining had little effect on the seepage characteristics of the water-rich bedrock layer at the bottom of the Zhiluo Formation in the range of not disturbed by mining and advanced influence.In accordance with the stress sensor data, when the working face passed a certain distance, the bottom plate of the extraction area was compacted by the falling gangue, and the sensor pressure data did not change with the mining face. At this time, the grouting of the fracturing area of the flexible isolation layer corresponding to the above goaf was not affected by the mining face. For example, the stress in the goaf of 1200 mm had no clear change. Therefore, the first grouting was conducted in the fracturing area. After the solidification of the grouting slurry, the water flow of monitoring lines No. 1 and No. 2 decreased significantly. This minimized the impact on the original geological environment and at the same time reduced the goaf water drainage of the working face. The sealing effect of the isolation layer has an important influence on promoting water-retaining coal mining.The experimental application of the flexible isolation layer has realized its feasibility from the physical simulation test method in this paper. The realization of a flexible isolation layer requires premining fracturing and postmining isolation grouting. At present, premining fracturing can be achieved by directional drilling technology. There are also examples of roof separation grouting for postmining flexible isolation layer grouting28,29. Therefore, there is no technical bottleneck in field applications. Moreover, there is still a certain distance from the specific engineering application. According to the results of this study, it is predicted that the implementation of a flexible isolation layer will have great significance for water conservation coal mining in western China, which can reduce soil erosion and protect surface ecology. More

  • in

    Portugal leads with Europe’s largest marine reserve

    CORRESPONDENCE
    18 January 2022

    Portugal leads with Europe’s largest marine reserve

    Filipe Alves

     ORCID: http://orcid.org/0000-0003-3752-2745

    0
    ,

    João G. Monteiro

     ORCID: http://orcid.org/0000-0002-3401-6495

    1
    ,

    Paulo Oliveira

    2
    &

    João Canning-Clode

     ORCID: http://orcid.org/0000-0003-2143-6535

    3

    Filipe Alves

    MARE Marine and Environmental Sciences Centre, Madeira, Portugal.

    View author publications

    You can also search for this author in PubMed
     Google Scholar

    João G. Monteiro

    MARE Marine and Environmental Sciences Centre, Madeira, Portugal.

    View author publications

    You can also search for this author in PubMed
     Google Scholar

    Paulo Oliveira

    Institute of Forests and Nature Conservation (IFCN, IP-RAM), Funchal, Madeira, Portugal.

    View author publications

    You can also search for this author in PubMed
     Google Scholar

    João Canning-Clode

    Smithsonian Environmental Research Center, Maryland, USA.

    View author publications

    You can also search for this author in PubMed
     Google Scholar

    Twitter

    Facebook

    Email

    Marine conservation is central to the United Nations’ Sustainable Development Goals 13 (climate action) and 14 (life below water). Portugal has now created the largest marine reserve with full protection in Europe and the North Atlantic, an achievement that other nations could follow.

    Access options

    Access through your institution

    Change institution

    Buy or subscribe

    /* style specs start */
    style{display:none!important}.LiveAreaSection-193358632 *{align-content:stretch;align-items:stretch;align-self:auto;animation-delay:0s;animation-direction:normal;animation-duration:0s;animation-fill-mode:none;animation-iteration-count:1;animation-name:none;animation-play-state:running;animation-timing-function:ease;azimuth:center;backface-visibility:visible;background-attachment:scroll;background-blend-mode:normal;background-clip:borderBox;background-color:transparent;background-image:none;background-origin:paddingBox;background-position:0 0;background-repeat:repeat;background-size:auto auto;block-size:auto;border-block-end-color:currentcolor;border-block-end-style:none;border-block-end-width:medium;border-block-start-color:currentcolor;border-block-start-style:none;border-block-start-width:medium;border-bottom-color:currentcolor;border-bottom-left-radius:0;border-bottom-right-radius:0;border-bottom-style:none;border-bottom-width:medium;border-collapse:separate;border-image-outset:0s;border-image-repeat:stretch;border-image-slice:100%;border-image-source:none;border-image-width:1;border-inline-end-color:currentcolor;border-inline-end-style:none;border-inline-end-width:medium;border-inline-start-color:currentcolor;border-inline-start-style:none;border-inline-start-width:medium;border-left-color:currentcolor;border-left-style:none;border-left-width:medium;border-right-color:currentcolor;border-right-style:none;border-right-width:medium;border-spacing:0;border-top-color:currentcolor;border-top-left-radius:0;border-top-right-radius:0;border-top-style:none;border-top-width:medium;bottom:auto;box-decoration-break:slice;box-shadow:none;box-sizing:border-box;break-after:auto;break-before:auto;break-inside:auto;caption-side:top;caret-color:auto;clear:none;clip:auto;clip-path:none;color:initial;column-count:auto;column-fill:balance;column-gap:normal;column-rule-color:currentcolor;column-rule-style:none;column-rule-width:medium;column-span:none;column-width:auto;content:normal;counter-increment:none;counter-reset:none;cursor:auto;display:inline;empty-cells:show;filter:none;flex-basis:auto;flex-direction:row;flex-grow:0;flex-shrink:1;flex-wrap:nowrap;float:none;font-family:initial;font-feature-settings:normal;font-kerning:auto;font-language-override:normal;font-size:medium;font-size-adjust:none;font-stretch:normal;font-style:normal;font-synthesis:weight style;font-variant:normal;font-variant-alternates:normal;font-variant-caps:normal;font-variant-east-asian:normal;font-variant-ligatures:normal;font-variant-numeric:normal;font-variant-position:normal;font-weight:400;grid-auto-columns:auto;grid-auto-flow:row;grid-auto-rows:auto;grid-column-end:auto;grid-column-gap:0;grid-column-start:auto;grid-row-end:auto;grid-row-gap:0;grid-row-start:auto;grid-template-areas:none;grid-template-columns:none;grid-template-rows:none;height:auto;hyphens:manual;image-orientation:0deg;image-rendering:auto;image-resolution:1dppx;ime-mode:auto;inline-size:auto;isolation:auto;justify-content:flexStart;left:auto;letter-spacing:normal;line-break:auto;line-height:normal;list-style-image:none;list-style-position:outside;list-style-type:disc;margin-block-end:0;margin-block-start:0;margin-bottom:0;margin-inline-end:0;margin-inline-start:0;margin-left:0;margin-right:0;margin-top:0;mask-clip:borderBox;mask-composite:add;mask-image:none;mask-mode:matchSource;mask-origin:borderBox;mask-position:0% 0%;mask-repeat:repeat;mask-size:auto;mask-type:luminance;max-height:none;max-width:none;min-block-size:0;min-height:0;min-inline-size:0;min-width:0;mix-blend-mode:normal;object-fit:fill;object-position:50% 50%;offset-block-end:auto;offset-block-start:auto;offset-inline-end:auto;offset-inline-start:auto;opacity:1;order:0;orphans:2;outline-color:initial;outline-offset:0;outline-style:none;outline-width:medium;overflow:visible;overflow-wrap:normal;overflow-x:visible;overflow-y:visible;padding-block-end:0;padding-block-start:0;padding-bottom:0;padding-inline-end:0;padding-inline-start:0;padding-left:0;padding-right:0;padding-top:0;page-break-after:auto;page-break-before:auto;page-break-inside:auto;perspective:none;perspective-origin:50% 50%;pointer-events:auto;position:static;quotes:initial;resize:none;right:auto;ruby-align:spaceAround;ruby-merge:separate;ruby-position:over;scroll-behavior:auto;scroll-snap-coordinate:none;scroll-snap-destination:0 0;scroll-snap-points-x:none;scroll-snap-points-y:none;scroll-snap-type:none;shape-image-threshold:0;shape-margin:0;shape-outside:none;tab-size:8;table-layout:auto;text-align:initial;text-align-last:auto;text-combine-upright:none;text-decoration-color:currentcolor;text-decoration-line:none;text-decoration-style:solid;text-emphasis-color:currentcolor;text-emphasis-position:over right;text-emphasis-style:none;text-indent:0;text-justify:auto;text-orientation:mixed;text-overflow:clip;text-rendering:auto;text-shadow:none;text-transform:none;text-underline-position:auto;top:auto;touch-action:auto;transform:none;transform-box:borderBox;transform-origin:50% 50% 0;transform-style:flat;transition-delay:0s;transition-duration:0s;transition-property:all;transition-timing-function:ease;vertical-align:baseline;visibility:visible;white-space:normal;widows:2;width:auto;will-change:auto;word-break:normal;word-spacing:normal;word-wrap:normal;writing-mode:horizontalTb;z-index:auto;-webkit-appearance:none;-moz-appearance:none;-ms-appearance:none;appearance:none;margin:0}.LiveAreaSection-193358632{width:100%}.LiveAreaSection-193358632 .login-option-buybox{display:block;width:100%;font-size:17px;line-height:30px;color:#222;padding-top:30px;font-family:Harding,Palatino,serif}.LiveAreaSection-193358632 .additional-access-options{display:block;font-weight:700;font-size:17px;line-height:30px;color:#222;font-family:Harding,Palatino,serif}.LiveAreaSection-193358632 .additional-login >li:not(:first-child)::before{transform:translateY(-50%);content:”;height:1rem;position:absolute;top:50%;left:0;border-left:2px solid #999}.LiveAreaSection-193358632 .additional-login >li:not(:first-child){padding-left:10px}.LiveAreaSection-193358632 .additional-login >li{display:inline-block;position:relative;vertical-align:middle;padding-right:10px}.BuyBoxSection-683559780{display:flex;flex-wrap:wrap;flex:1;flex-direction:row-reverse;margin:-30px -15px 0}.BuyBoxSection-683559780 .box-inner{width:100%;height:100%}.BuyBoxSection-683559780 .readcube-buybox{background-color:#f3f3f3;flex-shrink:1;flex-grow:1;flex-basis:255px;background-clip:content-box;padding:0 15px;margin-top:30px}.BuyBoxSection-683559780 .subscribe-buybox{background-color:#f3f3f3;flex-shrink:1;flex-grow:4;flex-basis:300px;background-clip:content-box;padding:0 15px;margin-top:30px}.BuyBoxSection-683559780 .subscribe-buybox-nature-plus{background-color:#f3f3f3;flex-shrink:1;flex-grow:4;flex-basis:100%;background-clip:content-box;padding:0 15px;margin-top:30px}.BuyBoxSection-683559780 .title-readcube{display:block;margin:0;margin-right:20%;margin-left:20%;font-size:24px;line-height:32px;color:#222;padding-top:30px;text-align:center;font-family:Harding,Palatino,serif}.BuyBoxSection-683559780 .title-buybox{display:block;margin:0;margin-right:29%;margin-left:29%;font-size:24px;line-height:32px;color:#222;padding-top:30px;text-align:center;font-family:Harding,Palatino,serif}.BuyBoxSection-683559780 .title-asia-buybox{display:block;margin:0;margin-right:5%;margin-left:5%;font-size:24px;line-height:32px;color:#222;padding-top:30px;text-align:center;font-family:Harding,Palatino,serif}.BuyBoxSection-683559780 .asia-link{color:#069;cursor:pointer;text-decoration:none;font-size:1.05em;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:1.05em6}.BuyBoxSection-683559780 .access-readcube{display:block;margin:0;margin-right:10%;margin-left:10%;font-size:14px;color:#222;padding-top:10px;text-align:center;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:20px}.BuyBoxSection-683559780 .access-asia-buybox{display:block;margin:0;margin-right:5%;margin-left:5%;font-size:14px;color:#222;padding-top:10px;text-align:center;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:20px}.BuyBoxSection-683559780 .access-buybox{display:block;margin:0;margin-right:30%;margin-left:30%;font-size:14px;color:#222;opacity:.8px;padding-top:10px;text-align:center;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:20px}.BuyBoxSection-683559780 .usps-buybox{display:block;margin:0;margin-right:30%;margin-left:30%;font-size:14px;color:#222;opacity:.8px;text-align:center;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:20px}.BuyBoxSection-683559780 .price-buybox{display:block;font-size:30px;color:#222;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;padding-top:30px;text-align:center}.BuyBoxSection-683559780 .price-from{font-size:14px;padding-right:10px;color:#222;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:20px}.BuyBoxSection-683559780 .issue-buybox{display:block;font-size:13px;text-align:center;color:#222;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:19px}.BuyBoxSection-683559780 .no-price-buybox{display:block;font-size:13px;line-height:18px;text-align:center;padding-right:10%;padding-left:10%;padding-bottom:20px;padding-top:30px;color:#222;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif}.BuyBoxSection-683559780 .vat-buybox{display:block;margin-top:5px;margin-right:20%;margin-left:20%;font-size:11px;color:#222;padding-top:10px;padding-bottom:15px;text-align:center;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;line-height:17px}.BuyBoxSection-683559780 .button-container{display:flex;padding-right:20px;padding-left:20px;justify-content:center}.BuyBoxSection-683559780 .button-container >*{flex:1px}.BuyBoxSection-683559780 .button-container >a:hover,.Button-505204839:hover,.Button-1078489254:hover,.Button-2808614501:hover{text-decoration:none}.BuyBoxSection-683559780 .readcube-button{background:#fff;margin-top:30px}.BuyBoxSection-683559780 .button-asia{background:#069;border:1px solid #069;border-radius:0;cursor:pointer;display:block;padding:9px;outline:0;text-align:center;text-decoration:none;min-width:80px;margin-top:75px}.BuyBoxSection-683559780 .button-label-asia,.ButtonLabel-3869432492,.ButtonLabel-3296148077,.ButtonLabel-1566022830{display:block;color:#fff;font-size:17px;line-height:20px;font-family:-apple-system,BlinkMacSystemFont,”Segoe UI”,Roboto,Oxygen-Sans,Ubuntu,Cantarell,”Helvetica Neue”,sans-serif;text-align:center;text-decoration:none;cursor:pointer}.Button-505204839,.Button-1078489254,.Button-2808614501{background:#069;border:1px solid #069;border-radius:0;cursor:pointer;display:block;padding:9px;outline:0;text-align:center;text-decoration:none;min-width:80px;max-width:320px;margin-top:10px}.Button-505204839 .readcube-label,.Button-1078489254 .readcube-label,.Button-2808614501 .readcube-label{color:#069}
    /* style specs end */Subscribe to nature+Get immediate online access to the entire Nature family of 50+ journals$29.99monthlySubscribeSubscribe to JournalGet full journal access for 1 year$199.00only $3.90 per issueSubscribeAll prices are NET prices. VAT will be added later in the checkout.Tax calculation will be finalised during checkout.Buy articleGet time limited or full article access on ReadCube.$32.00BuyAll prices are NET prices.

    Additional access options:

    Log in

    Learn about institutional subscriptions

    Nature 601, 318 (2022)
    doi: https://doi.org/10.1038/d41586-022-00093-8

    Competing Interests
    The authors declare no competing interests.

    Related Articles

    See more letters to the editor

    Subjects

    Conservation biology

    Ecology

    Sustainability

    Latest on:

    Ecology

    Wind power versus wildlife: root mitigation in evidence
    Correspondence 11 JAN 22

    Two million species catalogued by 500 experts
    Correspondence 11 JAN 22

    EU Nature Restoration Law needs ambitious and binding targets
    Correspondence 11 JAN 22

    Sustainability

    Sustainability at the crossroads
    Editorial 21 DEC 21

    The UN must get on with appointing its new science board
    Editorial 08 DEC 21

    Battery-powered trains offer a cost-effective ride to a cleaner world
    Research Highlight 22 NOV 21

    Jobs

    JUNIOR PROFESSOR IN MECHANICAL ENGINEERING: DIGITISATION FOR SMART OPERATIONS AND MAINTENANCE OF MACHINES

    KU Leuven
    Leuven, Belgium

    Director of IOCB Prague

    Institute of Organic Chemistry and Biochemistry of the CAS (IOCB Prague)
    Prague, Czech Republic

    Director of IOCB Prague

    Institute of Organic Chemistry and Biochemistry of the CAS (IOCB Prague)
    Prague, Czech Republic

    Chair, Department of Physiology

    Tulane University School of Medicine (SOM)
    New Orleans, LA, United States More

  • in

    Fishing activity before closure, during closure, and after reopening of the Northeast Canyons and Seamounts Marine National Monument

    Data and softwareThis analysis used two main data sources: (1) annual (through 2020) summaries of landings by species and by region provided by the Atlantic Coastal Cooperative Statistics Program (ACCSP), and (2) vessel-tracking data provided by Global Fishing Watch. The ACCSP is a cooperative state-federal program of U.S. states and the District of Columbia; it was established in 1995 to be the principal source of fisheries-dependent information on the Atlantic Coast of the United States. For the ACCSP data, I obtained annual landings by species for the North Atlantic region, Mid Atlantic region, and South Atlantic region (excluding landings from the Gulf of Mexico). The weekly cumulative landings data was obtained from the NOAA Fisheries Greater Atlantic Quota Monitoring website. Global Fishing Watch is an organization that provides access to information on commercial fishing activities, in particular information on the identity and location of fishing vessels34. Many large vessels use a system known as the Automatic Identification System (AIS) to avoid collisions at sea, broadcast their location to port authorities and other vessels, and to view other vessels in their vicinity. Vessels fitted with AIS transceivers can be observed by AIS base stations and by satellites fitted with AIS receivers. The US Coast Guard requires all vessels larger than 65 feet to have an AIS receiver onboard. Global Fishing Watch obtains AIS data for fishing vessels and enables users with Internet access to monitor fishing activity globally, and to view individual vessel tracks. They also partner with academic researchers to provide more fine-scale data.To obtain the vessel-tracking data for the relevant fisheries, I reviewed NOAA databases of squid and mackerel permits (2019 version; vessels with squid permits are automatically issued a Butterfish permit), and the Atlantic tuna permits (2020 version) and matched each permitted vessel to its unique Maritime Mobile Service Identity (MMSI) number, which is associated with Global Fishing Watch tracking information. I was able to identify 84% (187/224) of squid/butterfish permitted vessels (I focused on the SMB1A (Tier 1) permit category associated with the vast majority ( approx. 99%) of squid catch35), 100% of Tier 1 and Tier 2 mackerel-permitted vessels (56/56 vessels), and 74% of active tuna longline vessels (100/135 vessels). “Active” is defined as having reported successfully setting pelagic longline gear at least once between 2006 and 201236. This translates to a total of 17.55 million observations on fishing vessel locations for all three fisheries. I drop any observations that are missing either a latitude or longitude entry. For the squid and mackerel fisheries, I drop any observations with unusual longitudes ((ge 0^{circ }) and ( More

  • in

    Iterative data-driven forecasting of the transmission and management of SARS-CoV-2/COVID-19 using social interventions at the county-level

    1.Ebrahim, S. H., Ahmed, Q. A., Gozzer, E., Schlagenhauf, P. & Memish, Z. A. Covid-19 and community mitigation strategies in a pandemic. BMJ 368, m1066. https://doi.org/10.1136/bmj.m1066 (2020).Article 
    PubMed 

    Google Scholar 
    2.Ebrahim, S. H. et al. All hands on deck: A synchronized whole-of-world approach for COVID-19 mitigation. Int. J. Infect. Dis. 98, 208–215. https://doi.org/10.1016/j.ijid.2020.06.049 (2020).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    3.Kantner, M. & Koprucki, T. Beyond just “flattening the curve”: Optimal control of epidemics with purely non-pharmaceutical interventions. J. Math. Ind. https://doi.org/10.1186/s13362-020-00091-3 (2020).MathSciNet 
    Article 
    PubMed 
    PubMed Central 
    MATH 

    Google Scholar 
    4.Kupferschmidt, K. The lockdowns worked-but what comes next?. Science 368, 218–219. https://doi.org/10.1126/science.368.6488.218 (2020).ADS 
    CAS 
    Article 
    PubMed 

    Google Scholar 
    5.Byambasuren, O. et al. Estimating the seroprevalence of SARS-CoV-2 infections: Systematic review. medRxiv. https://doi.org/10.1101/2020.07.13.20153163 (2020).Article 

    Google Scholar 
    6.Fontanet, A. & Cauchemez, S. COVID-19 herd immunity: Where are we?. Nat. Rev. Immunol. 20, 583–584. https://doi.org/10.1038/s41577-020-00451-5 (2020).CAS 
    Article 
    PubMed 

    Google Scholar 
    7.Chowdhury, R. et al. Dynamic interventions to control COVID-19 pandemic: A multivariate prediction modelling study comparing 16 worldwide countries. Eur. J. Epidemiol. 35, 389–399. https://doi.org/10.1007/s10654-020-00649-w (2020).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    8.Giordano, G. et al. Modelling the COVID-19 epidemic and implementation of population-wide interventions in Italy. Nat. Med. 26, 855–860. https://doi.org/10.1038/s41591-020-0883-7 (2020).CAS 
    Article 
    PubMed 

    Google Scholar 
    9.Kissler, S. M., Tedijanto, C., Goldstein, E., Grad, Y. H. & Lipsitch, M. Projecting the transmission dynamics of SARS-CoV-2 through the postpandemic period. Science 368, 860. https://doi.org/10.1126/science.abb5793 (2020).ADS 
    CAS 
    Article 
    PubMed 

    Google Scholar 
    10.Prem, K. et al. The effect of control strategies to reduce social mixing on outcomes of the COVID-19 epidemic in Wuhan, China: A modelling study. Lancet Public Health 5, e261–e270. https://doi.org/10.1016/S2468-2667(20)30073-6 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    11.Leung, K., Wu, J. T., Liu, D. & Leung, G. M. First-wave COVID-19 transmissibility and severity in China outside Hubei after control measures, and second-wave scenario planning: A modelling impact assessment. Lancet 395, 1382–1393. https://doi.org/10.1016/S0140-6736(20)30746-7 (2020).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    12.Peng, L., Yang, W., Zhang, D., Zhuge, C. & Hong, L. Epidemic analysis of COVID-19 in China by dynamical modeling. medRxiv. https://doi.org/10.1101/2020.02.16.20023465 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    13.Read, J. M., Bridgen, J. R. E., Cummings, D. A. T., Ho, A. & Jewell, C. P. Novel coronavirus 2019-nCoV: Early estimation of epidemiological parameters and epidemic predictions. medRxiv. https://doi.org/10.1101/2020.01.23.20018549 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    14.Roda, W. C., Varughese, M. B., Han, D. & Li, M. Y. Why is it difficult to accurately predict the COVID-19 epidemic?. Infect. Dis. Model 5, 271–281. https://doi.org/10.1016/j.idm.2020.03.001 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    15.Wu, J. T., Leung, K. & Leung, G. M. Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: A modelling study. Lancet 395, 689–697. https://doi.org/10.1016/S0140-6736(20)30260-9 (2020).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    16.Perc, M., Gorišek Miksić, N., Slavinec, M. & Stožer, A. Forecasting COVID-19. Front. Phys. https://doi.org/10.3389/fphy.2020.00127 (2020).Article 

    Google Scholar 
    17.Er, S., Yang, S. & Zhao, T. COUnty aggRegation mixup AuGmEntation (COURAGE) COVID-19 prediction. Sci. Rep. 11, 14262. https://doi.org/10.1038/s41598-021-93545-6 (2021).ADS 
    CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    18.Hunter, E., Mac Namee, B. & Kelleher, J. An open-data-driven agent-based model to simulate infectious disease outbreaks. PLoS One. https://doi.org/10.1371/journal.pone.0208775 (2018).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    19.Venkatramanan, S. et al. Using data-driven agent-based models for forecasting emerging infectious diseases. Epidemics 22, 43–49. https://doi.org/10.1016/j.epidem.2017.02.010 (2018).Article 
    PubMed 

    Google Scholar 
    20.Brett, T. S. & Rohani, P. Transmission dynamics reveal the impracticality of COVID-19 herd immunity strategies. Proc. Natl. Acad. Sci. U. S. A. https://doi.org/10.1073/pnas.2008087117 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    21.Britton, T., Ball, F. & Trapman, P. A mathematical model reveals the influence of population heterogeneity on herd immunity to SARS-CoV-2. Science 369, 846–849. https://doi.org/10.1126/science.abc6810 (2020).ADS 
    MathSciNet 
    CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    22.Beven, K. Environmental Modelling: An Uncertain Future? (CRC Press, 2010).
    Google Scholar 
    23.Dietze, M. C. Prediction in ecology: A first-principles framework. Ecol. Appl. 27, 2048–2060. https://doi.org/10.1002/eap.1589 (2017).Article 
    PubMed 

    Google Scholar 
    24.Dietze, M. C. et al. Iterative near-term ecological forecasting: Needs, opportunities, and challenges. Proc. Natl. Acad. Sci. 115, 1424. https://doi.org/10.1073/pnas.1710231115 (2018).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    25.Keenan, T. F., Carbone, M. S., Reichstein, M. & Richardson, A. D. The model-data fusion pitfall: Assuming certainty in an uncertain world. Oecologia 167, 587–597. https://doi.org/10.1007/s00442-011-2106-x (2011).ADS 
    Article 
    PubMed 

    Google Scholar 
    26.Niu, S. et al. The role of data assimilation in predictive ecology. Ecosphere 5, art65. https://doi.org/10.1890/ES13-00273.1 (2014).Article 

    Google Scholar 
    27.White, E. P. et al. Developing an automated iterative near-term forecasting system for an ecological study. Methods Ecol. Evol. 10, 332–344. https://doi.org/10.1111/2041-210X.13104 (2019).Article 

    Google Scholar 
    28.Luo, Y. et al. Ecological forecasting and data assimilation in a data-rich era. Ecol. Appl. 21, 1429–1442. https://doi.org/10.1890/09-1275.1 (2011).Article 
    PubMed 

    Google Scholar 
    29.White, B. G. et al. Short-term forecast validation of six models. Weather Forecast. 14, 84–108. https://doi.org/10.1175/1520-0434(1999)014%3C0084:STFVOS%3E2.0.CO;2 (1999).ADS 
    Article 

    Google Scholar 
    30.Calvetti, D., Hoover, A. P., Rose, J. & Somersalo, E. Metapopulation network models for understanding, predicting, and managing the coronavirus disease COVID-19. Front. Phys. https://doi.org/10.3389/fphy.2020.00261 (2020).Article 

    Google Scholar 
    31.O’Sullivan, D., Gahegan, M., Exeter, D. J. & Adams, B. Spatially explicit models for exploring COVID-19 lockdown strategies. T Gis 24, 967–1000. https://doi.org/10.1111/tgis.12660 (2020).Article 

    Google Scholar 
    32.James, N., Menzies, M. & Bondell, H. Understanding spatial propagation using metric geometry with application to the spread of COVID-19 in the United States. EPL (Europhys. Lett.) 135, 48004. https://doi.org/10.1209/0295-5075/ac2752 (2021).ADS 
    CAS 
    Article 

    Google Scholar 
    33.Li, D. et al. Identifying US County-level characteristics associated with high COVID-19 burden. BMC Public Health 21, 1007. https://doi.org/10.1186/s12889-021-11060-9 (2021).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    34.Bisset, K. R. et al. INDEMICS: An interactive high-performance computing framework for data-intensive epidemic modeling. ACM Trans. Model Comput. Simul. https://doi.org/10.1145/2501602 (2014).MathSciNet 
    Article 
    PubMed 
    PubMed Central 
    MATH 

    Google Scholar 
    35.Chao, D. L., Halloran, M. E., Obenchain, V. J. & Longini, I. M. Jr. FluTE, a publicly available stochastic influenza epidemic simulation model. PLoS Comput. Biol. 6, e1000656. https://doi.org/10.1371/journal.pcbi.1000656 (2010).ADS 
    MathSciNet 
    CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    36.Marathe, M. V. & Ramakrishnan, N. Recent advances in computational epidemiology. IEEE Intell. Syst. 28, 96–101. https://doi.org/10.1109/MIS.2013.114 (2013).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    37.Dowd, M. A sequential Monte Carlo approach for marine ecological prediction. Environmetrics 17, 435–455. https://doi.org/10.1002/env.780 (2006).MathSciNet 
    Article 

    Google Scholar 
    38.Gu, F. On-demand data assimilation of large-scale spatial temporal systems using sequential Monte Carlo methods. Simul. Model. Pract. Theory 85, 1–14. https://doi.org/10.1016/j.simpat.2018.03.007 (2018).Article 

    Google Scholar 
    39.Michael, E. et al. Continental-scale, data-driven predictive assessment of eliminating the vector-borne disease, lymphatic filariasis, in sub-Saharan Africa by 2020. BMC Med. 15, 176. https://doi.org/10.1186/s12916-017-0933-2 (2017).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    40.Poole, D. & Raftery, A. E. Inference for deterministic simulation models: The Bayesian melding approach. J. Am. Stat. Assoc. 95, 1244–1255. https://doi.org/10.1080/01621459.2000.10474324 (2000).MathSciNet 
    Article 
    MATH 

    Google Scholar 
    41.Singh, B. K. & Michael, E. Bayesian calibration of simulation models for supporting management of the elimination of the macroparasitic disease, Lymphatic Filariasis. Parasites Vectors 8, 522. https://doi.org/10.1186/s13071-015-1132-7 (2015).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    42.Sisson, S. A., Fan, Y. & Tanaka, M. M. Sequential Monte Carlo without likelihoods. Proc. Natl. Acad. Sci. 104, 1760. https://doi.org/10.1073/pnas.0607208104 (2007).ADS 
    MathSciNet 
    CAS 
    Article 
    PubMed 
    PubMed Central 
    MATH 

    Google Scholar 
    43.Spear, R. C., Hubbard, A., Liang, S. & Seto, E. Disease transmission models for public health decision making: Toward an approach for designing intervention strategies for Schistosomiasis japonica. Environ. Health Perspect. 110, 907–915. https://doi.org/10.1289/ehp.02110907 (2002).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    44.Taylor, S. D. & White, E. P. Automated data-intensive forecasting of plant phenology throughout the United States. Ecol. Appl. 30, e02025. https://doi.org/10.1002/eap.2025 (2020).Article 
    PubMed 

    Google Scholar 
    45.Beaulieu-Jones, B. K. & Greene, C. S. Reproducibility of computational workflows is automated using continuous analysis. Nat. Biotechnol. 35, 342–346. https://doi.org/10.1038/nbt.3780 (2017).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    46.Delgoshaei, P., Austin, M. A. & Pertzborn, A. J. A semantic framework for modeling and simulation of cyber-physical systems. Int. J. Adv. Sys. Measure. 7, 223–237 (2014).
    Google Scholar 
    47.Dong, E., Du, H. & Gardner, L. An interactive web-based dashboard to track COVID-19 in real time. Lancet Infect. Dis. 20, 533–534. https://doi.org/10.1016/S1473-3099(20)30120-1 (2020).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    48.Henkel, R., Wolkenhauer, O. & Waltemath, D. Combining computational models, semantic annotations and simulation experiments in a graph database. Database https://doi.org/10.1093/database/bau130 (2015).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    49.Merkel, D. Docker: Lightweight linux containers for consistent development and deployment. Linux J. 2014, 2 (2014).
    Google Scholar 
    50.Nakamura, K., Higuchi, T. & Hirose, N. Sequential data assimilation: Information fusion of a numerical simulation and large scale observation data. J. UCS 12, 608–626. https://doi.org/10.3217/jucs-012-06-0608 (2006).Article 

    Google Scholar 
    51.Stodden, V. & Miguez, S. Best practices for computational science: Software infrastructure and environments for reproducible and extensible research. J. Open Res. Softw. https://doi.org/10.5334/jors.ay (2014).Article 

    Google Scholar 
    52.Unacast. Social distancing scoreboard. https://www.unacast.com/covid19/social-distancing-scoreboard (2020).53.Willem, L. et al. SOCRATES: An online tool leveraging a social contact data sharing initiative to assess mitigation strategies for COVID-19. BMC Res. Notes 13, 293. https://doi.org/10.1186/s13104-020-05136-9 (2020).CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    54.Iboi, E. A., Ngonghala, C. N. & Gumel, A. B. Will an imperfect vaccine curtail the COVID-19 pandemic in the U.S.?. Infect. Dis. Model 5, 510–524. https://doi.org/10.1016/j.idm.2020.07.006 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    55.Badr, H. S. et al. Association between mobility patterns and COVID-19 transmission in the USA: A mathematical modelling study. Lancet Infect. Dis. https://doi.org/10.1016/S1473-3099(20)30553-3 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    56.Contreras, S., Villavicencio, H. A., Medina-Ortiz, D., Biron-Lattes, J. P. & Olivera-Nappa, A. A multi-group SEIRA model for the spread of COVID-19 among heterogeneous populations. Chaos Solitons Fractals 136, 109925. https://doi.org/10.1016/j.chaos.2020.109925 (2020).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    57.Mossong, J. et al. Social contacts and mixing patterns relevant to the spread of infectious diseases. PLoS Med. 5, e74. https://doi.org/10.1371/journal.pmed.0050074 (2008).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    58.Chen, R. Markov Chain Monte Carlo Vol. Volume 7 Lecture Notes Series, Institute for Mathematical Sciences, National University of Singapore 147–182 (Co-Published with Singapore University Press, 2005).59.Doucet, A., Godsill, S. & Andrieu, C. On sequential Monte Carlo sampling methods for Bayesian filtering. Stat. Comput. 10, 197–208. https://doi.org/10.1023/A:1008935410038 (2000).Article 

    Google Scholar 
    60.Fearnhead, P. & Kunsch, H. R. Particle filters and data assimilation. Annu. Rev. Stat. Appl. 5, 421–449. https://doi.org/10.1146/annurev-statistics-031017-100232 (2018).MathSciNet 
    Article 

    Google Scholar 
    61.Gu, F., Butt, M., Ai, C., Shen, X. & Xiao, J. Proceedings of the Conference on Summer Computer Simulation 1–10 (Society for Computer Simulation International, 2015).62.Florida Agency for Health Care Administration. https://ahca.myflorida.com/ (2020).63.Polonsky, J. A. et al. Outbreak analytics: A developing data science for informing the response to emerging pathogens. Philos. Trans. R. Soc. B. https://doi.org/10.1098/rstb.2018.0276 (2019).Article 

    Google Scholar 
    64.Gambhir, M. et al. Geographic and ecologic heterogeneity in elimination thresholds for the major vector-borne helminthic disease, lymphatic filariasis. BMC Biol. 8, 22. https://doi.org/10.1186/1741-7007-8-22 (2010).Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    65.Spear, R. C. & Hubbard, A. Modelling Parasite Transmission and Control 99–111 (Springer, 2010).66.James, N. & Menzies, M. COVID-19 in the United States: Trajectories and second surge behavior. Chaos Interdiscip. J. Nonlinear Sci. 30, 091102. https://doi.org/10.1063/5.0024204 (2020).MathSciNet 
    CAS 
    Article 

    Google Scholar 
    67.Chang, S. et al. Mobility network models of COVID-19 explain inequities and inform reopening. Nature 589, 82–87. https://doi.org/10.1038/s41586-020-2923-3 (2021).ADS 
    CAS 
    Article 
    PubMed 
    PubMed Central 

    Google Scholar 
    68.James, N. & Menzies, M. Efficiency of communities and financial markets during the 2020 pandemic. Chaos Interdiscip. J. Nonlinear Sci. 31, 083116. https://doi.org/10.1063/5.0054493 (2021).MathSciNet 
    Article 

    Google Scholar 
    69.Yilmazkuday, H. Stay-at-home works to fight against COVID-19: International evidence from Google mobility data. J. Hum. Behav. Soc. Environ. 31, 210–220. https://doi.org/10.1080/10911359.2020.1845903 (2021).Article 

    Google Scholar 
    70.Brienen, N. C., Timen, A., Wallinga, J., Van Steenbergen, J. E. & Teunis, P. F. The effect of mask use on the spread of influenza during a pandemic. Risk Anal. 30, 1210–1218. https://doi.org/10.1111/j.1539-6924.2010.01428.x (2010).Article 
    PubMed 
    PubMed Central 

    Google Scholar  More