Ecology

Effect of warming on physical and chemical conditionsThe water temperature in the warmed treatment was increased by 2.87 ± 0.20 °C in spring and 3.04 ± 0.08 °C in fall, compared to the control (Fig. 1a,b, Table 1). The average temperature in the control treatment, throughout the duration of the experiment, was about 4 °C cooler in spring (14.84 ± 0.03 °C) than in fall (19.01 ± 0.02 °C). In spring, the temperature naturally increased by approximately 4.19 °C from day (d) 10, until the end of the experiment, whereas it remained relatively constant in the fall experiment. It displayed higher diurnal variations in spring than in fall: over the course of the experiments, daily temperature variation ranged from 0.87 to 1.98 °C in spring and from 0.23 to 1.12 °C in fall. The average Daily Light Integral (DLI) in the control treatment, was almost twice as high in the spring experiment (7.93 ± 0.61 mol m−2 d−1) than during the fall (4.61 ± 0.52 mol m−2 d−1) (Fig. 1c,d). Warming did not significantly alter the DLI in fall (Table 1); however, the DLI could not be measured in the warm mesocosms in spring, owing to technical problems.Figure 1Time series of physical and chemical variables. Water temperature (a, b), Daily Light Integral (DLI, c, d), ammonium (NH4+, e, f), nitrates (NO3− + NO2−, g, h), orthophosphate (PO43−, i, j), silicate (SiO2, k, l) concentrations, and N/P ratio (m, n) over the course of the spring (a, c,e, g, i, k, m) and fall (b, d, f, h, j, l, n) experiments in the control (black) and the warmed (orange) treatments. Error bars represent range of observation for the two mesocosms per treatment in spring and the standard deviation for the three mesocosms per treatment in fall. Dotted lines represent the missing data on d10 of the fall experiment due to bad weather conditions. Due to technical difficulties, DLI could not be calculated in the warmed mesocosms of the spring experiment.Full size imageTable 1 Summary of the p values obtained by Repeated Measures Analyses Of VAriance (RM-ANOVA, with treatment as fixed factor and time as random factor) comparing physical parameters and nutrient concentrations in the warmed and control mesocosms.Full size tableNutrient concentrations were measured daily in all mesocosms (Fig. 1, Table 1). Ammonium (NH4+) concentrations were higher in spring than in fall in the controls (0.45 ± 0.08 µM, and 0.41 ± 0.05 µM, respectively). Ammonium concentrations were significantly different, between the control and warmed mesocosms, only at the end of the fall experiment (warmed with a mean of 0.58 ± 0.23 µM, between d11–17; and control with a mean of 0.21 ± 0.09 µM, between d11–17). In contrast, ammonium concentrations did not vary between the control and warmed mesocosms in spring (Table 1). Cohen’s effect size (d) was used to evaluate the magnitude of the effect of warming. Regarding ammonium concentrations, the values were ten times larger in spring than in fall.The nitrate + nitrite (NO3− + NO2−) concentrations in the control treatments were higher in the spring, compared to the fall experiment (0.71 ± 0.08 µM and 0.23 ± 0.02 µM, respectively). Moreover, warming had different effects, depending on the experiment. In spring, the nitrate + nitrite concentrations were significantly higher in the warmed mesocosms from d8 until the end of the experiment, with an average difference of 540.5% between the warmed and control mesocosms, corresponding to a very large d. In fall, nitrate + nitrite concentrations were significantly lower in the warmed mesocosms than in the control, with an average difference of 20.4%, and a medium-sized d.Similar to the nitrate + nitrite concentrations, the orthophosphate (PO43−) concentrations in the control treatment were higher in spring (0.55 ± 0.07 µM and 0.17 ± 0.01 µM, respectively) than in fall. The concentrations were negatively affected by warming throughout the spring experiment, with an average decrease of 9.3%. However, the largest negative effect of experimental warming was observed at the end of fall, with an average decrease of 16.7%, between d15 and d17.Contrary to the nitrate + nitrite and orthophosphate concentrations, the silicate (SiO2) concentrations in the control treatments were, on average, lower in spring (3.31 ± 0.18 µM and 10.42 ± 0.15 µM, respectively) than in fall. Warming had a significant positive effect during the second part of the spring experiment (from d10 to d17), with average concentrations being 27.8% higher in the warmed mesocosms, than in the control. In contrast, the strongest effect of warming on silicate concentrations was observed at the end of the fall experiment, when the silicate concentrations were significantly lower in the warmed mesocosms, by an average of 10.8%, between d15 and d17.The N/P ratio, calculated as the sum of nitrate, nitrite and ammonium concentrations divided by orthophosphate concentration, was 1.99 and 2.39 on average in the control treatment of the spring and fall experiments, respectively (Fig. 1m,n). It was significantly higher over the entire experiments in the warmed treatment by on average 191.5% and 58.8% in spring and fall, respectively (Table 1). In fall, the highest difference between treatments was seen during the second half of the experiment, when the ratio was significantly higher by 133.5% from day 11 to 17.Effects of warming on gross primary production and respiration rates derived from oxygen sensor dataIn the spring experiment, daily GPP varied between 0.19 ± 0.01 and 1.72 ± 0.15 gO2 m−3 d−1 in the control mesocosms (Fig. 2A). It increased during the first half of the experiment (d2–d10), then decreased toward the end of the experiment. In the fall experiment, the daily GPP was lower than what was observed in the control mesocosms in spring and varied between 0.12 ± 0.02 and 0.96 ± 0.16 gO2 m−3 d−1 (Fig. 2B). In spring, warming significantly reduced GPP by 50.9% over the entire experiment, while in fall, warming enhanced GPP by 21.1% over the entire experiment, 32.3% from d4 to d7, and 44.1% from d12 to d17 (Table 2). In spring, when GPP was normalized by the chl-a measured by the high-frequency sensors, it was not significantly different between the treatments, over the entire experiment (Fig. 2C, Table 2). However, it was significantly higher (138%) in the warmed treatment, during the second half of the experiment (d10– d17). In fall, GPP normalized by the chl-a was also significantly enhanced (12%) by warming (Fig. 2D, Table 2).Figure 2Plankton oxygen metabolism parameters. Gross Primary Production (GPP, A, B), GPP normalized by the chlorophyll-a fluorescence (C, D), Respiration (R, E, F), R normalized by the chlorophyll-a fluorescence (G, H), and GPP:R ratio (I, J) in the control (black) and warmed (orange) treatments. Error bars represent range of observation for the two mesocosms per treatment in spring and the standard deviation for the three mesocosms per treatment in fall. In the spring experiment, GPP: Chl-a and R: Chl-a could not be estimated on d1 and d2.Full size imageTable 2 Summary table of the p values and the F-values obtained with the RM-ANOVA (with treatment as fixed factor and time as random factor) comparing the chl-a fluorescence, µ, l, the µ:l ratio, and pigment concentrations in the warmed and in the control treatments over the entire spring and fall experiments or over specific periods defined after trends observed in the data.Full size tableDaily R varied between 0.27 ± 0.02 and 1.92 ± 0.20 gO2 m−3 d−1 in the spring control mesocosms (Fig. 2E). Similar to the daily GPP, it increased during the first half of the experiment (d2– d10), with a strong increase between d8 and d10, before decreasing slowly until the end of the experiment. In the fall experiment, the daily R was lower than in spring, varying from 0.19 ± 0.02 and 1.09 ± 0.14 gO2 m−3 d−1, in the control mesocosms (Fig. 2F). Warming significantly reduced the daily R by an average of 47.9% in spring, while no significant differences were found in fall (Table 2). During both experiments, when daily R was normalized by chl-a, it was not significantly different between treatments over the entire experimental period (Figs. 2G,H), but it was significantly enhanced by warming during the second half of the experiments, by 172% and 49.6%, from d10–17 in spring, and d11–17 in fall, respectively (Table 2).The GPP:R ratio was on average 1.01 and 1.08 in the spring and fall control treatments, respectively (Fig. 2I,J). Consequently, because warming decreased GPP and R to a similar extent in spring, it did not significantly change the GPP:R ratio. Warming significantly increased GPP:R, by an average of 32% in fall (Table 2).Effects of warming on phytoplankton biomass (chlorophyll-a), growth, and loss rates derived from the chlorophyll-a sensor dataThe chl-a fluorescence data was measured using high-frequency sensors, which were inter-calibrated before and after the experiments, and were corrected by the chl-a concentration measured daily by HPLC (see “Methods”). It is hereafter referred to as chl-a. In the spring experiment, the daily chl-a was 5.28 ± 0.21 µg L−1 in the control mesocosms (Fig. 3a,c). A phytoplankton bloom dynamic was observed, with increasing concentrations from d2 to d10, reaching a maximum value of 8.62 ± 0.15 µg L−1, and decreasing concentrations from d10 to d17. The average daily chl-a was lower in fall than in the spring experiment (4.30 ± 0.59 µg L−1) (Fig. 3b,d), and displayed a relatively flat dynamic during the entire experiment, with maximum values on d8 (5.53 ± 0.58 µg L−1).Figure 3Phytoplankton chlorophyll-a, growth and loss rates. High-frequency chlorophyll-a data, uncorrected for Non Photochemical Quenching (NPQ) (a, b), daily average chlorophyll-a data corrected for the NPQ(c, d), phytoplankton growth rate (µ, e, f), loss rate (l, g, h), and µ:l ratio (i, j) in the control (black) and warmed (orange) treatments. Error bars represent range of observation for the two mesocosms per treatment in spring and the standard deviation for the three mesocosms per treatment in fall. In the spring experiment, µ and l could not be estimated on d1 and d12 and, for the latter, the missing data are represented as dotted lines.Full size imageWarming significantly reduced chl-a in both experiments (Table 2): an average of 69.5% from d5 to the end of the spring, and 31.7% from d8 to 15, in the fall experiment. Conversely, warming significantly enhanced chl-a concentrations at the beginning of the fall experiment (19.4% between d2 and d6). Generally, the magnitude of the effect was larger in spring than in fall (Table 2).In the control treatment, µ was higher in spring than in fall (0.44 ± 0.04 d−1 and 0.32 ± 0.05 d−1, respectively; Fig. 3e,f). During both seasons, the maximum µ was observed during the first half of the experiment (spring d7, 0.99 ± 0.01 d−1; fall d4, 0.61 ± 0.03 d−1). Warming enhanced µ by an average of 18.3% and 28.1%, over the entire spring and fall experiments, respectively, and by an average of 56.8% and 50.9%, respectively, from d8 until the end of the experiment (Table 2). The effect size was higher in fall than in spring (Table 2). However, contrary to the general trend of the entire experiment, during spring, warming significantly reduced µ during the first part of the experiment (d2–d7), with an 18.8% mean difference between the treatments.In contrast to µ, l was almost similar between the seasons, with average values of 0.39 ± 0.04 d−1 and 0.40 ± 0.07 d−1, in the control treatments for spring and fall, respectively (Fig. 3g,h). In the spring experiment, warming had a positive effect on the mean l across the study period (37.1%), and even more from d8 to d17 (59.1%), which was larger than the positive effect found for µ. The effect size of warming was not as large in fall, and l was significantly higher in the warmed treatment, although only in the middle of the experiment (20.4% from d7 to d11).When comparing µ and l, the results showed that in spring, µ was higher than l in the control treatment, during the first part of the experiment (d2–d9), and lower during the latter half of the experiment (d10–d17). Warming significantly decreased the µ:l ratio by 28.9%, during the first half of the experiment (D 3–8, Fig. 3i, Table 2), whereas no significant effect was observed in the rest of the experiment. In the fall control treatment, the µ:l ratio was generally lower than that of the spring control (Fig. 3j). Contrary to what was observed in the spring warming, this ratio significantly increased by an average of 92.9%, in the second half of the experiment (d11–d17, Table 2).Effects of warming on phytoplankton pigment concentrationsPhytoplankton pigment composition varied between seasons (Fig. 4). In the spring control treatment, the predominant pigments were fucoxanthin and 19′-hexanoyloxyfucoxanthin (19′-HF), which are mostly associated with diatoms (1.14 ± 0.10 µg L−1) and prymnesiophytes (19′-HF, 2.91 ± 0.14 µg L−1)23,24, respectively (Figs. 4A,B,G,H). The other pigments that were present included peridinin (0.18 ± 0.01 µg L−1), Chl-b (0.14 ± 0.01 µg L−1), zeaxanthin (0.08 ± 0.01 µg L−1), and the specific accessory pigment prasinoxanthin (0.06 ± 0.01 µg L−1), which are associated with dinoflagellates, green algae, cyanobacteria, and prasinophytes, respectively (Figs. 4C–F,I,J)23,24,25.Figure 4Phytoplankton pigment concentrations. Daily pigment concentrations (µg L−1) in the control (black) and warmed (orange) treatments for the spring (A–F) and fall (G–J) experiments. Error bars represent range of observation for the two mesocosms per treatment in spring and the standard deviation for the three mesocosms per treatment in fall. Dotted lines represent the missing data on d10 of the fall experiment due to bad weather conditions. (A, G) fucoxanthin; (B, H) 19′-hexanoyloxyfucoxanthin; (C, I) zeaxanthin; (D, J) chlorophyll-b; (E) peridinin, and (F) prasinoxanthin. Corresponding phytoplankton functional groups are indicated in parentheses.Full size imageIn the fall control treatment, the dominant pigments were the cyanobacteria-associated zeaxanthin (1.78 ± 0.22 µg L−1), the diatom-associated fucoxanthin (1.07 ± 0.27 µg L−1), the green algae-associated Chl-b (0.69 ± 0.14 µg L−1), and the prymnesiophyte-associated 19′-HF (0.68 ± 0.16 µg L−1). Among the main pigments that were identified in the spring experiment, peridinin and prasinoxanthin were either not detected or detected at negligible concentrations in the fall experiment, whereas lutein was detected in fall but not in spring (data not shown).Warming had seasonal effects on pigment concentrations (Table 2). In the spring experiment, warming had a large and significant negative effect on 19′-HF and zeaxanthin concentrations, with mean concentrations decreasing by 75.4% and 75.2%, respectively. Conversely, warming had moderately significant positive effects on peridinin concentration, which increased by an average of 101%.In the fall experiment, warming had a significant negative effect on Chl-b concentration, which decreased by 19.5%, and on zeaxanthin concentration, which significantly decreased in the middle of the experiment (43.4% from d11 to d15). In contrast, a significant positive effect was observed on fucoxanthin concentration, which increased by 210.7%, during the second part of the experiment (between d13 and d17).Relationships between plankton processes, pigment concentrations and environmental parametersPrincipal component analyses (PCA) were used to project plankton processes, pigment concentrations and environmental parameters in a multidimensional space in order to illustrate relationships among variables in both experiments (Fig. 5). For both experiments, GPP and R were clustered together along the first PCA axis, although they appeared closer in spring than in fall (Fig. 5A,B). Conversely, µ was close to ammonium for both experiments, to silicate in spring and to nitrate and nitrite in fall; and l was part of this cluster in spring but not in fall. Concerning phytoplankton pigment composition, in spring, zeaxanthin, associated with cyanobacteria, and 19′-HF, associated with prymnesiophytes, were part of a group together with temperature (Fig. 5C). Similarly, prasinoxanthin, associated with prasinophytes, and Chl-b, associated with green algae, were grouped with DLI and orthophosphate. Finally, peridinin, associated with dinoflagellates, and silicate were clustered together and opposed to fucoxanthin, which is associated with diatoms. In fall, zeaxanthin and Chl-b, representing cyanobacteria and green algae, were part of a group opposed to N-nutrients and temperature, while fucoxanthin was opposed to DLI, silicate and orthophosphate (Fig. 5D).Figure 5Principal component analyses (PCA) of logarithm response ratio (LRR) of plankton processes (A, B) and pigment concentrations (C, D) with environmental parameters for the spring (A, C) and fall (B, D) experiments. GPP: Gross Primary Production, R: Respiration, µ: Phytoplankton growth rate, l: Phytoplankton loss rate, Chl-b: Chlorophyll-b, 19′-HF: 19′-Hexanoyloxyfucoxanthin, Fuco: Fucoxanthin, Prasino: Prasinoxanthin, Zea: Zeaxanthin, DLI: Daily Light Integral.Full size imageTo evaluate specific relationships between phytoplankton processes, environmental variables, and phytoplankton community composition, ordinary least squares linear relationships were assessed for the effects of warming (expressed as the logarithmic response ratio) on GPP, R, µ, and l, nutrient concentrations, DLI, and pigment concentrations (Fig. 6). A significant positive relationship was found between the effects of warming on GPP versus R, and µ versus l (Fig. 6A,B). Moreover, the effect of warming on µ was positively and linearly related to the effects of warming on ammonium in both seasons (Fig. 6C), and to nitrate + nitrite concentrations in spring (Fig. 6D). There was no relationship between the effects of µ on pigment concentrations in the spring experiment, but its effects were positively correlated with the diatom-associated pigment fucoxanthin in fall (Fig. 6E). Similarly, R was positively correlated with fucoxanthin in fall (Fig. 6F). In contrast, significant negative relationships were found between the effects of warming on µ, Chl-b, and zeaxanthin, the pigments associated with green algae and cyanobacteria, respectively (Fig. 6G,H). Similarly, the effect of warming on R was negatively correlated to orthophosphate (Fig. 6I), and the effect on GPP to nitrate + nitrite and peridinin concentrations (Fig. 6J,K).Figure 6Linear relationships between the effect of warming on plankton processes, environment variables and pigment concentrations. Ordinary least squares linear relationships between the effect of warming, expressed as the log response ratio, on GPP, R, µ, and l, and the effect of warming on environmental and pigment variables for the spring (blue circles) and fall (green squares) experiments. Relationships were individually assessed for each experiment. Only statistically significant relationships (p  More

Jones, J. W. et al. The DSSAT cropping system model. Eur. J. Agron. 18, 235–265 (2003).Article 

Google Scholar 
van Diepen, C. A., Wolf, J., van Keulen, H. & Rappoldt, C. WOFOST: a simulation model of crop production. Soil Use Manag. 5, 16–24 (1989).Article 

Google Scholar 
Cao, J. et al. Integrating multi-source data for rice yield prediction across China using machine learning and deep learning approaches. Agric. For. Meteorol. 297, 108275 (2021).ADS 
Article 

Google Scholar 
Khanal, S., Kushal, K. C., Fulton, J. P., Shearer, S. & Ozkan, E. Remote sensing in agriculture—accomplishments, limitations, and opportunities. Remote Sens. 12, 3783 (2020).ADS 
Article 

Google Scholar 
Maas, S. J. Parameterised model of gramineous crop growth: II. within-season simulation calibration. Agron. J. 85, 354–358 (1993).Article 

Google Scholar 
Nguyen, V., Jeong, S., Ko, J., Ng, C. & Yeom, J. Mathematical integration of remotely-sensed information into a crop modelling process for mapping crop productivity. Remote Sens. 11, 2131 (2019).Article 

Google Scholar 
Huang, J. et al. Assimilation of remote sensing into crop growth models: current status and perspectives. Agric. For. Meteorol. 276–277, 107609 (2019).ADS 
Article 

Google Scholar 
Jin, X. et al. A review of data assimilation of remote sensing and crop models. Eur. J. Agron. 92, 141–152 (2018).Article 

Google Scholar 
Shawon, A. R. et al. Assessment of a proximal sensing-integrated crop model for simulation of soybean growth and yield. Remote Sens. 12, 410 (2020).ADS 
Article 

Google Scholar 
Shawon, A. R. et al. Two-dimensional simulation of barley growth and yield using a model integrated with remote-controlled aerial imagery. Remote Sens. 12, 3766 (2020).ADS 
Article 

Google Scholar 
Shin, T. et al. Simulation of wheat productivity using a model integrated with proximal and remotely controlled aerial sensing information. Front. Plant Sci. https://doi.org/10.3389/fpls.2021.649660 (2021).Article 
PubMed 
PubMed Central 

Google Scholar 
Huang, J. et al. Assimilating a synthetic Kalman filter leaf area index series into the WOFOST model to improve regional winter wheat yield estimation. Agric. For. Meteorol. 216, 188–202 (2016).ADS 
Article 

Google Scholar 
Khaki, S., Wang, L. & Archontoulis, S. V. A CNN-RNN framework for crop yield prediction. Front. Plant Sci. https://doi.org/10.3389/fpls.2019.01750 (2020).Article 
PubMed 
PubMed Central 

Google Scholar 
Kim, N. et al. An artificial intelligence approach to prediction of corn yields under extreme weather conditions using satellite and meteorological data. Appl. Sci. 10, 3785 (2020).CAS 
Article 

Google Scholar 
Kumar, P. et al. Comprehensive evaluation of soil moisture retrieval models under different crop cover types using C-band synthetic aperture radar data. Geocarto Int. 34, 1022–1041 (2019).Article 

Google Scholar 
Everingham, Y., Sexton, J., Skocaj, D. & Inman-Bamber, G. Accurate prediction of sugarcane yield using a random forest algorithm. Agron. Sustain. Dev. 36, 27 (2016).Article 

Google Scholar 
Feng, P., Wang, B., Li Liu, D., Waters, C. & Yu, Q. Incorporating machine learning with biophysical model can improve the evaluation of climate extremes impacts on wheat yield in south-eastern Australia. Agric. For. Meteorol. 275, 100–113 (2019).ADS 
Article 

Google Scholar 
Shahhosseini, M., Hu, G., Huber, I. & Archontoulis, S. V. Coupling machine learning and crop modeling improves crop yield prediction in the US Corn Belt. Sci. Rep. 11, 1606 (2021).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Cai, Y. et al. Detecting in-season crop nitrogen stress of corn for field trials using UAV- and CubeSat-based multispectral sensing. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 12, 5153–5166 (2019).ADS 
Article 

Google Scholar 
van Klompenburg, T., Kassahun, A. & Catal, C. Crop yield prediction using machine learning: a systematic literature review. Comput. Electron. Agric. 177, 105709 (2020).Article 

Google Scholar 
Kamilaris, A. & Prenafeta-Boldú, F. X. Deep learning in agriculture: a survey. Comput. Electron. Agric. 147, 70–90 (2018).Article 

Google Scholar 
Bui, D. T., Tsangaratos, P., Nguyen, V.-T., Liem, N. V. & Trinh, P. T. Comparing the prediction performance of a deep learning neural network model with conventional machine learning models in landslide susceptibility assessment. CATENA 188, 104426 (2020).Article 

Google Scholar 
Sahoo, A. K., Pradhan, C. & Das, H. Performance evaluation of different machine learning methods and deep-learning based convolutional neural network for health decision making. In Nature Inspired Computing for Data Science (eds Rout, M. et al.) (Springer International Publishing, 2020).
Google Scholar 
Jeong, S. et al. Development of Variable Threshold Models for detection of irrigated paddy rice fields and irrigation timing in heterogeneous land cover. Agric. Water Manag. 115, 83–91 (2012).Article 

Google Scholar 
Peng, D., Huete, A. R., Huang, J., Wang, F. & Sun, H. Detection and estimation of mixed paddy rice cropping patterns with MODIS data. Int. J. Appl. Earth Obs. Geoinf. 13, 13–23 (2011).ADS 

Google Scholar 
Jeong, S., Ko, J. & Yeom, J.-M. Nationwide projection of rice yield using a crop model integrated with geostationary satellite imagery: a case study in South Korea. Remote Sens. 10, 1665 (2018).ADS 
Article 

Google Scholar 
Xiao, X. et al. Mapping paddy rice agriculture in South and Southeast Asia using multi-temporal MODIS images. Remote Sens. Environ. 100, 95–113 (2006).ADS 
Article 

Google Scholar 
Ozdogan, M. & Gutman, G. A new methodology to map irrigated areas using multi-temporal MODIS and ancillary data: an application example in the continental US. Remote Sens. Environ. 112, 3520–3537 (2008).ADS 
Article 

Google Scholar 
Yeom, J.-M., Jeong, S., Deo, R. C. & Ko, J. Mapping rice area and yield in northeastern Asia by incorporating a crop model with dense vegetation index profiles from a geostationary satellite. GISci. Remote Sens. 58, 1–27 (2021).Article 

Google Scholar 
Yeom, J.-M. et al. Monitoring paddy productivity in North Korea employing geostationary satellite images integrated with GRAMI-rice model. Sci. Rep. 8, 16121 (2018).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Jeong, S., Ko, J., Choi, J., Xue, W. & Yeom, J.-M. Application of an unmanned aerial system for monitoring paddy productivity using the GRAMI-rice model. Int. J. Remote Sens. 39, 2441–2462 (2018).Article 

Google Scholar 
Jeong, S. et al. Geographical variations in gross primary production and evapotranspiration of paddy rice in the Korean Peninsula. Sci. Total Environ. 714, 136632 (2020).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Roger, P., Vermote, E. & Ray, J. MODIS Surface Reflectance User’s Guide. Collection 6 (2015).Scharlemann, J. P. W. et al. Global data for ecology and epidemiology: a novel algorithm for temporal Fourier processing MODIS data. PLoS ONE 3, e1408 (2008).ADS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Pede, T. & Mountrakis, G. An empirical comparison of interpolation methods for MODIS 8-day land surface temperature composites across the conterminous Unites States. ISPRS J. Photogramm. Remote Sens. 142, 137–150 (2018).ADS 
Article 

Google Scholar 
Kilibarda, M. et al. Spatio-temporal interpolation of daily temperatures for global land areas at 1 km resolution. J. Geophys. Res. Atmos. 119, 2294–2313 (2014).ADS 
Article 

Google Scholar 
Nunez, M. The development of a satellite-based insolation model for the tropical western Pacific Ocean. Int. J. Climatol. 13, 607–627 (1993).Article 

Google Scholar 
Otkin, J. A., Anderson, M. C., Mecikalski, J. R. & Diak, G. R. Validation of GOES-based insolation estimates using data from the U.S. Climate reference network. J. Hydrometeorol. 6, 460–475 (2005).ADS 
Article 

Google Scholar 
Pinker, R. & Laszlo, I. Modeling surface solar irradiance for satellite applications on a global scale. J. Appl. Meteorol. 31, 194–211 (1992).ADS 
Article 

Google Scholar 
Kawamura, H., Tanahashi, S. & Takahashi, T. Estimation of insolation over the Pacific Ocean off the Sanriku coast. J. Oceanogr. 54, 457–464 (1998).Article 

Google Scholar 
Yeom, J.-M., Seo, Y.-K., Kim, D.-S. & Han, K.-S. Solar radiation received by slopes using COMS imagery, a physically based radiation model, and GLOBE. J. Sens. 2016, 1–15 (2016).Article 

Google Scholar 
Yeom, J.-M., Han, K.-S. & Kim, J.-J. Evaluation on penetration rate of cloud for incoming solar radiation using geostationary satellite data. Asia-Pac. J. Atmos. Sci. 48, 115–123 (2012).ADS 
Article 

Google Scholar 
Kawai, Y. & Kawamura, H. Validation and improvement of satellite-derived surface solar radiation over the Northwestern Pacific Ocean. J. Oceanogr. 61, 79–89 (2005).Article 

Google Scholar 
Tanahashi, S., Kawamura, H., Matsuura, T., Takahashi, T. & Yusa, H. A system to distribute satellite incident solar radiation in real-time. Remote Sens. Environ. 75, 412–422 (2001).ADS 
Article 

Google Scholar 
Elbern, H., Schmidt, H., Talagrand, O. & Ebel, A. 4D-variational data assimilation with an adjoint air quality model for emission analysis. Environ. Model. Softw. 15, 539–548 (2000).Article 

Google Scholar 
Press, W. H., Teukolsky, S. A., Vetterling, W. T. & Flannery, B. P. Numerical Recipes: The Art of Scientific Computing (Cambridge University Press, 1992).MATH 

Google Scholar 
Ko, J. et al. Simulation and mapping of rice growth and yield based on remote sensing. J. Appl. Remote Sens. 9, 096067 (2015).Article 

Google Scholar 
Emami Javanmard, M., Ghaderi, S. F. & Hoseinzadeh, M. Data mining with 12 machine learning algorithms for predict costs and carbon dioxide emission in integrated energy-water optimization model in buildings. Energy Convers. Manag. 238, 114153 (2021).CAS 
Article 

Google Scholar 
Diebold, F. X. & Shin, M. Machine learning for regularized survey forecast combination: partially-egalitarian LASSO and its derivatives. Int. J. Forecast. 35, 1679–1691 (2019).Article 

Google Scholar 
Khosla, E., Dharavath, R. & Priya, R. Crop yield prediction using aggregated rainfall-based modular artificial neural networks and support vector regression. Environ. Dev. Sustain. 22, 5687–5708 (2020).Article 

Google Scholar 
Wang, S., Azzari, G. & Lobell, D. B. Crop type mapping without field-level labels: random forest transfer and unsupervised clustering techniques. Remote Sens. Environ. 222, 303–317 (2019).ADS 
Article 

Google Scholar 
Ustuner, M. & Balik, S. F. Polarimetric target decompositions and light gradient boosting machine for crop classification: a comparative evaluation. ISPRS Int. J. Geo Inf. 8, 97 (2019).Article 

Google Scholar 
Jeong, S., Ko, J. & Yeom, J.-M. Predicting rice yield at pixel scale through synthetic use of crop and deep learning models with satellite data in South and North Korea. Sci. Total Environ. 802, 149726 (2022).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Nash, J. E. & Sutcliffe, J. V. River flow forecasting through conceptual models part I: a discussion of principles. J. Hydrol. 10, 282–290 (1970).ADS 
Article 

Google Scholar  More

These authors contributed equally: Gábor Pozsgai, Ibtissem Ben Fekih.State Key Laboratory of Ecological Pest Control for Fujian and Taiwan Crops, Institute of Applied Ecology, Fujian Agriculture and Forestry University, Fuzhou, 350002, ChinaGábor Pozsgai, Ibtissem Ben Fekih, Jie Zhang & Minsheng YouJoint international Research Laboratory of Ecological Pest Control, Ministry of Education, Fuzhou, 350002, ChinaGábor Pozsgai, Gábor L. Lövei & Minsheng YouCE3C – Centre for Ecology, Evolution and Environmental Changes, Azorean Biodiversity Group and Universidade dos Açores, Angra do Heroísmo, 9700-042, Azores, PortugalGábor PozsgaiInstitute of Environmental Microbiology, College of Resources and Environment, Fujian Agriculture and Forestry University, Fuzhou, 350002, ChinaIbtissem Ben Fekih & Christopher RensingBasic Forestry and Proteomics Research Center, College of Life Science, Fujian Provincial Key Laboratory of Haixia Applied Plant Systems Biology, Fujian Agriculture and Forestry University, Fuzhou, 350002, ChinaMarkus V. KohnenLaboratoire de Biologie et de Physiologie des Organismes, Faculté des Sciences Biologiques, Université des Sciences et de la Technologie Houari Boumediène, BP 32 El Alia, Alger, 16111, AlgeriaSaid AmraniDuna-Ipoly National Park Directorate, Költő u. 21, H-1121, Budapest, HungarySándor BércesJuhász-Nagy Pál Doctoral School, University of Debrecen, Egyetem tér 1, H-4032, Debrecen, HungarySándor BércesDepartment of Zoology, Plant Protection Institute, Centre for Agricultural Research, Nagykovácsi út 26-30, H-1029, Budapest, HungaryDávid FülöpFujian University Key Laboratory for Plant-Microbe Interaction, College of Plant Protection, Fujian Agriculture and Forestry University, Fuzhou, 350002, ChinaMohammed Y. M. JaberDepartment of Plant and Environmental Sciences, University of Copenhagen, Thorvaldsensvej 40, 1871, Frederiksberg C, DenmarkNicolai Vitt MeylingDepartment of Algology and Mycology Faculty of Biology and Environmental Protection, University of Łódź, Banacha 12/16, PL-90-237, Łódź, PolandMalgorzata Ruszkiewicz-MichalskaDepartment of Molecular Biotechnology and Microbiology, University of Debrecen, Egyetem tér 1, Debrecen, H-4032, HungaryWalter P. PflieglerFujian Provincial Key Laboratory of Insect Ecology, College of Plant Protection, Fujian Agriculture and Forestry University, Fuzhou, 350002, ChinaFrancisco Javier Sánchez-GarcíaÁrea de Biología Animal, Departamento de Zoología y Antropología Física, Facultad de Veterinaria, Universidad de Murcia, Murcia, 30100, SpainFrancisco Javier Sánchez-GarcíaDepartment of Agroecology, Aarhus University, Flakkebjerg Research Centre, Forsøgsvej 1, DK-4200, Slagelse, DenmarkGábor L. Lövei More

Choice of biowaste treatment plants and sample identifiersCompost samples were collected from four central municipal biowaste treatment plants (denominated as #1 to #4) in Baden-Wurttemberg, Germany (Table 1). All plants used a state-of-the-art two-stage biowaste treatment process comprising of (a) anaerobic digestion/biogas production and (b) subsequent composting of the solid digestate to produce a high-quality mature compost sold for direct use as fertilizer in agriculture. The composts were regularly analyzed by an independent laboratory for quality and residual contamination and consistently fulfilled the quality requirements of the label RAL-GZ 251 Gütezeichen Kompost of the German Bundesgütegemeinschaft Kompost e.V. (www.gz-kompost.de). Plants #1 and #3 produce in addition a liquid fertilizer, which is separated from the solid digestate at the end of stage a) by press filtration and which is also intended for direct use on agricultural soil (replacement of liquid manure). In case of plants #1, #3, and #4 up to 25 wt% of shrub/tree cuttings were added to the solid digestate for composting. All plants used sieving (typically with a 12 or a 20 mm mesh) at the end of the process to assure the necessary purity of their finished composts. Whenever technically possible, we as well took samples of the pre-compost immediately before this final sieving step to evaluate its contribution to the removal of residual BPD fragments. For analysis, composts were passed consecutively through two sieves with mesh sizes of 5 mm and 1 mm, yielding two fragment preparations for IR-analysis namely a > 5 mm fraction corresponding to the contamination by residual “macroplastic” (5 mm is a commonly used upper size limit for “microplastic”, anything larger is macroplastic) and a 1–5 mm fraction corresponding to the regulatory relevant residual contamination by microplastic. The lower limit of 1 mm rather than 2 mm was chosen in anticipation of the expected changes in regulation, where the replacement of the 2 mm limit by a 1 mm limit is imminent.Table 1 Technical data of the investigated plants and incidence of BDP fragments in the sampled composts.Full size tableOccurrence of plastic fragments  > 1 mm in the sampled compostsComposting times of 5–9 weeks were used in the investigated plants (Table 1), which is shorter than the 12 weeks indicated in EN 13432 for the 90% disintegration of a compostable plastic material, but a realistic time span for state-of-the-art technical waste treatment. Since we were not in a position to estimate the quantity of BDP entering the plants, since for technical reasons we were unable to obtain a representative sample, we cannot say, whether any residual BDP detected by us in the finished composts was due to a yet incomplete disintegration process or whether it corresponds to the 10% material still permissible by EN 13432 even after the full composting step. However, in 7 out of the 12 sampled composts and pre-composts fragments with chemical signatures corresponding to the BDPs poly (lactic acid) (PLA) and poly (butylene-adipate-co-terephthalate) (PBAT) were identified in the > 5 mm and/or the 1–5 mm sieving fractions using FTIR analysis3 (Fig. 1; Table 1). All recovered fragments appeared to stem from foils, bags or packaging, since they were thin compared to their length and width (see Suppl Figure S1 for typical examples). Fragments with overlapping signatures, most likely PBAT/PLA mixtures or blends, were also found (see Suppl Figure S2 for the interpretation of the spectra). In addition, the recorded BDP fragment spectra (Fig. 1A) showed high similarity to the FTIR spectra of commercial compostable bags sold in the vicinity of the biowaste treatment plants (Fig. 1B), which together with the geometry of the recovered fragments led us to assuming that the majority of the BDP entered the biowaste in the form of such bags.Figure 1FTIR spectra of BDP fragments from composts and commercial bags. (A) BDP fragments recovered from the composts and (B) the commercial compostable bags. Fragments were coded as follows: p or f for pre-compost or finished compost, followed by the plant number (#1 to #4), an indication of the size fraction ( > 5 mm or 1–5 mm) in which the fragment was found, and finally, the fragment number. Fragment F#1_5mm_4 therefore represents the 4th fragment collected in the  > 5 mm size fraction from the finished compost of plant number 1. Bags were arbitrarily numbered 1–10, see Suppl Table S1 for supplier information. The spectra (in grey) of the reference materials for PLA and PBAT are given as basis for the interpretation. Spectra in red refer to test samples consisting only of PBAT, while those in blue indicate samples composed of PBAT/PLA mixtures.Full size imageThe BDP fragments were found alongside fragments of commodity plastics (mostly PE) in all cases. Finished composts tended to contain fewer and smaller fragments than the corresponding pre-composts. The final sieving of the pre-composts to prepare the finished composts hence appears to be quite effective in removing such fragments, in particular those from the > 5 mm size fraction (Table 1) and for that reason has become state-of-the-art in preparing quality composts (contamination by plastic fragments > 2 mm of less than 0.1 wt%). Given that the size of the fragments is a crucial factor regarding ecological risk, we analyzed the sizes (length Î width) of the BDP fragments in comparison to that of the plastic fragments with signatures of commodity plastics such as PE (Fig. 2). BDP fragments found in a given compost sample tended to be smaller than the fragments stemming from non-BDP materials, which may indicate that BDPs degrade faster or tend to disintegrate into tinier particles than commodity plastics. This may also explain why in the compost from plant #2, no BDP fragments were found in the particle fraction retained by the 5 mm sieve ( > 5 mm fraction), while 19 such particles were found in the fraction then retained by the 1 mm sieve (1–5 mm fraction). Interestingly, plant #2 is the only one included in our study that uses no mechanical breakdown of the incoming biowaste. This reduces the mechanical stress on the incoming material. Mechanical stress can alter the properties of plastic foils such as the crystallinity whereby crystallinity has been shown to influence the biological degradation of BDP such as PLA7.Figure 2Size distribution of plastic fragments  > 1 mm. (A) Fragments found in the finished compost from plant #1, (B) in the finished compost from plant #2, and (C) in the pre-compost from plant #3. For reasons of statistical relevance, only samples containing more than 20 BDP fragments per kg of compost were included in the analysis.Full size imageMaterial characteristics of BDP fragments in comparison to those of commercial biodegradable bagsIn order to verify whether the BDP fragments recovered from the composts differed from the compostable bags in any parameter with possible relevance for biodegradation and environmental impact16, the physico-chemical properties of bags and fragments were studied in detail. Since we wanted to have a maximum of information of the BDP fragments, size/weight was a limiting factor in selecting fragments for analysis. Fragments of at least 1 mg were required for the FT-IR analysis. 5 mg-fragments could be analyzed in addition by 1H-NMR, while the full set of analytics (FT-IR, 1H-NMR, and DSC) required at least 10 mg of sample.For insight into the chemical composition, 1H-NMR spectra of the commercial bags and all suitable BDP fragments were compared (Fig. 3). In case of material mixtures and blends, the 1H-NMR analysis allows quantification of the PBAT/PLA weight ratio in the materials and also of the ratio of the butylene terephthalate (BT) and butylene adipate (BA) units in the involved PBAT polyesters.Figure 31H NMR spectra of BDP fragments from composts and commercial bags. (A) BDP fragments recovered from the composts and (B) the commercial compostable bags. Fragments were coded as follows: p or f for pre-compost or finished compost, followed by the plant number (#1 to #4), an indication of the size fraction ( > 5 mm or 1–5 mm) in which the fragment was found, and finally, the fragment number. Bags were arbitrarily numbered 1–10, see Suppl Table S1 for supplier information. The spectra (in grey) of the reference materials for PLA and PBAT are given as basis for the interpretation. Spectra in red refer to test samples consisting only of PBAT, while those in blue indicate samples composed of PBAT/PLA mixtures. (C) Chemical structures of PLA and PBAT, chemical shifts of the protons are assigned as indicated in the reference spectra in (B).Full size imageThe 1H-NMR spectra corroborate the FTIR measurements in that all investigated commercial bags were made from PBAT/PLA mixtures of varied composition (Table 2). By comparison, some of the fragments, for instance, f#1_5mm_4, appeared to consist of only PBAT. Other fragments, e.g., f#1_1mm_9, were mixtures of PLA and PBAT (Table 2). However, even in the case of PBAT/PLA mixtures, the average PBAT content tended to be higher in the fragments than in the bags, while the BT/BA monomer ratio in the respective PBATs, was also significantly higher in the fragments than in the bags. If we assume the fragments to stem from similar compostable bags as the ones included in our comparison, this would mean that during composting of such a bag, the PLA degrades more quickly than the PBAT, whereas within a given PBAT polyester, the BA unit is more easily degraded than the BT unit. Evidence can indeed be found in the pertinent literature that PLA has faster biodegradation kinetics than PBAT, while BT is more resistant to mineralization than BA17,18.Table 2 Composition of commercial compostable bags and BDP fragments recovered from the composts as analyzed by 1H-NMR.Full size tableNext, differential scanning calorimetry (DSC) was used to analyze fragments compared to commercial bags in regard to the presence of amorphous vs. crystalline domains, a parameter expected to affect biodegradation kinetics and therefore the putative environmental impact of the produced microplastic16 upon release into the environment with the composts. Whereas amorphous domains show glass transition, crystalline domains show melting, both of which can be discerned by the respective phase transition enthalpy in the DSC curves (Fig. 4).Figure 4DSC curves of BDP fragments and compostable bags #1 and #7. Curves for the reference materials (in grey) for PLA and PBAT are given for comparison. Curves were recorded during the first heating run (temperature range: − 50 °C to 200 °C, heating rate: 10 °C min−1). (A) and (B) curves in red refer to test samples consisting only of PBAT, while those in blue indicate samples composed of PBAT/PLA mixtures. Fragments were coded as follows: p or f for pre-compost or finished compost, followed by the plant number (#1 to #4), an indication of the size fraction ( > 5 mm or 1–5 mm) in which the fragment was found, and finally, the fragment number.Full size imageThe curve for the reference PBAT shows a glass transition temperature (Tg) of − 29 °C and a broad melting range between 100 and 140 °C for the crystalline domains, while that of the PLA reference shows a glass transition temperature of 58 °C and a narrower melting peak between 144 °C and 162 °C. The curve for commercial bag #1, which had a comparatively high PLA content, shows a pronounced melting peak in the expected range; the same is the case for fragment p#3_5mm_1 and to a lesser extent for fragment p#3_5mm_9, two fragments, which also have high PLA contents. The DSC curves of the other fragments and bag #1 are undefined in comparison, which is due to their high PBAT content. According to the DSC curves, most of the investigated materials are semicrystalline, i.e., contain both amorphous (glass transition) and crystalline (melting) domains. However, the DCS data alone allow only a qualitative discussion of the differences between fragments and bags.To obtain quantitative data on the crystallinity differences, wide angle X-ray scattering (WAXS) spectra were recorded. WAXS requires fragments at least 3 cm long, which restricted the number of fragment samples to three, all of which were found in pre-compost samples. The corresponding curves are shown in Fig. 5A–C. The spectra of the commercial biodegradable bags are shown in Suppl Figure S3. Foils were in addition prepared by heat pressing from the reference materials for PLA and PBAT in order to include them into the WAXS measurements (Fig. 5D). While the foils produced from the PBAT reference material produced crystallinity peaks at 16.2°, 17.3°, 20.4°, 23.2°, and 24.8°, the foil prepared from the PLA reference material showed only an amorphous halo at 15.5° and 31.5°, which is in accordance with values published in the literature19. A more pronounced crystallinity peak was obtained in the case of an additionally annealed PLA foil.Figure 5WAXS curves with Lorenz fitting for (A) fragment p#3_5mm_1, (B) fragment p#3_5mm_9, and (C) fragment p#4_5mm_2. (D) WAXS curves for foils produced from the PBAT and PLA reference materials; the percent values indicate the crystallinity. The dash lines are the fitting peak curves for the XRD spectrum. Crystallinity can be obtained by dividing the integration area of the fitted peaks by the integration area of the entire spectrum. Fragments were coded as follows: p or f for pre-compost or finished compost, followed by the plant number (#1 to #4), an indication of the size fraction ( > 5 mm or 1–5 mm) in which the fragment was found, and finally, the fragment number.Full size imageIn case of the fragments and bags, the peaks of PLA and PBAT overlapped to some extent in the WAXS spectra, but by conducting Lorenz fitting using Origin software, the overall crystallinity could be calculated as follows:$$chi = { 1}00% , *{text{ Aa}}/left( {{text{Aa }} + {text{ Ac}}} right)$$where χ is the crystallinity and Aa and Ac represent the areas of the amorphous and crystalline peaks.Using this equation, crystallinities of 55% (fragments p#3_5mm_1), 34% (p#3_5mm_9), and 34% (p#4_5mm_2) were calculated for the fragments. The foils prepared in house for the reference materials had similar crystallinities (43% in case of the annealed PLA foil and 26% of the PBAT foil), while the simple PLA foil was amorphous. By comparison, for eight of the commercial bags, crystallinities in the range from 1% to 7% were calculated, whereas these values were 14% and 15% for the remaining two bag types (Suppl Figure S3).The high crystallinity of the larger fragments recovered from the pre-compost samples suggests that crystalline domains of BDP materials may indeed disintegrate more slowly than the amorphous ones, as prior studies on microbial biodegradation have suggested7,8. Admittedly, such large fragments per se would not enter the environment, since the final sieving step used to prepare the finished composts is quite efficient at removing them. However, it is tempting to extrapolate that residual BDP in general are remnants of the more crystal domains of the original material, even though experimental proof of this assumption is at present not possible. 10 wt% of a BDP bag is allowed to remain after standard composting. It is usually assumed that any such residues continue to degrade with comparable speed. However, should these residues correspond to the more crystalline domains, rather than degrading with similar speed as the bulk material, the more crystalline fragments can be expected to persist for a much longer and at present unpredictable length of time in the environment, e.g. when applied to the soil with the composts; in particular, when they are also enriched in PBAT and BT units as suggested by our analysis of the chemical composition. Data from the use of biodegradable foils in agriculture show that the degradation in the environment may take years20. Altogether this may have unforeseen economic and environmental consequences, especially when considering the high fraction of BDP fragments < 5 mm. Putative consequences include changes in soil properties, the soil microbiome and therefore in plant performance21, a factor indispensable for worldwide nutrition.Residues of BDP fragments  1 mm were found in the collected LF samples. This is hardly surprising, given that the LF is produced by press filtration of the digestate after the anaerobic stage. Such a filtration step can be expected to retain fragments > 1 mm in the produced filter cake, which goes into the composting step, leaving the filtrate, i.e. the LF, essentially free of such particles. Anaerobic digestion is currently not assumed to contribute significantly to the degradation of BDP17,22, but the process conditions (mixing, pumping) may promote breakdown of larger fragments, particularly when additives such as plasticizers23 leach out of the material.Since the residual solids content of the LF is low (plant #1: 8.6 wt%, plant #3: 5.8 wt%), a combination of enzymatic-oxidative treatment and µFTIR imaging originally developed for environmental samples from aqueous systems24,25 could be adapted for the analysis (size and chemical signature) of particles in the LF down to a size of 10 µm. The corresponding data are compiled in Table 3. In all cases, residual fragments from PBAT-based polymers represented the dominant plastic fraction in the investigated samples; i.e. approximately 53% of all plastic particles in the LF from plant #1 (11,520 BDP particles per liter) and 65% in the case of plant #3 (12,480 BDP particles per liter). Liquid manure is applied several times a year to fields at a concentration of 2–3 L m−2. According to our analysis > 20,000 BDP microparticles of a size ranging from 10 µm to 500 µm enter each m2 of agricultural soil whenever LF is applied on agricultural surfaces.Table 3 Microplastic fragments (BDP/all) found per liter of liquid fertilizer.Full size tableDue to the complexity of the matrix, a similar analysis of individual plastic fragments  1 mm. Six compost samples representing the more contaminated ones based on the content of fragments > 1 mm, namely, f#1, f#2, p#3, f#3, p#4 and f#4 (nomenclature: f or p for finished or pre-compost, followed by plant number), were extracted with a 90/10 vol% chloroform/methanol mixture. The amounts of PBAT and PLA in the obtained extracts were then quantified via 1H-NMR (Table 4). Briefly, the intensity of characteristic signals in the extract spectra of the compost samples (see Suppl Figure S4) were compared to peak intensities produced by calibration standards of the pure polymer dissolved at a known concentration in the chloroform/methanol. All samples and standards were normalized using the 1,2-dichloroethan signal at 3.73 ppm as internal standard. See also Suppl Figure S5 for an exemplification of the quantification of the PBAT/PLA ratios. Based on the amounts of PBAT and PLA extracted from a known amount of compost, the total mass concentration (wt% dry weight) of these polymers in the composts was calculated.Table 4 Evidence of PBAT and PLA residues caused by fragments  2 mm. Moreover, residues of PBAT and PLA were found in all investigated compost samples, including the finished compost from plant #4, which had shown no contamination by larger BPD fragments (Table 1). The pre-compost from that plant had shown a few contaminating BDP fragments in the > 5 mm fraction. However, in regard to the fragments More

Department of Environmental Science, Policy and Management, UC Berkeley, Berkeley, CA, USAT. F. Keenan, X. Luo, Y. Zhang & S. ZhouClimate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USAT. F. Keenan, X. Luo, Y. Zhang & S. ZhouDepartment of Geography, National University of, Singapore, SingaporeX. LuoARC Centre of Excellence for Climate Extremes, Sydney, New South Wales, AustraliaM. G. De KauweClimate Change Research Centre, University of New South Wales, Sydney, New South Wales, AustraliaM. G. De KauweSchool of Biological Sciences, University of Bristol, Bristol, UKM. G. De KauweHawkesbury Institute for the Environment, Western Sydney University, Penrith, New South Wales, AustraliaB. E. MedlynDepartment of Life Sciences, Imperial College London, Ascot, UKI. C. PrenticeDepartment of Biological Sciences, Macquarie University, North Ryde, New South Wales, AustraliaI. C. PrenticeDepartment of Earth System Science, Tsinghua University, Haidian, Beijing, ChinaI. C. Prentice & H. WangDepartment of Environmental Systems Science, ETH, Zurich, SwitzerlandB. D. StockerSwiss Federal Institute for Forest, Snow and Landscape Research WSL, Birmensdorf, SwitzerlandB. D. StockerDepartment of Biological Sciences, Texas Tech University, Lubbock, TX, USAN. G. SmithPhysical and Life Sciences Directorate, Lawrence Livermore National Laboratory, Livermore, CA, USAC. TerrerDepartment of Civil and Environmental Engineering, Massachusetts Institute of Technology, Boston, MA, USAC. TerrerSino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing, ChinaY. ZhangLamont-Doherty Earth Observatory of Columbia University, Palisades, NY, USAS. ZhouEarth Institute, Columbia University, New York, NY, USAS. ZhouDepartment of Earth and Environmental Engineering, Columbia University, New York, NY, USAS. ZhouState Key Laboratory of Earth Surface Processes and Resources Ecology, Faculty of Geographical Science, Beijing Normal University, Beijing, ChinaS. Zhou More

Rendell, L. & Whitehead, H. Culture in whales and dolphins. Behav. Brain Sci. 24, 309–324 (2001).CAS 
Article 

Google Scholar 
Krützen, M. et al. Cultural transmission of tool use in bottlenose dolphins. Proc. Natl. Acad. Sci. U.S.A. 102, 8939–8943 (2005).ADS 
Article 

Google Scholar 
Kawai, M. Newly-acquired pre-cultural behavior of the natural troop of Japanese monkeys on Koshima Islet. Primates 6, 1–30 (1965).Article 

Google Scholar 
Slater, P. The cultural transmission of bird song. Trends Ecol. Evol. 1, 94–97 (1986).CAS 
Article 

Google Scholar 
Whitehead, H. & Rendell, L. The cultural lives of whales and dolphins. (University of Chicago Press, 2014).Whiten, A. The identification and differentiation of culture in chimpanzees and other animals: from natural history to diffusion experiments. The question of animal culture, 99–124 (2009).Allen, J. A. Community through culture: from insects to whales: How social learning and culture manifest across diverse animal communities. BioEssays 41, 1900060 (2019).Article 

Google Scholar 
Allen, J., Weinrich, M., Hoppitt, W. & Rendell, L. Network-based diffusion analysis reveals cultural transmission of lobtail feeding in humpback whales. Science 340, 485–488 (2013).ADS 
CAS 
Article 

Google Scholar 
Baker, C. Migratory movement and population structure of humpback whales (Megaptera novaeangliae) in the central and eastern North Pacific. Mar Ecol Prog Ser 31, 105–119 (1986).ADS 
Article 

Google Scholar 
Garrigue, C. et al. Movement of individual humpback whales between wintering grounds of Oceania (South Pacific), 1999 to 2004. J. Cetacean Res. Manage 3, 275–281 (2011).
Google Scholar 
Rosenbaum, H. C. et al. First circumglobal assessment of Southern Hemisphere humpback whale mitochondrial genetic variation and implications for management. Endang. Spec. Res. 32, 551–567 (2017).Article 

Google Scholar 
Noad, M. J., Cato, D. H., Bryden, M. M., Jenner, M. N. & Jenner, K. C. Cultural revolution in whale songs. Nature 408, 537. https://doi.org/10.1038/35046199 (2000).ADS 
CAS 
Article 
PubMed 

Google Scholar 
Payne, R. S. & McVay, S. Songs of humpback whales. Science 173, 585–597 (1971).ADS 
CAS 
Article 

Google Scholar 
Garland, E. C. et al. Dynamic horizontal cultural transmission of humpback whale song at the ocean basin scale. Curr. Biol. 21, 687–691. https://doi.org/10.1016/j.cub.2011.03.019 (2011).CAS 
Article 
PubMed 

Google Scholar 
Payne, R. & Guinee, L. N. Humpback whale (Megaptera novaeangliae) songs as an indicator of “stocks”. Communication and behavior of whales, 333–358 (1983).Garrigue, C. et al. Movements of humpback whales in Oceania, South Pacific. J. Cetac. Res. Manage. 4, 255–260 (2002).
Google Scholar 
Derville, S., Torres, L. G., Zerbini, A. N., Oremus, M. & Garrigue, C. Horizontal and vertical movements of humpback whales inform the use of critical pelagic habitats in the western South Pacific. Sci. Rep. 10, 1–13 (2020).Article 

Google Scholar 
Garrigue, C. et al. First assessment of interchange of humpback whales between Oceania and the East coast of Australia. J. Cetac. Res. Manage. 3, 269–274 (2011).
Google Scholar 
Steel, D. et al. Migratory connections between humpback whales from South Pacific breeding grounds and Antarctic feeding areas based on genotype matching. Int. Whal. Comm. (2008).Constantine, R., Russell, K., Gibbs, N., Childerhouse, S. & Baker, C. S. Photo-identification of humpback whales (Megaptera novaeangliae) in New Zealand waters and their migratory connections to breeding grounds of Oceania. Mar. Mam. Sci. 23, 715–720 (2007).Article 

Google Scholar 
Garland, E. C. et al. Humpback whale song on the southern ocean feeding grounds: implications for cultural transmission. PLoS ONE 8, e79422 (2013).ADS 
Article 

Google Scholar 
Garland, E. C. et al. Population structure of humpback whales in the western and central South Pacific Ocean as determined by vocal exchange among populations. Conserv. Biol. 29, 1198–1207 (2015).Article 

Google Scholar 
Cholewiak, D. M., Sousa-Lima, R. S. & Cerchio, S. Humpback whale song hierarchical structure: Historical context and discussion of current classification issues. Mar. Mam. Sci. 29, E312–E332. https://doi.org/10.1111/mms.12005 (2013).Article 

Google Scholar 
Payne, K., Tyack, P. & Payne, R. Progressive changes in the songs of humpback whales (Megaptera novaeangliae): a detailed analysis of two seasons in Hawaii. Communication and behavior of whales, 9–57 (1983).Payne, K. & Payne, R. Large scale changes over 19 years in songs of humpback whales in Bermuda. Ethology 68, 89–114 (1985).
Google Scholar 
Allen, J. A., Garland, E. C., Dunlop, R. A. & Noad, M. J. Cultural revolutions reduce complexity in the songs of humpback whales. Proc. R. Soc. B: Biol. Sci. 285, 20182088. https://doi.org/10.1098/rspb.2018.2088 (2018).Article 

Google Scholar 
Allen, J. A., Garland, E. C., Murray, A., Noad, M. J. & Dunlop, R. Using self-organizing maps to classify humpback whale song units and quantify their similarity. J. Acoust. Soc. Am. 142, 1943–1952 (2017).ADS 
Article 

Google Scholar 
Murray, A., Dunlop, R. A., Noad, M. J. & Goldizen, A. W. Stereotypic and complex phrase types provide structural evidence for a multi-message display in humpback whales (Megaptera novaeangliae). J Acoust Soc Am. 143, 980–994 (2018).ADS 
Article 

Google Scholar 
Garland, E. C. et al. Redefining western and central South Pacific humpback whale population structure based on vocal cultural exchange. (2013).Rekdahl, M. Humpback whale vocal communication: Use and stability of social calls and revolutions in the songs of east Australian whales. (2012).Templeton, C. N., Laland, K. N. & Boogert, N. J. Does song complexity correlate with problem-solving performance in flocks of zebra finches?. Anim. Behav. 92, 63–71 (2014).Article 

Google Scholar 
Boogert, N. J., Giraldeau, L.-A. & Lefebvre, L. Song complexity correlates with learning ability in zebra finch males. Anim. Behav. 76, 1735–1741 (2008).Article 

Google Scholar 
Winn, H. & Winn, L. The song of the humpback whale Megaptera novaeangliae in the West Indies. Mar. Biol. 47, 97–114 (1978).Article 

Google Scholar 
Girola, E., Noad, M. J., Dunlop, R. A. & Cato, D. H. Source levels of humpback whales decrease with frequency suggesting an air-filled resonator is used in sound production. The Journal of the Acoustical Society of America (In Review).Warren, V. E., Constantine, R., Noad, M., Garrigue, C. & Garland, E. C. Migratory insights from singing humpback whales recorded around central New Zealand. R. Soc. Open Sci. 7, 201084 (2020).ADS 
Article 

Google Scholar 
Garland, E. C. et al. Quantifying humpback whale song sequences to understand the dynamics of song exchange at the ocean basin scale. J. Acoust. Soc. Am. 133, 560–569. https://doi.org/10.1121/1.4770232 (2013).ADS 
Article 
PubMed 

Google Scholar 
Owen, C. et al. Migratory convergence facilitates cultural transmission of humpback whale song. R. Soc. Open Sci. 6, 190337 (2019).ADS 
Article 

Google Scholar 
Garland, E. C., Rendell, L., Lamoni, L., Poole, M. M. & Noad, M. Song hybridization events during revolutionary song change provide insights into cultural transmission in humpback whales. Proc. Natl. Acad. Sci. 114, 7822–7829 (2017).CAS 
Article 

Google Scholar 
Noad, M. & Cato, D. A combined acoustic and visual survey of humpback whales off southeast Queensland. Mem. Qld. Mus. 47, 507–523 (2001).
Google Scholar 
Spierings, M., de Weger, A. & Ten Cate, C. Pauses enhance chunk recognition in song element strings by zebra finches. Anim. Cogn. 18, 867–874 (2015).Article 

Google Scholar 
Doupe, A. J. & Kuhl, P. K. Birdsong and human speech: common themes and mechanisms. Annu. Rev. Neurosci. 22, 567–631 (1999).CAS 
Article 

Google Scholar 
Allen, J. A., Garland, E. C., Dunlop, R. A. & Noad, M. J. Network analysis reveals underlying syntactic features in a vocally learnt mammalian display, humpback whale song. Proc. R. Soc. B 286, 20192014 (2019).Article 

Google Scholar 
Barón Birchenall, L. Animal communication and human language: An overview. International Journal of Comparative Psychology 29 (2016).Noad, M. J. The use of song by humpback whales (Megaptera novaeangliae) during migration off the east coast of Australia (doctoral dissertation) Doctor of Philosophy thesis, University of Sydney, (2002).Catchpole, C. Song and female choice: good genes and big brains?. Trends Ecol. Evol. 11, 358–360. https://doi.org/10.1016/0169-5347(96)30042-6 (1996).CAS 
Article 
PubMed 

Google Scholar 
Nowicki, S., Hasselquist, D., Bensch, S. & Peters, S. Nestling growth and song repertoire size in great reed warblers: evidence for song learning as an indicator mechanism in mate choice. Proc. R. Soc. London B: Biol. Sci. 267, 2419–2424 (2000).CAS 
Article 

Google Scholar 
NOAA. Vol. 81 (ed National Marine Fisheries Service) 62260–62320 (Department of Commerce, Federal Register, 2016).Noad, M. J., Kniest, E. & Dunlop, R. A. Boom to bust? Implications for the continued rapid growth of the eastern Australian humpback whale population despite recovery. Popul. Ecol. 61(2), 198–209 (2019).Article 

Google Scholar 
Garrigue, C., Albertson, R. & Jackson, J. A. An anomalous increase in the New Caledonia humpback whales breeding sub-stock E2. Scientific Committee of the International Whaling Commission, Paper (2012).Garland, E. C. & McGregor, P. K. Cultural transmission, evolution, and revolution in vocal displays: Insights from bird and whale song. Front. Psychol. 11, 2387 (2020).Article 

Google Scholar 
Zandberg, L., Lachlan, R. F., Lamoni, L. & Garland, E. C. Global cultural evolutionary model of humpback whale song. Philos. Trans. R. Soc. B 376, 20200242 (2021).Article 

Google Scholar 
Crates, R. et al. Loss of vocal culture and fitness costs in a critically endangered songbird. Proc. R. Soc. B 288, 20210225 (2021).Article 

Google Scholar 
Garland, E. C., Garrigue, C. & Noad, M. J. When does cultural evolution become cumulative culture? A case study of humpback whale song. Philos. Trans. R. Soc. B 377, 20200313 (2022).Article 

Google Scholar 
Garland, E. C. et al. Improved versions of the Levenshtein distance method for comparing sequence information in animals’ vocalisations: tests using humpback whale song. Behaviour 149, 1413–1441 (2012).Article 

Google Scholar 
Garland, E. C. et al. The devil is in the detail: quantifying vocal variation in a complex, multileveled, and rapidly evolving display. J. Acoust. Soc. Am. 142, 460–472 (2017).ADS 
Article 

Google Scholar 
Rekdahl, M. L. et al. Culturally transmitted song exchange between humpback whales (Megaptera novaeangliae) in the southeast Atlantic and southwest Indian Ocean basins. R. Soc. Open Sci. 5, 172305 (2018).ADS 
Article 

Google Scholar 
Suzuki, R. & Shimodaira, H. Pvclust: an R package for assessing the uncertainty in hierarchical clustering. Bioinformatics 22, 1540–1542 (2006).CAS 
Article 

Google Scholar 
Sokal, R. R. & Rohlf, F. J. The comparison of dendrograms by objective methods. Taxon 11(2), 33–40 (1962).Article 

Google Scholar 
R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. 2015. URL http (R Core Development Team, 2016). More

Friedlingstein, P. et al. Global carbon budget 2020. Earth Syst. Sci. Data 12, 3269–3340 (2020).
Google Scholar 
Myneni, R. B. et al. A large carbon sink in the woody biomass of northern forests. Proc. Natl Acad. Sci. USA 98, 14784–14789 (2001).CAS 

Google Scholar 
Sitch, S. et al. Recent trends and drivers of regional sources and sinks of carbon dioxide. Biogeosciences 12, 653–679 (2015).
Google Scholar 
Kauppi, P. E. et al. Large impacts of climatic warming on growth of boreal forests since 1960. PLoS ONE 9, e111340 (2014).
Google Scholar 
Zhu, Z. C. et al. Greening of the Earth and its drivers. Nat. Clim. Change 6, 791–795 (2016).CAS 

Google Scholar 
Piao, S. L. et al. Characteristics, drivers and feedbacks of global greening. Nat. Rev. Earth Environ. 1, 14–27 (2020).
Google Scholar 
IPCC Climate Change 2013: The Physical Science Basis (eds Stocker, T. F. et al.) (Cambridge Univ. Press, 2013).Penuelas, J. et al. Shifting from a fertilization-dominated to warming-dominated period. Nat. Ecol. Evol. 1, 1438–1445 (2017).
Google Scholar 
D’Arrigo, R., Wilson, R., Liepert, B. & Cherubini, P. On the ‘divergence problem’ in northern forests: a review of the tree-ring evidence and possible causes. Glob. Planet. Change 60, 289–305 (2008).
Google Scholar 
Beck, P. S. A. et al. Changes in forest productivity across Alaska consistent with biome shift. Ecol. Lett. 14, 373–379 (2011).
Google Scholar 
Vickers, H. et al. Changes in greening in the High Arctic: insights from a 30-year AVHRR max NDVI dataset for Svalbard. Environ. Res. Lett. 11, 105004 (2016).
Google Scholar 
Piao, S. L. et al. Evidence for a weakening relationship between interannual temperature variability and northern vegetation activity. Nat. Commun. 5, 5018 (2014).CAS 

Google Scholar 
Duffy, K. A. et al. How close are we to the temperature tipping point of the terrestrial biosphere? Sci. Adv. 7, eaay1052 (2021).CAS 

Google Scholar 
Liu, Y. W. et al. Seasonal responses of terrestrial carbon cycle to climate variations in CMIP5 models: evaluation and projection. J. Clim. 30, 6481–6503 (2017).
Google Scholar 
Huang, M. T. et al. Air temperature optima of vegetation productivity across global biomes. Nat. Ecol. Evol. 3, 772–779 (2019).
Google Scholar 
Park, T. et al. Changes in timing of seasonal peak photosynthetic activity in northern ecosystems. Glob. Change Biol. 25, 2382–2395 (2019).
Google Scholar 
Keeling, C. D., Chin, J. F. S. & Whorf, T. P. Increased activity of northern vegetation inferred from atmospheric CO2 measurements. Nature 382, 146–149 (1996).CAS 

Google Scholar 
Piao, S. L., Friedlingstein, P., Ciais, P., Viovy, N. & Demarty, J. Growing season extension and its impact on terrestrial carbon cycle in the Northern Hemisphere over the past 2 decades. Glob. Biogeochem. Cycles 3, GB3018 (2007).
Google Scholar 
Keenan, T. F. et al. Net carbon uptake has increased through warming-induced changes in temperate forest phenology. Nat. Clim. Change 4, 598–604 (2014).CAS 

Google Scholar 
Xia, J. Y. et al. Joint control of terrestrial gross primary productivity by plant phenology and physiology. Proc. Natl Acad. Sci. USA 112, 2788–2793 (2015).CAS 

Google Scholar 
Eyring, V. et al. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev. 9, 1937–1958 (2016).
Google Scholar 
Meehl, G. A. & Tebaldi, C. More intense, more frequent, and longer lasting heat waves in the 21st century. Science 305, 994–997 (2004).CAS 

Google Scholar 
Yamori, W., Hikosaka, K. & Way, D. A. Temperature response of photosynthesis in C3, C4, and CAM plants: temperature acclimation and temperature adaptation. Photosynth. Res. 119, 101–117 (2014).CAS 

Google Scholar 
Berry, J. & Bjorkman, O. Photosynthetic response and adaptation to temperature in higher plants. Annu. Rev. Plant Physiol. Plant Mol. Biol. 31, 491–543 (1980).
Google Scholar 
Chen, A., Huang, L., Liu, Q. & Piao, S. L. Optimal temperature of vegetation productivity and its linkage with climate and elevation on the Tibetan Plateau. Glob. Change Biol. 27, 1942–1951 (2021).
Google Scholar 
Smith, N. G., Lombardozzi, D., Tawfik, A., Bonan, G. & Dukes, J. S. Biophysical consequences of photosynthetic temperature acclimation for climate. J. Adv. Model. Earth Syst. 9, 536–547 (2017).
Google Scholar 
Chen, M. & Zhuang, Q. L. Modelling temperature acclimation effects on the carbon dynamics of forest ecosystems in the conterminous United States. Tellus B 65, 19156 (2013).
Google Scholar 
Crous, K. Y. Plant responses to climate warming: physiological adjustments and implications for plant functioning in a future, warmer world. Botany 106, 1049–1051 (2019).
Google Scholar 
Conley, M. M. et al. CO2 enrichment increases water-use efficiency in sorghum. New Phytol. 151, 407–412 (2001).
Google Scholar 
Norby, R. J. & Zak, D. R. Ecological lessons from free-air CO2 enrichment (FACE) experiments. Annu. Rev. Ecol. Evol. Syst. 42, 181–203 (2011).
Google Scholar 
Huang, M. T. et al. Change in terrestrial ecosystem water-use efficiency over the last three decades. Glob. Change Biol. 21, 2366–2378 (2015).
Google Scholar 
Gonsamo, A. et al. Greening drylands despite warming consistent with carbon dioxide fertilization effect. Glob. Change Biol. 7, 3336–3349 (2021).
Google Scholar 
Keenan, T. F. et al. Increase in forest water-use efficiency as atmospheric carbon dioxide concentrations rise. Nature 499, 324–327 (2013).CAS 

Google Scholar 
Lemordant, L. et al. Modification of land–atmosphere interactions by CO2 effects: implications for summer dryness and heat wave amplitude. Geophys. Res. Lett. 43, 10240–10248 (2016).CAS 

Google Scholar 
Lian, X. et al. Summer soil drying exacerbated by earlier spring greening of northern vegetation. Sci. Adv. 6, eaax0255 (2020).
Google Scholar 
Lian, X. et al. Multifaceted characteristics of dryland aridity changes in a warming world. Nat. Rev. Earth Environ. 4, 232–250 (2021).
Google Scholar 
Druel, A., Ciais, P., Krinner, G. & Peylin, P. Modeling the vegetation dynamics of northern shrubs and mosses in the ORCHIDEE land surface model. J. Adv. Model. Earth Syst. 11, 2020–2035 (2019).
Google Scholar 
Scheiter, S., Langan, L. & Higgins, S. I. Next-generation dynamic global vegetation models: learning from community ecology. New Phytol. 198, 957–969 (2013).
Google Scholar 
Mod, H. K. & Luoto, M. Arctic shrubification mediates the impacts of warming climate on changes to tundra vegetation. Environ. Res. Lett. 12, 124028 (2016).
Google Scholar 
Zhang, Y., Commane, R., Zhou, S., Williams, A. P. & Gentine, P. Light limitation regulates the response of autumn terrestrial carbon uptake to warming. Nat. Clim. Change 10, 739–743 (2020).CAS 

Google Scholar 
Bauerle, W. L. et al. Photoperiodic regulation of the seasonal pattern of photosynthetic capacity and the implications for carbon cycling. Proc. Natl Acad. Sci. USA 109, 8612–8617 (2012).CAS 

Google Scholar 
Zhang, Y., Parazoo, N. C., Williams, A. P., Zhou, S. & Gentine, P. Large and projected strengthening moisture limitation on end-of-season photosynthesis. Proc. Natl Acad. Sci. USA 117, 9216–9222 (2020).CAS 

Google Scholar 
Fritz, M. et al. Brief communication: future avenues for permafrost science from the perspective of early career researchers. Cryosphere 9, 1715–1720 (2015).
Google Scholar 
Jin, X. Y. et al. Impacts of climate-induced permafrost degradation on vegetation: a review. Adv. Clim. Change Res. 12, 29–47 (2021).
Google Scholar 
Tramontana, G. et al. Predicting carbon dioxide and energy fluxes across global FLUXNET sites with regression algorithms. Biogeosciences 13, 4291–4313 (2016).CAS 

Google Scholar 
Zhang, Y., Joiner, J., Alemohammad, S. H., Zhou, S. & Gentine, P. A global spatially contiguous solar-induced fluorescence (CSIF) dataset using neural networks. Biogeosciences 15, 5779–5800 (2018).CAS 

Google Scholar 
Navarro-Racines, C., Tarapues, J., Thornton, P., Jarvis, A. & Ramirez-Villegas, J. High-resolution and bias-corrected CMIP5 projections for climate change impact assessments. Sci. Data 7, 7 (2020).
Google Scholar  More

Illustration of the compound event indicesBuilding on earlier studies24,25, we develop two univariate indices to model concurrent climate conditions, i.e., a CWD index that varies from compound cold-wet conditions to CWD conditions, and a CCD index that varies from compound warm-wet conditions to CCD conditions (see “Methods”). The two indices incorporate the dependence between temperature and precipitation and are a measure of how warm/cold and dry a point is relative to the distribution of climate conditions at a given location. We illustrate the two indices on two grid points that have strong but opposite temperature-precipitation correlation. In the case where temperature and precipitation are strongly negatively correlated, the CWD index is well aligned with the primary axis of the bivariate distribution (Fig. 1a). In the case where temperature and precipitation are strongly positively correlated, the same holds for the CCD index (Fig. 1d). As illustrated for several concurrent hot-dry and cold-dry events that occurred around the globe, the two indices well capture these events (Supplementary Figs. 1 and 2).Fig. 1: The relationship between precipitation and temperature and compound indices.a Scatter plot of summer precipitation and temperature anomalies (z-score) with corresponding CWD index in color (see “Methods”). The location is at 97.25°W and 33.75°N from 1901 to 2018. b The same as a but for spring at 84.75°E and 66.75°N. c Same distribution as in a but colored based on the CCD index. d Same distribution as in b but colored based on the CCD index.Full size imageNotably, in the case where precipitation and temperature are strongly positively correlated, the CWD index indicates the relative anomalies of bivariate joint distribution, and some counterintuitive situations might occur relative to the univariate marginals (Fig. 1b). For instance, points might be labeled as strong CWD events (CWD index > 1.5) even though temperature is anomalously cold (temperature anomalies < 0, red dots in lower left quadrant of Fig. 1b). The CCD index exhibits similar behavior (Fig. 1c). This indicates an interesting property of the compound indices to identify strong compound conditions relative to bivariate distribution that are not necessarily extreme from a univariate perspective3,24,26,27.Widespread direct and lagged impacts of springtime compound climate conditionsTo evaluate the lagged summer vegetation responses to spring compound climate conditions, we compute partial correlation between CWD (CCD) spring and subsequent summer vegetation variation by controlling for the influence of summer compound climate conditions on these correlations (see “Methods”). Results show widespread negative associations between CWD spring and subsequent summer vegetation in the mid-latitudes (50°N).a–c The average standardized anomalies (z-score) of GPP during CWD spring but subsequent non-CWD summer (a), non-CWD spring but subsequent CWD summer (b), and consecutive CWD spring and summer (c) for areas in Fig. 2a where summer vegetation responds positively (r ≥ 0.22) to spring CWD climate conditions. d–f The same as a–c, but for soil moisture. g–i The same as a–c, but for runoff. The bar plots with dash lines (without dash line) indicate the average anomalies of multiple observation-based (model-based) products, and the circles indicate the average anomalies of each product. GLASS, LUE, NIRv, Flux-CRU, and Flux-ERA5 are observation-based GPP products, while model simulations are taken from TRENDYv6. GLEAM is observation-based soil moisture. GRUN represents observation-based runoff. GLDAS-VIC, GLDAS-Noah, GLDAS-Catchment, and FLDAS indicate assimilatory soil moisture and runoff that incorporate satellite- and ground-based observational products.Full size imageFig. 4: The responses of vegetation productivity and hydrological variables to CWD events in mid-latitudes (23.5–50°N/S).a–c The average standardized anomalies (z-score) of GPP during CWD spring but subsequent non-CWD summer (a), non-CWD spring but subsequent CWD summer (b), and consecutive CWD spring and summer (c) for areas in Fig. 2a where summer vegetation responds negatively (r ≤ −0.22) to spring CWD climate conditions. d–f The same as a–c, but for soil moisture. g–i The same as a–c, but for runoff. The bar plots with dash lines (without dash line) indicate the average anomalies of multiple observation-based (model-based) products, and the circles indicate the average anomalies of each product. For details on data see Fig. 3.Full size imageFig. 5: The effects of CCD events on vegetation productivity and hydrological variables in mid-to-high latitudes.a–c The average standardized anomalies (z-score) of GPP during CCD spring but subsequent non-CCD summer (a), non-CCD spring but subsequent CCD summer (b), and consecutive CCD spring and summer (c) for areas in Fig. 2b where summer vegetation responds negatively (r ≤ −0.22) to spring CCD climate conditions. d–f The same as a–c, but for soil moisture. g–i The same as a–c, but for runoff. The bar plots with dash lines (without dash line) indicate the average anomalies of multiple observation-based (model-based) products, and the circles indicate the average anomalies of each product. For details on data see Fig. 3.Full size imageCWD events increase vegetation productivity in high latitudesWe first analyze the direct responses of productivity to springtime and summertime CWD events across high latitudes ( >50°N, Fig. 3). Productivity increases during CWD spring and summer (Fig. 3a–c), which is consistent with vegetation responses (Supplementary Fig. 8a–c). Despite elevated spring greenness, spring water overall shows positive anomalies during CWD spring (Fig. 3d, f, g, i). This result indicates that spring greenness during CWD conditions is not associated with dry spring across high latitudes, which is further confirmed by similar anomalies in springtime TWS (Supplementary Fig. 8d, f). In contrast, severe water reduction is found in CWD summer (Fig. 3e, f, h, i). This suggests that despite the beneficial effects of CWD events on productivity in summer, they are associated with summer water deficit.Next, to analyze the lagged effects of springtime CWD events, we investigate the productivity anomalies in summer under three cases, namely CWD spring but non-CWD summer, non-CWD spring but CWD summer, and consecutive CWD spring and summer. Our results indicate that springtime CWD events have positive lagged effects on summer productivity across high latitudes (Fig. 3). Specifically, we find that during non-CWD summer (that is not favorable for summer vegetation growth) preceded by CWD spring, positive anomalies are still found in summer productivity (Fig. 3a). In contrast, during CWD summer (preceded by non-CWD spring), some models and observation-based products exhibit a reduction in summer productivity (Fig. 3b). We further find that summer productivity highly increases during consecutive events (Fig. 3c). Vegetation anomalies show similar behaviors (Supplementary Fig. 8a–c). Regarding the lagged responses of hydrological variables, CWD springs followed by non-CWD summers do not lead to water dryness, despite increased vegetation greenness (Fig. 3d, g). The magnitude of summer water deficit is similar for both cases that include CWD summer (Fig. 3e, f, h, i) and is consistent with summer TWS anomalies (Supplementary Fig. 8e, f). These results imply that in high latitudes, summer water reductions characterized by TWS, soil moisture, and runoff are not associated with increased spring greenness but are primarily caused by summer precipitation deficit.The productivity responses to compound climate conditions may be stronger than that to individual events through the synergistic effects of temperature and precipitation28. To investigate this, we compute the average anomalies in GPP and soil moisture associated with univariate events across the focus areas, which are then compared with the effects of CWD and CCD events in high latitudes (see “Methods”). Warm events can not only directly increase productivity but also show positive lagged effects (Supplementary Fig. 9a, b). In contrast, dry events reduce productivity (Supplementary Fig. 9e, f). This indicates that the direct and lagged positive effects of CWD events across high latitudes are mainly dominated by the warm component, while dry conditions have negative effects. Therefore, the warm-induced increase in productivity slightly exceeds that associated with CWD events (Supplementary Fig. 9b). Soil moisture under warm springs shows positive anomalies (Supplementary Fig. 9c, d), while they slightly decline during dry spring (Supplementary Fig. 9g, h). This suggests that the positive anomalies in soil water during CWD spring are driven by the warm component.CWD events reduce vegetation productivity in mid-latitudesHere, we first investigate the direct effects of springtime and summertime CWD events across mid-latitudes (23.5–50°N/S). Springtime productivity exhibits little changes during CWD spring (Fig. 4a, c), despite dry spring (Fig. 4d, f, g, i). When considering the direct effects of CWD events in summer, the results are similar, whereas the negative magnitude of productivity in summer is larger than that in spring (Fig. 4b, c). This difference suggests CWD conditions in summer show more adverse effects on productivity than that in spring in mid-latitudes. The anomalies in vegetation and TWS are consistent (Supplementary Fig. 10).Next, the lagged effects of springtime CWD events in mid-latitudes are assessed. In cases with CWD spring but non-CWD summer, summer productivity exhibits slight anomalies (Fig. 4a), with slightly decreased summer water (Fig. 4d, g). Summer productivity and water show much higher reductions in case with consecutive events (Fig. 4c, f, i) than for the case with only CWD summer (Fig. 4b, e, h). These results are supported by the responses of vegetation indices and TWS (Supplementary Fig. 10), revealing that springtime CWD events in mid-latitudes have negative lagged effects on summer productivity and water availability.The direct and lagged effects of individual events are finally compared with that of CWD events in mid-latitudes. Dry conditions in spring and summer directly decrease productivity and cause soil water dryness (Supplementary Fig. 11a–d). Moreover, dry spring depletes soil moisture earlier, which, in turn, causes dry summer and reduction in productivity during non-dry summer (Supplementary Fig. 11a, c). This indicates that dry springs have negative lagged effects on summer productivity. In contrast, productivity and soil water show positive anomalies during warm springs, while they show negative anomalies in summer (Supplementary Fig. 11e–h). These results suggest that the direct and lagged negative effects of CWD springs are dominated by the dry component in mid-latitudes, while the warm component mitigates the negative effects of the dry component in spring. Accordingly, the decline in productivity due to dry conditions thus exceeds that triggered by CWD events (Supplementary Fig. 11b).Decreased vegetation productivity due to the negative synergistic effects of CCD eventsHere, we first investigate the direct effects of CCD events across mid-to-high latitudes. Productivity reductions are found during springtime and summertime CCD events (Fig. 5a–c) concurrent with water reductions (Fig. 5). Vegetation and TWS show similar behaviors during CCD spring and summer (Supplementary Fig. 12). These results reveal that CCD events in spring and summer can impose direct adverse impacts on productivity and soil water across mid-to-high latitudes. The productivity reductions in spring and summer are similar in magnitude (Fig. 5a, b), indicating that CCD events between spring and summer can cause similar damage to productivity.We then analyze the lagged effects of springtime CCD events. Our results indicate that springtime CCD events show negative lagged effects on summer productivity and cause summer water reductions in mid-to-high latitudes (Fig. 5). Specifically, we find that in cases with CCD spring but non-CCD summer, summer productivity and water exhibit strongly negative anomalies (Fig. 5a, d, g). Moreover, summer anomalies are higher during consecutive events (Fig. 5c, f, i) than the cases including only CCD summer (Fig. 5b, e, h). Vegetation indices and TWS show similar responses (Supplementary Fig. 12). Our results further indicate that CCD spring has more severe negative lagged effects on productivity than CWD spring. That is, we find that in comparison to cases with preceding CWD spring and consecutive CWD events, summer productivity shows higher reduction in cases with preceding CCD spring and consecutive CCD events (Fig. 4a, c versus Fig. 5a, c). Moreover, in cases with CCD spring but non-CCD summer (Fig. 5a, d, g), summer anomalies are close to those in scenarios with non-CCD spring but CCD summer (Fig. 5b, e, h). The vegetation and TWS anomalies further confirm this situation (Supplementary Fig. 12a, b, d, e). These results suggest that the lagged effects of CCD spring can be of similar magnitude as their direct adverse effects.We finally compare the direct and lagged effects of individual events with that of CCD events in mid-to-high latitudes. Cold conditions in spring and summer directly reduce productivity but show weak effects on soil moisture (Supplementary Fig. 13a–d), and cold spring shows negative lagged effects on summer productivity (Supplementary Fig. 13a). Dry events show direct and lagged negative effects on productivity and soil moisture (Supplementary Fig. 13e–h). These results imply that the negative lagged effects of CCD springs are dominated by both cold and dry components. The effects of CCD events on productivity mostly exceeds the individual dry or cold impacts (Supplementary Fig. 13a, b, e, f). More