More stories

  • in

    Drowning in seaweed: How to stop invasive Sargassum

    Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
    the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
    Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
    and JavaScript. More

  • in

    As good as human experts in detecting plant roots in minirhizotron images but efficient and reproducible: the convolutional neural network “RootDetector”

    DatasetsImage acquisitionFor this study, we assembled three datasets: one for training of the RootDetector Convolutional Neural Network (Training-Set), one for a performance comparison between humans and RootDetector in segmenting roots in minirhizotron images (Comparison-Set), and one for the validation of the algorithm (Validation-Set). The Training-Set contained 129 images comprised of 17 randomly selected minirhizotron images sampled in a mesocosm experiment (see “Mesocosm sampling” Section), 47 randomly selected minirhizotron images sampled in a field study (see “Field sampling” Section) as well as the 65 minirhizotron images of soy roots published by Wang et al.15. The Comparison-Set contained 25 randomly selected minirhizotron images from the field-study which all were not part of the images included in the Training- and Validation-Sets. The Validation-Set contained 10 randomly selected minirhizotron images from the same field study, which had not been used in the Training-Set. All images were recorded with 2550 ✕ 2273 pixels at 300 dpi with a CI-600 In-Situ Root Imager (CID Bio-Science Inc., Camas, WA, USA) and stored as .tiff files to reduce compression loss. For all training and evaluation purposes we used raw, unprocessed output images from the CI-600.Mesocosm samplingThe mesocosm experiment was established in 2018 on the premises of the Institute for Botany and Landscape Ecology of the University of Greifswald (Fig. S1). It features 108 heavy duty plastic buckets of 100 l each, filled to two thirds of their height with moderately decomposed sedge fen peat. Each mesocosm contained one minirhizotron (inner diameter: 64 mm, outer diameter: 70 mm, length: 650 mm) installed at a 45°angle and capped in order to avoid penetration by light. The mesocosms were planted with varying compositions of plant species that typically occur in north-east German sedge fens (Carex rostrata, Carex acutiformis, Glyceria maxima, Equisetum fluviatile, Juncus inflexus, Mentha aquatica, Acorus calamus and Lycopus europaeus). The mesocosms were subjected to three different water table regimes: stable at soil surface level, stable at 20 cm below soil surface and fluctuating between the two levels every two weeks. The minirhizotrons were scanned weekly at two levels of soil depth (0–20 cm and 15–35 cm) between April 2019 and December 2021, resulting in roughly 9500 minirhizotron images of 216 × 196 mm. Manual quantification of root length would, based on own experience, take approximately three hours per image, resulting in approximately 28,500 h of manual processing for the complete dataset. Specimens planted were identified by author Dr. Blume-Werry, however no voucher specimen were deposited. All methods were carried out in accordance with relevant institutional, national, and international guidelines and legislation.Field samplingThe field study was established as part of the Wetscapes project in 201716. The study sites were located in Mecklenburg-Vorpommern, Germany, in three of the most common wetland types of the region: alder forest, percolation fen and coastal fen (Fig. S2). For each wetland type, a pair of drained versus rewetted study sites was established. A detailed description of the study sites and the experimental setup can be found in Jurasinski et al.16. At each site, 15 minirhizotrons (same diameter as above, length: 1500 mm) were installed at 45° angle along a central boardwalk. The minirhizotrons have been scanned biweekly since April 2018, then monthly since January 2019 at two to four levels of soil depth (0–20 cm, 20–40 cm, 40–60 cm and 60–80 cm), resulting in roughly 12,000 minirhizotron images of 216 × 196 cm, i.e. an estimated 36,000 h of manual processing for the complete dataset. Permission for the study was obtained from the all field owners. Figure 1Overview of the RootDetector system. The main component is a semantic segmentation network based on the U-Net architecture. The root length is estimated by skeletonizing the segmentation output and applying the formula introduced by Kimura et al.17. During training only, a weight map puts more emphasis on fine roots.Full size imageThe CNN RootDetectorImage annotationFor the generation of training data for the CNN, human analysts manually masked all root pixels in the 74 images of the Training-Set using GIMP 2.10.12. The resulting ground truth data are binary, black-and-white images in Portable Network Graphics (.png) format, where white pixels represent root structures and black pixels represent non-root objects and soil (Fig. 2b). All training data were checked and, if required, corrected by an expert (see “Selection of participants” for definition). The Validation-Set was created in the same way but exclusively by experts.Figure 2Example of segmentation and result of skeletonization. A 1000 by 1000 pixel input image (a), the manually annotated ground truth image (b), the RootDetector estimation image (c), the combined representation image (error map, d with green indicating true positives, red indicating false positive, blue indicating false negatives), the skeletonized RootDetector estimation image (e), and the skeletonized ground truth image (f).Full size imageArchitectureRootDetector’s core consists of a Deep Neural Network (DNN) based on the U-Net image segmentation architecture[27]nd is implemented in TensorFlow and Keras frameworks18. Although U-Net was originally developed for biomedical applications, it has since been successfully applied to other domains due to its generic design.RootDetector is built up of four down-sampling blocks, four up-sampling blocks and a final output block (Fig. 1). Every block contains two 3 × 3 convolutional layers, each followed by rectified linear units (ReLU). The last output layer instead utilizes Sigmoid activation. Starting from initial 64 feature channels, this number is doubled in every down-block and the resolution is halved via 2 × 2 max-pooling. Every up-block again doubles the resolution via bilinear interpolation and a 1 × 1 convolution which halves the number of channels. Importantly, after each up-sampling step, the feature map is concatenated with the corresponding feature map from the down-sampling path. This is crucial to preserve fine spatial details.Our modifications from the original architecture include BatchNormalization19 after each convolutional layer which significantly helps to speed up the training process and zero-padding instead of cropping as suggested by Ronneberger, Fischer, & Brox20 to preserve the original image size.In addition to the root segmentation network, we trained a second network to detect foreign objects, specifically the adhesive tape that is used as a light barrier on the aboveground part of the minirhizotrons. We used the same network architecture as above and trained in a supervised fashion with the binary cross-entropy loss. During inference, the result is thresholded (predefined threshold value: 0.5) and used without post-processing.TrainingWe pre-trained RootDetector on the COCO dataset21 to generate a starting point. Although the COCO dataset contains a wide variety of image types and classes not specifically related to minirhizotron images, Majurski et al.22 showed, that for small annotation counts, transfer-learning even from unrelated datasets may improve a CNNs performance by up to 20%. We fine-tuned for our dataset with the Adam optimizer23 for 15 epochs and trained on a total of 129 images from the Training-Set (17 mesocosm images, 47 field-experiment images, 65 soy root images). To enhance the dataset size and reduce over-fitting effects, we performed a series of augmentation operations as described by Shorten & Khoshgoftaar24. In many images, relatively coarse roots ( > 3 mm) occupied a major part of the positive (white) pixel space, which might have caused RootDetector to underestimate fine root details overall. Similarly, negative space (black pixels) between tightly packed, parallel roots was often very small and might have impacted the training process to a lesser extent when compared to large areas with few or no roots (Fig. 2). To mitigate both effects, we multiplied the result of the cross-entropy loss map with a weight map which emphasizes positive–negative transitions. This weight map is generated by applying the following formula to the annotated ground truth images:$$omega left( x right) = 1 – left( {tanh left( {2tilde{x} – 1} right)} right)^{2}$$
    (1)
    where ω(x) is the average pixel value of the annotated weight map in a 5 × 5 neighborhood around pixel x. Ronneberger, Fischer, & Brox20 implemented a similar weight map, however with stronger emphasis on space between objects. As this requires computation of distances between two comparatively large sets of points, we adapted and simplified their formula to be computable in a single 5 × 5 convolution.For the loss function we applied a combination of cross-entropy and Dice loss 25:$${mathcal{L}} = {mathcal{L}}_{CE} + lambda {mathcal{L}}_{Dice} = – frac{1}{N}sumnolimits_{i} {wleft( {x_{i} } right)y_{i} log left( {x_{i} } right) + lambda frac{{2sumnolimits_{i} {x_{i} y_{i} } }}{{sumnolimits_{i} {x_{i}^{2} sumnolimits_{i} {y_{i}^{2} } } }}}$$
    (2)

    where x are the predicted pixels, y the corresponding ground truth labels, N the number of pixels in an image and λ a balancing factor which we set to 0.01. This value was derived empirically. The Dice loss is applied per-image to counteract the usually high positive-to-negative pixel imbalance. Since this may produce overly confident outputs and restrict the application of weight maps, we used a relatively low value for λ.Output and post-processingRootDetector generates two types of output. The first type of output are greyscale .png files in which white pixels represent pixels associated with root structures and black pixels represent non-root structures and soil (Fig. 2c). The advantage of .png images is their lossless ad artifact-free compression at relatively small file sizes. RootDetector further skeletonizes the output images and reduces root-structures to single-pixel representations using the skeletonize function of scikit-image v. 0.17.1 (26; Fig. 2e,f). This helps to reduce the impact of large diameter roots or root-like structures such as rhizomes in subsequent analyses and is directly comparable to estimations of root length. The second type of output is a Comma-separated values (.csv) file, with numerical values indicating the number of identified root pixels, the number of root pixels after skeletonization, the number of orthogonal and diagonal connections between pixels after skeletonization and an estimation of the physical combined length of all roots for each processed image. The latter is a metric commonly used in root research as in many species, fine roots provide most vital functions such as nutrient and water transport3. Therefore, the combined length of all roots in a given space puts an emphasis on fine roots as they typically occupy a relatively smaller fraction of the area in a 2D image compared to often much thicker coarse roots. To derive physical length estimates from skeletonized images, RootDetector counts orthogonal- and diagonal connections between pixels of skeletonized images and employs the formula proposed by Kimura et al.17 (Eq. 3).$$L = left[ {N_{d}^{2} + left( {N_{d} + N_{o} /2} right)^{2} } right]^{{1/2}} + N_{o} /2$$
    (3)
    where Nd is the number of diagonally connected and No the number of orthogonally connected skeleton pixels. To compute Nd we convolve the skeletonized image with two 2 × 2 binary kernels, one for top-left-to-bottom-right connections and another for bottom-left-to-top-right connections and count the number of pixels with maximum response in the convolution result. Similarly, No is computed with a 1 × 2 and a 2 × 1 convolutional kernels.Performance comparisonSelection of participantsFor the performance comparison, we selected 10 human analysts and divided them into three groups of different expertise levels in plant physiology and with the usage of digital root measuring tools. The novice group consisted of 3 ecology students (2 bachelor’s, 1 master’s) who had taken or were taking courses in plant physiology but had no prior experience with minirhizotron images or digital root measuring tools. This group represents undergraduate students producing data for a Bachelor thesis or student assistants employed to process data. The advanced group consisted of 3 ecology students (1 bachelor’s, 2 master’s) who had already taken courses in plant physiology and had at least 100 h of experience with minirhizotron images and digital root measuring tools. The expert group consisted of 4 scientists (2 PhD, 2 PhD candidates) who had extensive experience in root science and at least 250 h of experience with digital root measuring tools. All methods were carried out in accordance with relevant institutional, national, and international guidelines and legislation and informed consent was obtained from all participants.Instruction and root tracingAll three groups were instructed by showing them a 60 min live demo of an expert tracing roots in minirhizotron images, during which commonly encountered challenges and pitfalls were thoroughly discussed. Additionally, all participants were provided with a previously generated, in-depth manual containing guidelines on the identification of root structures, the correct operation of the root tracing program and examples of often encountered challenges and suggested solutions. Before working on the Comparison-Set, all participants traced roots in one smaller-size sample image and received feedback from one expert.Image preparation and root tracingBecause the minirhizotron images acquired in the field covered a variety of different substrates, roots of different plant species, variance in image quality, and because tracing roots is very time consuming, we decided to maximize the number of images by tracing roots only in small sections, in order to cover the largest number of cases possible. To do this, we placed a box of 1000 × 1000 pixels (8.47 × 8.47 cm) at a random location in each of the images in the Comparison-Set and instructed participants to trace only roots within that box. Similarly, we provided RootDetector images where the parts of the image outside the rectangle were occluded. All groups used RootSnap! 1.3.2.25 (CID Bio-Science Inc., Camas, WA, USA;27), a vector based tool to manually trace roots in each of the 25 images in the comparison set. We decided on RootSnap! due to our previous good experience with the software and its’ relative ease of use. The combined length of all roots was then exported as a csv file for each person and image and compared to RootDetector’s output of the Kimura root length.ValidationWe tested the accuracy of RootDetector on a set of 10 image segments of 1000 by 1000 pixels cropped from random locations of the 10 images of the Validation-Set. These images were annotated by a human expert without knowledge of the estimations by the algorithm and were exempted from the training process. As commonly applied in binary classification, we use the F1 score as a metric to evaluate the performance RootDetector. F1 is calculated from precision (Eq. 4) and recall (Eq. 5) and represents their harmonic mean (Eq. 6). Ranging from 0 to 1, higher values indicate high classification (segmentation) performance. As one of the 10 image sections contained no roots and thus no F1 Score was calculable, it was excluded from the validation. We calculated the F1 score for each of the nine remaining image sections and averaged the values as a metric for overall segmentation performance.$$Precision;(P) = frac{{tp}}{{tp + fp}}$$
    (4)
    $$Recall;(R) = frac{{tp}}{{tp + fn}}$$
    (5)
    $$F1 = 2*frac{{P*R}}{{P + R}}$$
    (6)
    where P = precision, R = recall, tp = true positives; fp = false positives, fn = false negatives.Statistical analysisWe used R Version 4.1.2 (R Core Team, 2021) for all statistical analyses and R package ggplot2 Version 3.2.128 for visualizations. Pixel identification-performance comparisons were based on least-squares fit and the Pearson method. Root length estimation-performance comparisons between groups of human analysts (novice, advanced, expert) and RootDetector were based on the respective estimates of total root length plotted over the minirhizotron images in increasing order of total root length. Linear models were calculated using the lm function for each group of analysts. To determine significant differences between the groups and the algorithm, 95% CIs as well as 83% CIs were displayed and RootDetector root length outside the 95% CI were considered significantly different from the group estimate at α = 0.0529. The groups of human analysts were considered significantly different if their 83% CIs did not overlap, as the comparison of two 83% CIs approximates an alpha level of 5%30,31.This study is approved by Ethikkommission der Universitätsmedizin Greifswald, University of Greifswald, Germany. More

  • in

    Household energy-saving behavior, its consumption, and life satisfaction in 37 countries

    Figure 1 presents the average monthly energy expenditure at the household level based on USD across the 37 surveyed nations. The households in Singapore expend the most amount of energy, that is, 748 USD each month on average. The energy consumption appears positively associated with the economic development level; for example, households from high-income countries, including France, Italy, Japan and the US, tend to consume more energy than those from low-income countries (e.g., Kazakhstan, Myanmar, and Mongolia). In India, Indonesia, and Vietnam, households with higher income expend more on energy than rural/slum households. For the energy expenditure to household income ratio, strong trends were not found between developing and developed countries. Notably, middle-income countries (e.g., Greece, Chile, Brazil, Egypt) spend a relatively higher share of total income on energy.Figure 1Average monthly energy expenditure at the household level across the 37 surveyed nations. Data source: Original survey.Full size imageThe relationship between subjective well-being and energy consumption expenditure based on the ordered logit, ordered probit, and OLS models is shown in Table 2, panel A. The LR Chi-Square test and Pseudo R-squared for the ordered logistic regression model and the ordered probit model were applied to measure the goodness of the fit, whereas F-statistics and adjusted R-squared were used for the OLS model. For the validation of the measurement of subjective well-being, life satisfaction and happiness measures were used. Importantly, the results from variated regression models are consistent, indicating a positive relationship between household energy consumption expenditure and the improvement of individuals’ subjective well-being. Regarding the model’s goodness of fit, the LR Chi-Square test with ordered logit and probit models, and the F-statistic in the OLS model are all statistically significant at 0.1%, which validates the regression model. As the consistency of the robustness results is derived from different models, the ordered logit model is applied in Table 2 (Panel B).Table 2 Association between energy consumption expenditure and subjective well-being in high- and non-high-income countries.Full size tableWith the control variables being constant, energy consumption expenditure improves subjective well-being, including life satisfaction and happiness. The coefficients for the relationship of energy consumption with life satisfaction and with happiness are 0.018 and 0.008, respectively, and they are statistically significant at the 1% level; in other words, there is increased energy consumption for people who are satisfied with their lives and are happier. This is because electricity, water, gas, or gasoline are indispensable consumption goods in daily life. The results suggest that when policies lead to a reduction in the consumption of these goods at the household level, the life satisfaction of citizens is likely to decrease. When reducing energy consumption at the household level to reduce the emission of greenhouse gases, the conflicts of interest of individuals in these households (given that they derive life satisfaction from energy consumption) pose a challenge to policymakers; therefore, policymakers should devise strategies to improve both citizens’ living standards and environmental preservation.Referring to the criteria developed by the World Bank, the standard classification of high-income nations and non-high-income nations is as follows. Based on the 2017 gross national income (GNI) per capita, the World Bank List of Economies (June 2018) presented the following criteria for nations to be classified as high-income and non-high-income nations, respectively: a GNI per capita of $12,056 or higher, and less than $12,056. According to this standard of classification, in this study, high-income nations comprise Japan, Singapore, Chile, Australia, the United States, Germany, the United Kingdom, France, Spain, Italy, Sweden, Canada, Netherlands, Greece, Hungary, Poland, and the Czech Republic, whereas non-high-income nations comprise Thailand, Malaysia, Indonesia, Vietnam, Philippines, Mexico, Venezuela, Brazil, Colombia, South Africa, India, Myanmar, Kazakhstan, Mongolia, Egypt, Russia, China, Turkey, Romania, and Sri Lanka.Regarding the comparison of high- and non-high-income countries, energy consumption at the household level is more likely to lead to life satisfaction in non-high-income than in high-income countries. In high-income countries, the coefficients for the relationship of energy consumption with life satisfaction and with happiness are 0.010 and 0.003, respectively; these coefficients are 0.035 and 0.015, respectively, among non-high-income countries. Hence, in both high-income and non-high-income countries, an increase in energy consumption leads to an increase in life satisfaction; nonetheless, energy consumption is more crucial for households in non-high-income countries. Compared to the effect of energy consumption on satisfaction in high-income countries and non-high-income countries, individuals living in less urbanized countries appear more satisfied with energy consumption.Table 3 presents the association between life satisfaction and energy consumption expenditure at the household level in each country by estimating Eq. (2) based on the ordered logit model for each country. There is a positive relationship between energy consumption expenditure and life satisfaction in 27 out of the 37 nations. For example, the coefficient of this relationship is 0.062 in Brazil, and is statistically significant at the 1% level. An increase in energy consumption expenditure positively impacts the life satisfaction of households in Brazil, meaning that individuals with greater energy expenditure tend to be satisfied with their lives. Similar results are found in other countries: Canada, Chile, China, Egypt, France, Germany, Greece, India, Indonesia, Italy, and Japan. As life satisfaction is a proxy of well-being, energy consumption is expected to increase when households can afford more energy to obtain higher life satisfaction. These results indicate that most of the developed and developing countries analyzed face a conflict of interest in addressing individuals’ life satisfaction and environment conservation goals; these countries include China and India that are home to large populations that have a positive desire for energy consumption.Table 3 Relationship between energy expenditure and life satisfaction for each country.Full size tableHowever, the association between life satisfaction and energy consumption expenditure at the household level was non-significant across some countries. In Australia, the coefficient of this association is positive but not statistically significant; hence, an increase in energy expenditure is not completely associated with life satisfaction at the household level here. Similar results are found in the Netherlands, Hungary, Sweden, Singapore, Poland, the Czech Republic, and Colombia. In these countries, energy consumption is at an adequate level, and additional energy consumption does not lead to higher life satisfaction. It may be that households consume an adequate amount of energy with their income and energy price.Tables 4, 5, 6, and 7 display the determinant factors of household energy consumption in 37 nations by estimating the energy demand equation for each country using Eq. (3). The key energy consumption metric is the quantity of energy consumed (e.g., kWh) across the targeted households. Since price information is limited, transforming consumption expenditure into a quantity (e.g., kWh) is problematic. As explained earlier, this study adopted the energy demand equation.Table 4 Household socioeconomic and demographic determinants of household energy consumption expenditure I.Full size tableTable 5 Household socioeconomic and demographic determinants of household energy consumption expenditure II.Full size tableTable 6 Household socioeconomic and demographic determinants of household energy consumption expenditure III.Full size tableTable 7 Household socioeconomic and demographic determinants of household energy consumption expenditure IV.Full size tableThere are positive relationships between energy consumption expenditure at the household level and household income across countries. If the coefficients for household income are positive and statistically significant, this means that energy consumption expenditure at the household level would increase with an increase in household income ensuing from economic development in the country, ceteris paribus. The positive coefficients for the association between energy consumption expenditure and household income range from 0.756 (Japan) to 3.613 (the Philippines) in our sample, indicating that an additional 10,000 USD would lead to an additional energy consumption expenditure at the household level of approximately 17.3% (Japan) – 445% (Mongolia). The number is calculated using the magnitude of the coefficient/energy consumption expenditure. The results also show that homeowners tend to consume more energy than renters in Australia, Brazil, Canada, Chile, China, Colombia, Germany, India, Italy, Japan, Malaysia, Mexico, Russia, the United States, and Vietnam. This indicates that if individuals live in their own houses, the household energy consumption expenditure tends to be higher owing to the wealth effect, as energy is a normal consumption good. Overall, the wealth effect on energy consumption expenditure at the household level is increasing in our sample, and with economic development, energy consumption may increase.The following factors are confirmed to reduce energy consumption at the household level: (1) energy-curtailment behavior regarding electricity, (2) higher education, and (3) age. The energy-saving effect is confirmed in households. In Canada, the coefficient of energy-saving behaviors is -0.642, indicating that households consume 12.5% less energy when they adopt both energy curtailment behavior and non-saving groups (64.2/513). The Canadian household average energy consumption is 513 USD. Similar results are seen in Colombia, Germany, India, Indonesia, Italy, Japan, the Netherlands, Poland, Russia, Turkey, the United Kingdom, and the United States. The magnitude of the effect of energy curtailment behavior ranged from 6.4% (Russia) to 32% (India) less energy consumption expenditure. Hence, energy-saving behaviors have a favorable effect on environmentally preferable outcomes. By contrast, households in Indonesia save electricity as they tend to spend more on purchasing energy.Individuals with higher education tend to save energy in 23 out of the 37 nations. For instance, the coefficient for individuals with university-level education is -2.292 and statistically significant at the 1% level. This suggests that households with individuals who have university-level education have less energy consumption expenditure than households with individuals with junior high school or lower levels of education. Similar results are seen in Brazil, Canada, Chile, Colombia, the Czech Republic, France, Germany, Hungary, India, Indonesia, Japan, Malaysia, the Netherlands, the Philippines, Poland, Russia, Singapore, South Africa, Spain, Sweden, Turkey, the United Kingdom, and the United States. Encouraging households to engage in energy curtailment behaviors and higher educational attainment may lead to environment-friendly outcomes.Surprisingly, purchasing energy-saving household products has a limited effect on reducing energy consumption expenditure at the household level. The coefficients for purchasing energy-saving household products are negative, ranging between -0.044 and -0.763, and are statistically significant in Australia, Canada, the Czech Republic, Italy, and Kazakhstan. Hence, the purchase of these products in these five countries decreases energy expenditure from 2.9% (China) to 14% (Australia). However, the relationship between energy consumption expenditure at the household level and purchasing energy-saving household products is non-significant in the other countries. Moreover, in Poland and Turkey, households that purchase these products consume more energy than those that do not. Therefore, purchasing energy-saving household products has a limited contribution to energy saving at the household level.The findings also show that older individuals tend to have lower energy consumption. The coefficients for the age variable are negative and statistically significant in 30 countries (out of 37). The effect of age on energy consumption expenditure ranges between -0.003 and -0.148, indicating that as the average age of individuals increases by one year, their monthly energy consumption expenditure reduces from 0.3–14.8 USD. This may be because older individuals are more likely to live frugally. More

  • in

    Bimodality and alternative equilibria do not help explain long-term patterns in shallow lake chlorophyll-a

    Real-world dataThe dataset consisted of 2986 observations from 902 freshwater shallow lakes in Denmark and North America (data extracted from the LAGOSNE database on 22 February 2022 via R LAGOSNE package version 2.0.2)56 (Supplementary Fig. 9). The Danish lakes were sampled for one or several years from 1984 to 2020 (data extracted in October 2021 from https://odaforalle.au.dk/main.aspx) (Supplementary Fig. 10). Prerequisites for inclusion in the analysis were that lakes had been sampled for physical and chemical variables at least four times or at least three times over the growing season (May to September) for the Danish or North American lakes, respectively, had a mean depth of less than 3 m and were freshwater. Water chemistry samples were analysed using standard methods and data for total phosphorus (TP), total nitrogen (TN) and chlorophyll-a are included here57. The mean and range of TP, TN and chlorophyll-a for the combined sites is given in Table 1, along with the values for each region separately.To gain a longer-term perspective on the relationship between nutrients and chlorophyll-a, we calculated the across-year averages of the summer means of TP, TN and chlorophyll-a, sequentially increasing numbers of years included in the mean up to a total of a five-year mean, at which point there were only 99 lakes left in the dataset. In calculating the multi-year means we allowed a maximum gap of 2 years between observations (i.e. two observations could cover 3 years) to avoid including time series with too many missing years in between. Hence, only lakes with sufficient numbers of sequential data were included, resulting in a large drop in lake numbers as the length of the multi-year mean increased (Table 2).Numerical methodsDiagnostic tests or proxies of alternative equilibriaWe modelled the response of chlorophyll-a to TP and TN using generalised-linear models58 with Gamma distribution and an identity link on untransformed data for single-year and multiple-year means up to 5-year means. We used the Gamma distribution, as chlorophyll fit this significantly better than a normal or log-normal distribution. We used psuedo R2 of the model along with the patterns of residuals, and finally, we plotted the kernel density of the chlorophyll-a values as diagnostics of the presence, absence or prevalence of alternative equilibria in the simulated and real work data.To test how appropriate these diagnostics or proxies of alternative stable states in terms of how well they identify the existence of alternative stable states in randomly sampled multi-year data, we

    1.

    Simulated two scenarios for the main manuscript, with and without alternative stable states in the data, which were as close to the real-life data as possible. The results of these scenarios appear in the main text (please see details below in the “Data Simulation” section).

    2.

    We provide multiple scenarios with different degrees, or prevalence, of alternative stable states in the data, see simulations of alternative stable state scenarios. The results of these scenarios appear in Supplementary note 2.

    Hierarchical bootstrap approachThere are a large number of permutations of data, both real-world and simulated, that can provide a mean of the two to five sequential years from each lake in the time series data. It was vital to have a method that selects the data for analysis that provides a valid and comparable representation of both real work and simulated data and the models’ errors. In order to provide this we used a non-parametric hierarchical bootstrap procedure38. The flowchart shows the data preparation and data analysis steps of the hierarchical bootstrap procedure (Fig. 4). In the first step (step 1 in Fig. 4), all possible longer-term means are calculated for each lake. To keep as much data as possible, we decided to allow for up to 2 years of gap in the data between years. Taking the 5-year mean data as an example, if data from a lake existed for the years 1991 and 1994−1997, a 5-year mean would be calculated for the years 1991, 1994, 1995, 1996 and 1997. However, if the time series would contain a larger gap, e.g. data would only exist for the years 1991 and 1995–1998, no 5-year mean could be calculated. After the selection procedure, all the 2-year, 3-year and 5-year means are transferred into a new table (step 2 in Fig. 4).Fig. 4Data preparation and analysis steps of the hierarchical bootstrap procedure.Full size imageThe procedure is the same for each temporal scale from 2-year means to 5-year means. For the example of 5 mean years, lakes are randomly sampled from the full 5-year mean dataset in step 2 (Fig. 4) with replacement up to the number of lakes as in the original dataset, for the 5-year means 99 (step 3a). Here, the same lake can appear multiple times or not at all. This step is common for every bootstrap procedure59. However, since we have nested data (5-year means within lakes), we need a second step, in which for every resampled lake in step 3a, one 5-year mean is chosen (step 3b in Fig. 4). Then the three GLM models are produced from the randomly selected data in step 3c (Fig. 4). These steps are then repeated 1000 times to get a good representation of the uncertainties of the model. To ensure a fair comparison between single-year data and their equivalent multi-year mean data, we repeated the bootstrap procedure with single years only using only the lakes for which we also calculated multi-year means. To take the five-year mean as an example, there were 99 lakes where we could calculate at least one 5-year mean observation. First, we ran the bootstrap procedure to calculate 5-year mean values of TP, TN and chlorophyll-a (1000 times) and then took single years’ values of TP, TN and chlorophyll-a (1000 times) from exactly the same 99 lakes. With this approach, exactly the same datasets with the same lakes and observations within lakes are used for the calculation of the multi-year means and their single-year counterparts, making for a robust analysis. The GLM models did not always converge. If either the TP, TN or TP*TN model with interaction did not converge, the iteration was not used in further analysis. The number of converging models equal for each iteration of random samples is given in the results.The described hierarchical approach is the best way to reflect the structure of the original data. A simple, non-hierarchical bootstrap would favour lakes with more five-year means over lakes with fewer five-year means, simply because these make up a larger part of the data. Furthermore, sampling without replacement at the lake level would result in five-year means from lakes with few data dominating the produced random dataset, as every lake would be sampled every time, which then would result in high model leverage of 5-year means from lakes with less data. In contrast, the hierarchical procedure ensures that every lake has the same chance to end up in the randomly sampled bootstrap, in the second step, it ensures that of each sampled lake, every 5-year mean has the same chance to end up in the random dataset. These notions are in agreement with the findings of an assessment on how to properly resample hierarchical data by non-parametric bootstrap38.Data simulationGeneral approach of simulation assumptions and proceduresWe generated random scatter for the generalised-linear model based on Gamma distributions for two different “populations” of lakes with two different intercepts and slopes. At first, we calculated the linear equations for the two populations:For each population i and j, 99 samples (equalling the number of lakes in real-life data with 5-year means, nlake) were generated with a specific number of data points depending on the scenario (nyear) each, hence nlake = 1−99 for each population of lakes, e.g. with 20 years (nyear = 20) each.We found the real nutrient data to be normally distributed, with total nitrogen (TN) having a range between 0.33 and 4.93 mg/L and a constant coefficient of variation (CV, with a mean CV of 0.35) across this range (the same is true for total phosphorus (TP) at a shorter range). Hence, for each nlake, the x for the nyear = 20 were generated based on the mean range (mean per lake of the real-life data) and CV (0.35) from the real-life TN concentration data, hence with a range of 0.33 to 4.33 mg/L. Therefore the values and random variability of x in the simulations are close to the true values of the TN concentrations. The x is then fed into the linear equations above.To the resulting yi and yj we added random noise based on the Gamma distribution (using the rgamma function in R). We used a Gamma distribution because the Chlorophyll-a concentration also follows a Gamma distribution. The variability of a Gamma distribution is expressed by the shape variable. The variability of chlorophyll-a, its shape value, equals 2.63. This shape value was used in the Gamma distribution of yi and yj. The final calculated yi and yj had therefore a random rate calculated as shape/yi or shape/yj. Hence, their variability in the y dimension was close to the true chlorophyll-a variability.The data from both lake populations were then pooled and randomly sampled using the same hierarchical bootstrap procedure with 500 iterations for the scenarios in the supplementary materials and with 1000 iterations for main text simulation scenarios, which is identical to what was done for the real-world data.Simulation scenarios based on characteristics of real-world dataThe real-world 5-year mean data consisted of 99 lakes with 5–20 years of data for each lake. For the simulation scenario in the main text, we therefore randomly sampled between 5 and 20 data points for each of the 99 simulated lakes based on the x distribution described above. Intercepts and slopes of the simulation, resembled the range of the true data (see scatter plots in Fig. 2 of the main manuscript).In the alternative stable state scenario, we chose two slopes and intercepts for different populations of lakes:

    Population i: ai = 0, bi = 40

    Population j: aj = 50, bj = 120

    We based the slopes and intercepts of the ASS scenarios on the diagnostic combination defined by Scheffer and Carpenter7 which propose an abrupt shift in (a) the time series, (b) the multimodal distribution of states and (c) the dual relationship to a controlling factor. Here, the idea is that an ecosystem will jump from one state to the next at the same (nutrient) conditions (different intercept and/or slope, condition a within ref. 7), where any change in the nutrient will have different effects on algae or macrophytes (best represented by different slope, condition c), resulting in a multimodal distribution of the response (condition b). Hence, simulations are in line with what is predicted for ASS, but we took great lengths to also show other possibilities with the simulations in the Supplementary information to ensure we did not overlook any occasional occurrence of alternative equilibria.Here, the appearance of alternative stable states in the data could happen at any point in the time series of a single lake, or the entire time series could include only one of the two alternative stable states. To accommodate these alternative stable state constellations (for each of which we made a separate simulation scenario, (see Supplementary Note 2, “Simulations of alternative stable state constellations”), we forced the alternative stable state scenario to be constructed of 1/3 of data with one state, 1/3 of data with the second state and 1/3 of data where both alternative states could occur. In the latter case, the alternative stable state appeared after the first 20% but before the last 20% of the time series. Since the variability and range of x (nutrient) and y (chlorophyll -a response) is simulated as close as possible to the real-world data in all scenarios, the measures taken here (variable time series and combination of different alternative stable state scenario constellations) produce a simulation as close to the real-world data as possible. Specifically, we found the real-world nutrient data to be normally distributed, with total nitrogen (TN) having a range between 0.33 and 4.93 mg/L and a constant coefficient of variation (CV, with a mean CV of 0.35) across this range (the same is true for total phosphorus (TP) at a shorter range). Hence, for each simulated lake, the x were generated based on this mean range and CV. Furthermore, the resulting yi and yj were randomised by using a Gamma distribution (using the rgamma function in R). We used a Gamma distribution because the chlorophyll-a concentration also follows a Gamma distribution. The variability of a Gamma distribution is expressed by the shape variable. The variability of chlorophyll-a, its shape value, equals 2.63. This shape value was used in the Gamma distribution of yi and y. The final calculated yi and yj had, therefore a random rate calculated as shape/yi or shape/y. Hence, their variability in the y dimension was close to the true chlorophyll-a variability.For the scenario without alternative stable states, both populations of data had the same intercept and slope:

    Population i: ai = 0, bi = 40

    Population j: aj = 0, bj = 40.

    Please see Supplementary Note 2 for further simulations of different potential constellations of alternative states. There we show that our approach finds alternative stable states in response to nutrient concentration, even if they appear in time series from different lakes.Assessment of diagnostic tests or proxies of alternative equilibriaWe modelled the response of chlorophyll-a to TP and TN using generalised-linear models3 with Gamma distribution and an identity link on untransformed data for single-year and multiple-year means up to 5-year means. We used the Gamma distribution, as chlorophyll fit this significantly better than a normal or log-normal distribution. We used R2 of the model along with the patterns of residuals, and finally, we plotted the kernel density of the chlorophyll-a values as diagnostics of the presence, absence or prevalence of alternative equilibria in the simulated and real work data.The comparison of how the diagnostics/proxies of alternative stable states respond to the variation in the prevalence of alternative equilibria in the simulated datasets provides a robust assessment of their ability to identify both the presence and absence of alternative equilibria. It is the response of these diagnostic tests over time, with the increase in the temporal perspective as more years are added to the mean values of TP, TN and chlorophyll-a, that are key to the identification of the presence and or absence of alternative equilibria in a given dataset. The simulations show that a dataset which contains alternative equilibria will show (1) no improvement in R2 as the temporal perspective of the data increases (more years in the multi-year mean); (2) an increased bimodality in the residuals of the models of nutrients predicting chlorophyll-a will increase as more years are added to the multi-year mean and (3) the kernel density function of chlorophyll-a will display increasingly bimodality as more years are added to the mean. In the absence of alternative equilibria, the patterns differ with an R2, and increase in unimodality of residuals and a consistent unimodal pattern in the kernel density function. Thus, the diagnostic tests provide a robust test of both the presence and absence of alternative equilibria in a given dataset.Alternative stable state assessment for real data with limited data rangeIt could be the case that alternative stable states do not appear in the full dataset but only in a limited TN and TP concentration range. We filtered and re-analyzed the data, only keeping data points within the following two ranges: – TN concentration = 0.5−2 mg/L–TP concentration = 0.05−0.4 mg/L. In the filtered data, 1329 out of the original 2876 single-year data points, 289 out of 1028 3-year mean data points and 212 out of the 864 five-mean year data points remained. The filtered data consisted of data points from 550, 48 and 27 lakes for the single-year data, 3-year means and 5-year means, respectively. The smaller range resulted in lower R² of the models, yet the pattern that multi-year means result in higher R² compared to single-year data was largely consistent, apart from the 5-year mean TN models for which both, the single-year and mean data resulted in very low R² (Supplementary Fig. 6). Furthermore, due to the lower number of samples, the errors of all proxies are higher, making conclusions more difficult than for the full data. Still, we do not see any clear indication of alternative stable states in the scatter plots (two groups of dots are not appearing (Supplementary Fig. 5), the Kernel density plots (or model residuals (Supplementary Fig. 6)). i.e. no signs of bimodality in residuals or Kernel density plots. Please see details on this analysis in the supplementary material.Details and the R code for the steps for the random multi-year sampling can be found in the supplementary materials.Reporting summaryFurther information on research design is available in the Nature Portfolio Reporting Summary linked to this article. More

  • in

    Marine protected areas, marine heatwaves, and the resilience of nearshore fish communities

    Lauchlan, S. S. & Nagelkerken, I. Species range shifts along multistressor mosaics in estuarine environments under future climate. Fish Fish. 21, 32–46 (2020).Article 

    Google Scholar 
    Gao, G., Zhao, X., Jiang, M. & Gao, L. Impacts of marine heatwaves on algal structure and carbon sequestration in conjunction with ocean warming and acidification. Front. Mar. Sci. 8, 758651 (2021).Article 

    Google Scholar 
    Asch, R. G. Climate change and decadal shifts in the phenology of larval fishes in the California Current ecosystem. Proc. Natl. Acad. Sci. 112, E4065–E4074 (2015).Article 
    ADS 
    CAS 

    Google Scholar 
    Lonhart, S. I., Jeppesen, R., Beas-Luna, R., Crooks, J. A. & Lorda, J. Shifts in the distribution and abundance of coastal marine species along the eastern Pacific Ocean during marine heatwaves from 2013 to 2018. Mar. Biodivers. Rec. 12, 13 (2019).Article 

    Google Scholar 
    Morley, J. W. et al. Projecting shifts in thermal habitat for 686 species on the North American continental shelf. PLoS ONE 13, e0196127 (2018).Article 

    Google Scholar 
    Vergés, A. et al. The tropicalization of temperate marine ecosystems: Climate-mediated changes in herbivory and community phase shifts. Proc. R. Soc. B 281, 20140846 (2014).Article 

    Google Scholar 
    Wernberg, T. et al. Climate-driven regime shift of a temperate marine ecosystem. Science 353, 169–172 (2016).Article 
    ADS 
    CAS 

    Google Scholar 
    Cheung, W. W. L. et al. Marine high temperature extremes amplify the impacts of climate change on fish and fisheries. Sci. Adv. https://doi.org/10.1126/sciadv.abh0895 (2021).Article 

    Google Scholar 
    Ling, S. D., Johnson, C. R., Frusher, S. D. & Ridgway, K. R. Overfishing reduces resilience of kelp beds to climate-driven catastrophic phase shift. Proc. Natl. Acad. Sci. 106, 22341–22345 (2009).Article 
    ADS 
    CAS 

    Google Scholar 
    Pessarrodona, A. et al. Tropicalization unlocks novel trophic pathways and enhances secondary productivity in temperate reefs. Funct. Ecol. 36, 659–673 (2022).Article 

    Google Scholar 
    Hobday, A. J. et al. A hierarchical approach to defining marine heatwaves. Prog. Oceanogr. 141, 227–238 (2016).Article 
    ADS 

    Google Scholar 
    Holbrook, N. J. et al. A global assessment of marine heatwaves and their drivers. Nat. Commun. 10, 2624 (2019).Article 
    ADS 

    Google Scholar 
    Smale, D. A. et al. Marine heatwaves threaten global biodiversity and the provision of ecosystem services. Nat. Clim. Chang. 9, 306–312 (2019).Article 
    ADS 

    Google Scholar 
    Cheung, W. W. L. & Frölicher, T. L. Marine heatwaves exacerbate climate change impacts for fisheries in the northeast Pacific. Sci. Rep. 10, 6678 (2020).Article 
    ADS 
    CAS 

    Google Scholar 
    Garrabou, J. et al. Marine heatwaves drive recurrent mass mortalities in the Mediterranean Sea. Glob. Change Biol. 28, 5708–5725 (2022).Article 
    CAS 

    Google Scholar 
    Wernberg, T. et al. An extreme climatic event alters marine ecosystem structure in a global biodiversity hotspot. Nat. Clim. Change 3, 78–82 (2013).Article 
    ADS 

    Google Scholar 
    Cure, K. et al. Distributional responses to marine heat waves: insights from length frequencies across the geographic range of the endemic reef fish Choerodon rubescens. Mar. Biol. 165, 1 (2018).Article 

    Google Scholar 
    Jacox, M. G., Tommasi, D., Alexander, M. A., Hervieux, G. & Stock, C. A. Predicting the evolution of the 2014–2016 California current system marine heatwave from an ensemble of coupled global climate forecasts. Front. Mar. Sci. https://doi.org/10.3389/fmars.2019.00497 (2019).Article 

    Google Scholar 
    Gentemann, C. L., Fewings, M. R. & García-Reyes, M. Satellite sea surface temperatures along the West Coast of the United States during the 2014–2016 northeast Pacific marine heat wave. Geophys. Res. Lett. 44, 312–319 (2017).Article 
    ADS 

    Google Scholar 
    Cavanaugh, K. C., Reed, D. C., Bell, T. W., Castorani, M. C. N. & Beas-Luna, R. Spatial variability in the resistance and resilience of giant kelp in southern and baja California to a multiyear heatwave. Front. Mar. Sci. https://doi.org/10.3389/fmars.2019.00413 (2019).Article 

    Google Scholar 
    Cavole, L. M. et al. Biological impacts of the 2013–2015 warm-water anomaly in the Northeast Pacific: Winners, losers, and the future. Oceanography 29, 273–285 (2016).Article 

    Google Scholar 
    Sen Gupta, A. et al. Drivers and impacts of the most extreme marine heatwave events. Sci. Rep. 10, 19359 (2020).Article 
    ADS 
    CAS 

    Google Scholar 
    Rykaczewski, R. R. & Checkley, D. M. Influence of ocean winds on the pelagic ecosystem in upwelling regions. Proc. Natl. Acad. Sci. 105, 1965–1970 (2008).Article 
    ADS 
    CAS 

    Google Scholar 
    Thompson, A. R. et al. Putting the Pacific marine heatwave into perspective: The response of larval fish off southern California to unprecedented warming in 2014–2016 relative to the previous 65 years. Glob. Change Biol. 28, 1766–1785 (2022).Article 
    CAS 

    Google Scholar 
    Suryan, R. M. et al. Ecosystem response persists after a prolonged marine heatwave. Sci. Rep. 11, 6235 (2021).Article 
    ADS 
    CAS 

    Google Scholar 
    Bates, A. E. et al. Resilience and signatures of tropicalization in protected reef fish communities. Nat. Clim. Change 4, 62–67 (2014).Article 
    ADS 

    Google Scholar 
    Behrens, M. & Lafferty, K. Effects of marine reserves and urchin disease on southern Californian rocky reef communities. Mar. Ecol. Prog. Ser. 279, 129–139 (2004).Article 
    ADS 

    Google Scholar 
    Bernhardt, J. R. & Leslie, H. M. Resilience to climate change in coastal marine ecosystems. Ann. Rev. Mar. Sci. 5, 371–392 (2013).Article 

    Google Scholar 
    Caselle, J. E., Davis, K. & Marks, L. M. Marine management affects the invasion success of a non-native species in a temperate reef system in California, USA. Ecol. Lett. 21, 43–53 (2018).Article 

    Google Scholar 
    Micheli, F. et al. Evidence that marine reserves enhance resilience to climatic impacts. PLoS ONE 7, e40832 (2012).Article 
    ADS 
    CAS 

    Google Scholar 
    Olds, A. D. et al. Marine reserves help coastal ecosystems cope with extreme weather. Glob. Change Biol. 20, 3050–3058 (2014).Article 
    ADS 

    Google Scholar 
    Freedman, R. M., Brown, J. A., Caldow, C. & Caselle, J. E. Marine protected areas do not prevent marine heatwave-induced fish community structure changes in a temperate transition zone. Sci. Rep. 10, 21081 (2020).Article 
    ADS 
    CAS 

    Google Scholar 
    Bates, A. E. et al. Climate resilience in marine protected areas and the ‘Protection Paradox’. Biol. Cons. 236, 305–314 (2019).Article 

    Google Scholar 
    Kirlin, J. et al. California’s Marine Life Protection Act Initiative: Supporting implementation of legislation establishing a statewide network of marine protected areas. Ocean Coast. Manag. 74, 3–13 (2013).Article 

    Google Scholar 
    Saarman, E. T. et al. An ecological framework for informing permitting decisions on scientific activities in protected areas. PLoS ONE 13, e0199126 (2018).Article 

    Google Scholar 
    Caselle, J. E., Rassweiler, A., Hamilton, S. L. & Warner, R. R. Recovery trajectories of kelp forest animals are rapid yet spatially variable across a network of temperate marine protected areas. Sci. Rep. 5, 14102 (2015).Article 
    ADS 
    CAS 

    Google Scholar 
    Hamilton, S. L., Caselle, J. E., Malone, D. P. & Carr, M. H. Incorporating biogeography into evaluations of the Channel Islands marine reserve network. PNAS 107, 18272–18277 (2010).Article 
    ADS 
    CAS 

    Google Scholar 
    Wendt, D. E. & Starr, R. M. Collaborative research: An effective way to collect data for stock assessments and evaluate marine protected areas in California. Mar. Coast. Fish. 1, 315–324 (2009).Article 

    Google Scholar 
    Côté, I. M. & Darling, E. S. Rethinking ecosystem resilience in the face of climate change. PLoS Biol. 8, e1000438 (2010).Article 

    Google Scholar 
    Holling, C. S. Resilience and stability of ecological systems. Annu. Rev. Ecol. Syst. 4, 1–23 (1973).Article 

    Google Scholar 
    Li, L. et al. Subregional differences in groundfish distributional responses to anomalous ocean bottom temperatures in the northeast Pacific. Glob. Change Biol. 25, 2560–2575 (2019).Article 
    ADS 

    Google Scholar 
    Dawson, M. N. Phylogeography in coastal marine animals: A solution from California?. J. Biogeogr. 28, 723–736 (2001).Article 

    Google Scholar 
    Horn, M. H., Allen, L. G. & Lea, R. N. Biogeography. In The Ecology of Marine Fishes: California and Adjacent Waters (ed. Allen, L.) 3–25 (University of California Press, 2006). https://doi.org/10.1525/california/9780520246539.003.0001.Chapter 

    Google Scholar 
    Horn, M. H. & Allen, L. G. A distributional analysis of California coastal marine fishes. J. Biogeogr. 5, 23–42 (1978).Article 

    Google Scholar 
    Garrabou, J. et al. Mass mortality in Northwestern Mediterranean rocky benthic communities: Effects of the 2003 heat wave. Glob. Change Biol. 15, 1090–1103 (2009).Article 
    ADS 

    Google Scholar 
    Smale, D. A. & Wernberg, T. Extreme climatic event drives range contraction of a habitat-forming species. Proc. R. Soc. B 280, 20122829 (2013).Article 

    Google Scholar 
    O’Leary, B. C. et al. Addressing criticisms of large-scale marine protected areas. Bioscience 68, 359–370 (2018).Article 

    Google Scholar 
    California Department of Fish and Wildlife. California Sheephead, Bodianus (formerly Semicossyphus) pulcher, Enhanced Status Report. (2021).Pinsky, M. L., Selden, R. L. & Kitchel, Z. J. Climate-driven shifts in marine species ranges: Scaling from organisms to communities. Ann. Rev. Mar. Sci. 12, 153–179 (2020).Article 

    Google Scholar 
    Francour, P., Mangialajo, L. & Pastor, J. Mediterranean marine protected areas and non-indigenous fish spreading. In Fish Invasions of the Mediterranean Sea: Change and Renewal (eds Golani, D. & Appelbaum-Golani, B.) 127–144 (Pensoft Publisher, 2010).
    Google Scholar 
    Couce, E., Ridgwell, A. & Hendy, E. J. Future habitat suitability for coral reef ecosystems under global warming and ocean acidification. Glob. Change Biol. 19, 3592–3606 (2013).Article 
    ADS 

    Google Scholar 
    Bennett, S., Wernberg, T., Harvey, E. S., Santana-Garcon, J. & Saunders, B. J. Tropical herbivores provide resilience to a climate-mediated phase shift on temperate reefs. Ecol. Lett. 18, 714–723 (2015).Article 

    Google Scholar 
    Trainer, V. L. et al. Pelagic harmful algal blooms and climate change: Lessons from nature’s experiments with extremes. Harmful Algae 91, 101591 (2020).Article 

    Google Scholar 
    Gliwicz, Z. M., Babkiewicz, E., Kumar, R., Kunjiappan, S. & Leniowski, K. Warming increases the number of apparent prey in reaction field volume of zooplanktivorous fish. Limnol. Oceanogr. 63, S30–S43 (2018).Article 
    ADS 

    Google Scholar 
    Nielsen, J. M. et al. Responses of ichthyoplankton assemblages to the recent marine heatwave and previous climate fluctuations in several Northeast Pacific marine ecosystems. Glob. Change Biol. 27, 506–520 (2021).Article 
    ADS 

    Google Scholar 
    du Pontavice, H., Gascuel, D., Reygondeau, G., Stock, C. & Cheung, W. W. L. Climate-induced decrease in biomass flow in marine food webs may severely affect predators and ecosystem production. Glob. Change Biol. 27, 2608–2622 (2021).Article 
    ADS 

    Google Scholar 
    Arimitsu, M. L. et al. Heatwave-induced synchrony within forage fish portfolio disrupts energy flow to top pelagic predators. Glob. Change Biol. 27, 1859–1878 (2021).Article 
    ADS 
    CAS 

    Google Scholar 
    Oken, K. L., Essington, T. E. & Fu, C. Variability and stability in predation landscapes: A cross-ecosystem comparison on the potential for predator control in temperate marine ecosystems. Fish Fish. 19, 489–501 (2018).Article 

    Google Scholar 
    Baum, J. K. & Worm, B. Cascading top-down effects of changing oceanic predator abundances. J. Anim. Ecol. 78, 699–714 (2009).Article 

    Google Scholar 
    Jacox, M. G. et al. Impacts of the 2015–2016 El Niño on the California current system: Early assessment and comparison to past events. Geophys. Res. Lett. 43, 7072–7080 (2016).Article 
    ADS 

    Google Scholar 
    Brodeur, R. D., Auth, T. D. & Phillips, A. J. Major shifts in pelagic micronekton and macrozooplankton community structure in an upwelling ecosystem related to an unprecedented marine heatwave. Front. Mar. Sci. https://doi.org/10.3389/fmars.2019.00212 (2019).Article 

    Google Scholar 
    Field, J. C. et al. Spatiotemporal patterns of variability in the abundance and distribution of winter-spawned pelagic juvenile rockfish in the California Current. PLoS ONE 16, e0251638 (2021).Article 
    CAS 

    Google Scholar 
    Schroeder, I. D. et al. Source water variability as a driver of rockfish recruitment in the California current ecosystem: Implications for climate change and fisheries management. Can. J. Fish. Aquat. Sci. 76, 950–960 (2019).Article 
    CAS 

    Google Scholar 
    Echeverria, T. W. Thirty-four species of California rockfishes: Maturity and seasonality of reproduction. Fish. Bull. 85, 229–250 (1987).
    Google Scholar 
    Miller, A. & Sydeman, W. Rockfish response to low-frequency ocean climate change as revealed by the diet of a marine bird over multiple time scales. Mar. Ecol. Prog. Ser. 281, 207–216 (2004).Article 
    ADS 

    Google Scholar 
    Johnson, K. F. et al. Status of lingcod (Ophiodon elongatus) along the southern U.S. west coast in 2021. 195 p. (2021).Winemiller, K. O. & Rose, K. A. Patterns of life-history diversification in North American fishes: Implications for population regulation. Can. J. Fish. Aquat. Sci. 49, 2196–2218 (1992).Article 

    Google Scholar 
    Stuart-Smith, R. D., Brown, C. J., Ceccarelli, D. M. & Edgar, G. J. Ecosystem restructuring along the great barrier reef following mass coral bleaching. Nature 560, 92–96 (2018).Article 
    ADS 
    CAS 

    Google Scholar 
    Starr, R. M. et al. Variation in responses of fishes across multiple reserves within a network of marine protected areas in temperate waters. PLoS ONE 10, e0118502 (2015).Article 

    Google Scholar 
    Ziegler, S. L. et al. External fishing effort regulates positive effects of no-take marine protected areas. Biol. Cons. 269, 109546 (2022).Article 

    Google Scholar 
    Jarvis, E. T. & Lowe, C. G. The effects of barotrauma on the catch-and-release survival of southern California nearshore and shelf rockfish (Scorpaenidae, Sebastes spp.). Can. J. Fish. Aquat. Sci. 65, 1286–1296 (2008).Article 

    Google Scholar 
    Brooks, R. et al. Nearshore Fishes Abundance and Distribution Data, California Collaborative Fisheries Research Program (CCFRP). (2022).García-Reyes, M. & Sydeman, W. J. California multivariate ocean climate indicator (MOCI) and marine ecosystem dynamics. Ecol. Ind. 72, 521–529 (2017).Article 

    Google Scholar 
    R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing. (2021).Oksanen, J. et al. vegan: Community Ecology Package. (2020).Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D. & R Core Team. nlme: Linear and Nonlinear Mixed Effects Models. (2021). More

  • in

    City comfort: weaker metabolic response to changes in ambient temperature in urban red squirrels

    Speakman, J. R. The cost of living: Field metabolic rates of small mammals. Adv. Ecol. Res. 30, 177–297 (1999).Article 

    Google Scholar 
    Brown, J. H., Gillooly, J. F., Allen, A. P., Savage, V. M. & West, G. B. Toward a metaboolic theory of ecology. Ecology 85(7), 1771–1789. https://doi.org/10.1890/03-9000 (2004).Article 

    Google Scholar 
    Larivée, M. L., Boutin, S., Speakman, J. R., McAdam, A. G. & Humphries, M. M. Associations between over-winter survival and resting metabolic rate in juvenile North American red squirrels. Funct. Ecol. 24(3), 597–607. https://doi.org/10.1111/j.1365-2435.2009.01680.x (2010).Article 

    Google Scholar 
    Corp, N., Gorman, M. L. & Speakman, J. R. Seasonal variation in the resting metabolic rate of male wood mice Apodemus sylvaticus from two contrasting habitats 15 km apart. J. Comp. Physiol. B 167(3), 229–239. https://doi.org/10.1007/s003600050069 (1997).Article 
    CAS 

    Google Scholar 
    Lehto Hürlimann, M., Martin, J. G. A. & Bize, P. Evidence of phenotypic correlation between exploration activity and resting metabolic rate among populations across an elevation gradient in a small rodent species. Behav. Ecol. Sociobiol. 73(9), 131. https://doi.org/10.1007/s00265-019-2740-6 (2019).Article 

    Google Scholar 
    Reher, S., Rabarison, H., Montero, B. K., Turner, J. M. & Dausmann, K. H. Disparate roost sites drive intraspecific physiological variation in a Malagasy bat. Oecologia 198(1), 35–52. https://doi.org/10.1007/s00442-021-05088-2 (2022).Article 
    ADS 

    Google Scholar 
    McDonald, R. I. et al. Research gaps in knowledge of the impact of urban growth on biodiversity. Nat. Sustain. https://doi.org/10.1038/s41893-019-0436-6 (2019).Article 

    Google Scholar 
    Shochat, E., Warren, P. S., Faeth, S. H., McIntyre, N. E. & Hope, D. From patterns to emerging processes in mechanistic urban ecology. Trends Ecol. Evol. 21(4), 186–191. https://doi.org/10.1016/j.tree.2005.11.019 (2006).Article 

    Google Scholar 
    United Nations, Department of Economic and Social Affairs, Population Division. World Urbanization Prospects 2018: Highlights. https://population.un.org/wup/Publications/ (2018).Alberti, M. et al. The complexity of urban eco-evolutionary dynamics. Bioscience 70(9), 772–793. https://doi.org/10.1093/biosci/biaa079 (2020).Article 

    Google Scholar 
    Birnie-Gauvin, K., Peiman, K. S., Gallagher, A. J., de Bruijn, R. & Cooke, S. J. Sublethal consequences of urban life for wild vertebrates. Environ. Rev. 24(4), 416–425. https://doi.org/10.1139/er-2016-0029 (2016).Article 

    Google Scholar 
    Diamond, S. E. & Martin, R. A. Physiological adaptation to cities as a proxy to forecast global-scale responses to climate change. J. Exp. Biol. 224((Suppl_1)), jeb22336. https://doi.org/10.1242/jeb.229336 (2021).Article 

    Google Scholar 
    Grimm, N. B. et al. Global change and the ecology of cities. Science 319(5864), 756–760. https://doi.org/10.1126/science.1150195 (2008).Article 
    ADS 
    CAS 

    Google Scholar 
    McDonnell, M. J. & Pickett, S. T. Ecosystem structure and function along urban-rural gradients: An unexploited opportunity for ecology. Ecology 71(4), 1232–1237. https://doi.org/10.2307/1938259 (1990).Article 

    Google Scholar 
    Francis, R. A. & Chadwick, M. A. What makes a species synurbic?. Appl. Geogr. 32(2), 514–521. https://doi.org/10.1016/j.apgeog.2011.06.013 (2012).Article 

    Google Scholar 
    Luniak, M. Synurbization–adaptation of animal wildlife to urban development. In Proc. 4th Int. Symposium Urban Wildl. Conserv (Tucson, University of Arizona, 2004).Coogan, S. C. P., Raubenheimer, D., Zantis, S. P. & Machovsky-Capuska, G. E. Multidimensional nutritional ecology and urban birds. Ecosphere 9(4), e02177. https://doi.org/10.1002/ecs2.2177 (2018).Article 

    Google Scholar 
    Lowry, H., Lill, A. & Wong, B. B. Behavioural responses of wildlife to urban environments. Biol. Rev. Camb. Philos. Soc. 88(3), 537–549. https://doi.org/10.1111/brv.12012 (2013).Article 

    Google Scholar 
    Łopucki, R., Klich, D., Ścibior, A. & Gołębiowska, D. Hormonal adjustments to urban conditions: Stress hormone levels in urban and rural populations of Apodemus agrarius. Urban Ecosyst. 22(3), 435–442. https://doi.org/10.1007/s11252-019-0832-8 (2019).Article 

    Google Scholar 
    McCleery, R. in Urban mammals in Urban Ecosystem Ecology (eds. Aitkenhead-Peterson, J., Volder, A.) 87–102 (American Society of Agronomy, 2010). https://doi.org/10.2134/agronmonogr55.c52010Uchida, K., Suzuki, K., Shimamoto, T., Yanagawa, H. & Koizumi, I. Seasonal variation of flight initiation distance in Eurasian red squirrels in urban versus rural habitat. J. Zool. 298(3), 225–231. https://doi.org/10.1111/jzo.12306 (2016).Article 

    Google Scholar 
    Kleerekoper, L., van Esch, M. & Salcedo, T. B. How to make a city climate-proof, addressing the urban heat island effect. Resour. Conserv. Recyl. 64, 30–38. https://doi.org/10.1016/j.resconrec.2011.06.004 (2012).Article 

    Google Scholar 
    Pickett, S. T. et al. Urban ecological systems: Scientific foundations and a decade of progress. J. Environ. Manag. 92(3), 331–362. https://doi.org/10.1016/j.jenvman.2010.08.022 (2011).Article 
    CAS 

    Google Scholar 
    Rizwan, A. M., Dennis, L. Y. & Chunho, L. A review on the generation, determination and mitigation of Urban Heat Island. J. Environ. Sci. 20(1), 120–128 (2008).Article 
    CAS 

    Google Scholar 
    Isaksson, C. Urban ecophysiology: Beyond costs, stress and biomarkers. J. Exp. Biol. 223(22), jeb203794. https://doi.org/10.1242/jeb.203794 (2020).Article 

    Google Scholar 
    Miles, L. S., Carlen, E. J., Winchell, K. M. & Johnson, M. T. J. Urban evolution comes into its own: Emerging themes and future directions of a burgeoning field. Evol. Appl. 14(1), 3–11. https://doi.org/10.1111/eva.13165 (2020).Article 

    Google Scholar 
    Gavett, A. P. & Wakeley, J. S. Blood constituents and their relation to diet in urban and rural house sparrows. Condor 88(3), 279–284. https://doi.org/10.2307/1368873 (1986).Article 

    Google Scholar 
    Murray, M. et al. Greater consumption of protein-poor anthropogenic food by urban relative to rural coyotes increases diet breadth and potential for human-wildlife conflict. Ecography 38(12), 1235–1242. https://doi.org/10.1111/ecog.01128 (2015).Article 

    Google Scholar 
    Pollock, C. J., Capilla-Lasheras, P., McGill, R. A. R., Helm, B. & Dominoni, D. M. Integrated behavioural and stable isotope data reveal altered diet linked to low breeding success in urban-dwelling blue tits (Cyanistes caeruleus). Sci. Rep. 7(1), 5014. https://doi.org/10.1038/s41598-017-04575-y (2017).Article 
    ADS 
    CAS 

    Google Scholar 
    Schulte-Hostedde, A. I., Mazal, Z., Jardine, C. M. & Gagnon, J. Enhanced access to anthropogenic food waste is related to hyperglycemia in raccoons (Procyon lotor). Conserv. Physiol. 6(1), coy026. https://doi.org/10.1093/conphys/coy026 (2018).Article 
    CAS 

    Google Scholar 
    Fingland, K., Ward, S. J., Bates, A. J. & Bremner-Harrison, S. A systematic review into the suitability of urban refugia for the Eurasian red squirrel Sciurus vulgaris. Mamm. Rev. 52(1), 26–38. https://doi.org/10.1111/mam.12264 (2021).Article 

    Google Scholar 
    Jokimäki, J., Selonen, V., Lehikoinen, A. & Kaisanlahti-Jokimäki, M.-L. The role of urban habitats in the abundance of red squirrels (Sciurus vulgaris, L.) in Finland. Urban For. Urban Green. 27, 100–108. https://doi.org/10.1016/j.ufug.2017.06.021 (2017).Article 

    Google Scholar 
    Dausmann, K. H., Wein, J., Turner, J. M. & Glos, J. Absence of heterothermy in the European red squirrel (Sciurus vulgaris). Mammal. Biol. 78(5), 332–335. https://doi.org/10.1016/j.mambio.2013.01.004 (2013).Article 

    Google Scholar 
    Turner, J. M., Reher, S., Warnecke, L. & Dausmann, K. H. Eurasian red squirrels show little seasonal variation in metabolism in food-enriched habitat. Physiol. Biochem. Zool. 90(6), 655–662. https://doi.org/10.1086/694847 (2017).Article 

    Google Scholar 
    McNab, B. K. On the comparative ecological and evolutionary significance of total and mass-specific rates of metabolism. Physiol. Biochem. Zool. 72(5), 642–644 (1999).Article 
    CAS 

    Google Scholar 
    Menzies, A. K. et al. Body temperature, heart rate, and activity patterns of two boreal homeotherms in winter: Homeostasis, allostasis, and ecological coexistence. Funct. Ecol. 34(11), 2292–2301. https://doi.org/10.1111/1365-2435.13640 (2020).Article 

    Google Scholar 
    Wauters, L. & Dhondt, A. Activity budget and foraging behaviour of the red squirrel (Sciurus vulgaris Linnaeus, 1758) in a coniferous habitat. Z. Säugetierkd. 52(6), 341–353 (1987).
    Google Scholar 
    Wauters, L., Swinnen, C. & Dhondt, A. A. Activity budget and foraging behaviour of red squirrels (Sciurus vulgaris) in coniferous and deciduous habitats. J. Zool. 227(1), 71–86. https://doi.org/10.1111/j.1469-7998.1992.tb04345.x (1992).Article 

    Google Scholar 
    Reher, S., Dausmann, K. H., Warnecke, L. & Turner, J. M. Food availability affects habitat use of Eurasian red squirrels (Sciurus vulgaris) in a semi-urban environment. J. Mammal. 97(6), 1543–1554. https://doi.org/10.1093/jmammal/gyw105 (2016).Article 

    Google Scholar 
    Moller, H. Foods and foraging behavior of red (Sciurus vulgaris) and grey (Sciurus carolinensis) squirrels. Mammal. Rev. 13(2–4), 81–98. https://doi.org/10.1111/j.1365-2907.1983.tb00270.x (1983).Article 

    Google Scholar 
    Krauze-Gryz, D. & Gryz, J. in A review of the diet of the red squirrel (Sciurus vulgaris) in different types of habitats in Red squirrels: Ecology, conservation & management in Europe (eds. Shuttleworth, C. M., Lurz, P. W. W., Hayward, M. W.) 39–50 (European Squirrel Initiative, London, 2015)Shuttleworth, C. M. in The effect of supplemental feeding on the red squirrel (Sciurus vulgaris), Doctoral dissertation (University of London, London, 1996).Birnie-Gauvin, K., Peiman, K. S., Raubenheimer, D. & Cooke, S. J. Nutritional physiology and ecology of wildlife in a changing world. Conserv. Physiol. https://doi.org/10.1093/conphys/cox030 (2017).Article 

    Google Scholar 
    Wist, B., Stolter, C. & Dausmann, K. H. Sugar addicted in the city: Impact of urbanisation on food choice and diet composition of the Eurasian red squirrel (Sciurus vulgaris). J. Urban Ecol. 8(1), juac012. https://doi.org/10.1093/jue/juac012 (2022).Article 

    Google Scholar 
    Burton, T., Killen, S. S., Armstrong, J. D. & Metcalfe, N. B. What causes intraspecific variation in resting metabolic rate and what are its ecological consequences?. Proc. Biol. Sci. 278(1724), 3465–3473. https://doi.org/10.1098/rspb.2011.1778 (2011).Article 
    CAS 

    Google Scholar 
    Clarke, A. Costs and consequences of evolutionary temperature adaptation. Trends Ecol. Evol. 18(11), 573–581. https://doi.org/10.1016/j.tree.2003.08.007 (2003).Article 

    Google Scholar 
    Lovegrove, B. G. The influence of climate on the basal metabolic rate of small mammals: A slow-fast metabolic continuum. J. Comp. Physiol. B 173(2), 87–112. https://doi.org/10.1007/s00360-002-0309-5 (2003).Article 
    CAS 

    Google Scholar 
    McNab, B. K. The energetics of endotherms. Ohio J. Sci. 74(6), 370–380 (1974).
    Google Scholar 
    Tattersall, G. J. et al. Coping with thermal challenges: Physiological adaptations to environmental temperatures. Compr. Physiol. 2(3), 2151–2202 (2012).Article 

    Google Scholar 
    Broggi, J. et al. Sources of variation in winter basal metabolic rate in the great tit. Funct. Ecol. 21(3), 528–533. https://doi.org/10.1111/j.1365-2435.2007.01255.x (2007).Article 

    Google Scholar 
    Schlünzen, K. H., Hoffmann, P., Rosenhagen, G. & Riecke, W. Long-term changes and regional differences in temperature and precipitation in the metropolitan area of Hamburg. Int. J. Climatol. 30(8), 1121–1136. https://doi.org/10.1002/joc.1968 (2010).Article 

    Google Scholar 
    Reher, S. & Dausmann, K. H. Tropical bats counter heat by combining torpor with adaptive hyperthermia. Proc. R. Soc. B Biol. Sci. 288(1942), 20202059. https://doi.org/10.1098/rspb.2020.2059 (2021).Article 

    Google Scholar 
    Rezende, E. L. & Bacigalupe, L. D. Thermoregulation in endotherms: Physiological principles and ecological consequences. J. Comp. Physiol. B 185(7), 709–727. https://doi.org/10.1007/s00360-015-0909-5 (2015).Article 
    CAS 

    Google Scholar 
    Scholander, P. F., Hock, R., Walters, V., Johnson, F. & Irving, L. Heat regulation in some arctic and tropical mammals and birds. Biol. Bull. 99(2), 237–258. https://doi.org/10.2307/1538741 (1950).Article 
    CAS 

    Google Scholar 
    Terblanche, J. S., Clusella-Trullas, S., Deere, J. A., Van Vuuren, B. J. & Chown, S. L. Directional evolution of the slope of the metabolic rate-temperature relationship is correlated with climate. Physiol. Biochem. Zool. 82(5), 495–503. https://doi.org/10.1086/605361 (2009).Article 

    Google Scholar 
    Gallo, K. P., Easterling, D. R. & Peterson, T. C. The influence of land use/land cover on climatological values of the diurnal temperature range. J. Clim. 9(11), 2941–2944. https://doi.org/10.1175/1520-0442(1996)009%3c2941:TIOLUC%3e2.0.CO;2 (1996).Article 
    ADS 

    Google Scholar 
    Wang, K. et al. Urbanization effect on the diurnal temperature range: Different roles under solar dimming and brightening. J. Clim. 25(3), 1022–1027. https://doi.org/10.1175/jcli-d-10-05030.1 (2012).Article 
    ADS 

    Google Scholar 
    Fristoe, T. S. et al. Metabolic heat production and thermal conductance are mass-independent adaptations to thermal environment in birds and mammals. Proc. Natl. Acad. Sci. USA 112(52), 15934–15939. https://doi.org/10.1073/pnas.1521662112 (2015).Article 
    ADS 
    CAS 

    Google Scholar 
    Sándor, K. et al. Urban nestlings have reduced number of feathers in Great Tits (Parus major). Ibis 163(4), 1369–1378. https://doi.org/10.1111/ibi.12948 (2021).Article 

    Google Scholar 
    Beliniak, A., Krauze-Gryz, D., Jasińska, K., Jankowska, K. & Gryz, J. Contrast in daily activity patterns of red squirrels inhabiting urban park and urban forest. Hystrix https://doi.org/10.4404/hystrix-00476-2021 (2021).Article 

    Google Scholar 
    Thomas, L. S., Teich, E., Dausmann, K., Reher, S. & Turner, J. M. Degree of urbanisation affects Eurasian red squirrel activity patterns. Hystrix 29(2), 175–180. https://doi.org/10.4404/hystrix-00065-2018 (2018).Article 

    Google Scholar 
    Krauze-Gryz, D., Gryz, J. & Brach, M. Spatial organization, behaviour and feeding habits of red squirrels: Differences between an urban park and an urban forest. J. Zool. 315(1), 69–78. https://doi.org/10.1111/jzo.12905 (2021).Article 

    Google Scholar 
    Jarman, T. E., Gartrell, B. D. & Battley, P. F. Differences in body composition between urban and rural mallards Anas platyrhynchos. J. Urban Ecol. 6(1), juaa011. https://doi.org/10.1093/jue/juaa011 (2020).Article 

    Google Scholar 
    Cruz-Neto, A. P. & Bozinovic, F. The relationship between diet quality and basal metabolic rate in endotherms: Insights from intraspecific analysis. Physiol. Biochem. Zool. 77(6), 877–889 (2004).Article 

    Google Scholar 
    Geluso, K. & Hayes, J. P. Effects of dietary quality on basal metabolic rate and internal morphology of European starlings (Sturnus vulgaris). Physiol. Biochem. Zool. 72(2), 189–197 (1999).Article 
    CAS 

    Google Scholar 
    Seebacher, F. Is endothermy an evolutionary by-product?. Trends Ecol. Evol. 35(6), 503–511. https://doi.org/10.1016/j.tree.2020.02.006 (2020).Article 

    Google Scholar 
    Perissinotti, P. P., Antenucci, C. D., Zenuto, R. & Luna, F. Effect of diet quality and soil hardness on metabolic rate in the subterranean rodent Ctenomys talarum. Comp. Biochem. Physiol. Mol. Integr. Physiol. 154(3), 298–307. https://doi.org/10.1016/j.cbpa.2009.05.013 (2009).Article 
    CAS 

    Google Scholar 
    Thorp, C. R., Ram, P. K. & Florant, G. L. Diet alters metabolic rate in the yellow-bellied marmot (Marmota flaviventris) during hibernation. Physiol. Zool. 67(5), 1213–1229. https://doi.org/10.1086/physzool.67.5.30163890 (1994).Article 

    Google Scholar 
    Silva, S. I., Jaksic, F. M. & Bozinovic, F. Interplay between metabolic rate and diet quality in the South American fox Pseudalopex culpaeus. Comp. Biochem. Physiol. Mol Integr. Physiol. 137(1), 33–38. https://doi.org/10.1016/j.cbpb.2003.09.022 (2004).Article 
    CAS 

    Google Scholar 
    Rewkiewicz-Dziarska, A., Wielopolska, A. & Gill, J. Hematological indices of Apodemus agrarius (Pallas, 1771) from different urban environments. Bull. Acad. Polon. Sci. Ser. Sci. Biol. 25(4), 261–268 (1977).CAS 

    Google Scholar 
    Ohrnberger, S. A., Hambly, C., Speakman, J. R. & Valencak, T. G. Limits to sustained energy intake XXXII: Hot again: Dorsal shaving increases energy intake and milk output in golden hamsters (Mesocricetus auratus). J Exp. Biol. https://doi.org/10.1242/jeb.230383 (2020).Article 

    Google Scholar 
    Speakman, J. R. & Król, E. The heat dissipation limit theory and evolution of life histories in endotherms—Time to dispose of the disposable soma theory?. Integr. Comp. Biol. 50(5), 793–807. https://doi.org/10.1093/icb/icq049 (2010).Article 

    Google Scholar 
    Diamond, S. E., Chick, L. D., Perez, A., Strickler, S. A. & Martin, R. A. Evolution of thermal tolerance and its fitness consequences: Parallel and non-parallel responses to urban heat islands across three cities. Proc. R. Soc. B Biol. Sci. 285(1882), 20180036. https://doi.org/10.1098/rspb.2018.0036 (2018).Article 

    Google Scholar 
    Isaksson, C. & Hahs, A. Urbanization, oxidative stress and inflammation: A question of evolving, acclimatizing or coping with urban environmental stress. Funct. Ecol. 29(7), 913–923. https://doi.org/10.1111/1365-2435.12477 (2015).Article 

    Google Scholar 
    Sokolova, I. M. & Lannig, G. Interactive effects of metal pollution and temperature on metabolism in aquatic ectotherms: Implications of global climate change. Clim. Res. 37(2–3), 181–201 (2008).Article 

    Google Scholar 
    Carey, H. V., Andrews, M. T. & Martin, S. L. Mammalian hibernation: Cellular and molecular responses to depressed metabolism and low temperature. Physiol. Rev. 83(4), 1153–1181 (2003).Article 
    CAS 

    Google Scholar 
    Pereira, M. E., Aines, J. & Scheckter, J. L. Tactics of heterothermy in eastern gray squirrels (Sciurus carolinensis). J. Mammal. 83(2), 467–477 (2002).Article 

    Google Scholar 
    Breuner, C. W., Wingfield, J. C. & Romero, L. M. Diel rhythms of basal and stress-induced corticosterone in a wild, seasonal vertebrate. Gambel’s white-crowned sparrow. J Exp. Zool. 284(3), 334–342. https://doi.org/10.1002/(SICI)1097-010X(19990801)284:3%3c334::AID-JEZ11%3e3.0.CO;2-# (1999).Article 
    CAS 

    Google Scholar 
    Careau, V., Thomas, D., Humphries, M. M. & Réale, D. Energy metabolism and animal personality. Oikos 117(5), 641–653. https://doi.org/10.1111/j.0030-1299.2008.16513.x (2008).Article 

    Google Scholar 
    Fletcher, Q. E. et al. Seasonal stage differences overwhelm environmental and individual factors as determinants of energy expenditure in free-ranging red squirrels. Funct. Ecol. 26(3), 677–687. https://doi.org/10.1111/j.1365-2435.2012.01975.x (2012).Article 

    Google Scholar 
    Barthel, L. & Berger, A. Unexpected gene-flow in urban environments: The example of the European Hedgehog. Animals 10(12), 2315. https://doi.org/10.3390/ani10122315 (2020).Article 

    Google Scholar 
    Fusco, N. A., Carlen, E. J. & Munshi-South, J. Urban landscape genetics: are biologists keeping up with the pace of urbanization?. Current Landsc. Ecol. Rep. 6(2), 35–45. https://doi.org/10.1007/s40823-021-00062-3 (2021).Article 

    Google Scholar 
    Ziege, M. et al. Population genetics of the European rabbit along a rural-to-urban gradient. Sci. Rep. 10(1), 2448. https://doi.org/10.1038/s41598-020-57962-3 (2020).Article 
    ADS 
    CAS 

    Google Scholar 
    Morash, A. J., Neufeld, C., MacCormack, T. J. & Currie, S. The importance of incorporating natural thermal variation when evaluating physiological performance in wild species. J. Exp. Biol. 221(14), jeb164673. https://doi.org/10.1242/jeb.164673 (2018).Article 

    Google Scholar 
    Pörtner, H.-O., et al. Climate change 2022: Impacts, adaptation and vulnerability. IPCC Sixth Assessment Report (2022).Anderies, J. M., Katti, M. & Shochat, E. Living in the city: Resource availability, predation, and bird population dynamics in urban areas. J. Theor. Biol. 247(1), 36–49. https://doi.org/10.1016/j.jtbi.2007.01.030 (2007).Article 
    ADS 
    MATH 

    Google Scholar 
    Shochat, E. Credit or debit? Resource input changes population dynamics of city-slicker birds. Oikos 106(3), 622–626. https://doi.org/10.1111/j.0030-1299.2004.13159.x (2004).Article 

    Google Scholar 
    Koprowski, J. L. Handling tree squirrels with a safe and efficient restraint. Wildl. Soc. B 30(1), 101–103. https://doi.org/10.2307/3784642 (2002).Article 

    Google Scholar 
    Magris, L. & Gurnell, J. Population ecology of the red squirrel (Sciurus vulgaris) in a fragmented woodland ecosystem on the Island of Jersey Channel Islands. J. Zool. 256(1), 99–112. https://doi.org/10.1017/s0952836902000134 (2002).Article 

    Google Scholar 
    Bethge, J., Wist, B., Stalenberg, E. & Dausmann, K. Seasonal adaptations in energy budgeting in the primate Lepilemur leucopus. J Comp. Physiol. B 187(5–6), 827–834. https://doi.org/10.1007/s00360-017-1082-9 (2017).Article 

    Google Scholar 
    Dausmann, K. H., Glos, J. & Heldmaier, G. Energetics of tropical hibernation. J Comp. Physiol. B 179(3), 345–357. https://doi.org/10.1007/s00360-008-0318-0 (2009).Article 
    CAS 

    Google Scholar 
    Kobbe, S., Nowack, J. & Dausmann, K. H. Torpor is not the only option: Seasonal variations of the thermoneutral zone in a small primate. J. Comp. Physiol. B 184(6), 789–797. https://doi.org/10.1007/s00360-014-0834-z (2014).Article 

    Google Scholar 
    Lighton, J. R. Measuring Metabolic Rates: A Manual for Scientists (Oxford University Press, 2018).Book 

    Google Scholar 
    Bethge, J., Razafimampiandra, J. C., Wulff, A. & Dausmann, K. H. Sportive lemurs elevate their metabolic rate during challenging seasons and do not enter regular heterothermy. Conserv. Physiol. 9(1), coab075. https://doi.org/10.1093/conphys/coab075 (2021).Article 

    Google Scholar 
    Reher, S., Ehlers, J., Rabarison, H. & Dausmann, K. H. Short and hyperthermic torpor responses in the Malagasy bat Macronycteris commersoni reveal a broader hypometabolic scope in heterotherms. J. Comp. Physiol. B 188(6), 1015–1027. https://doi.org/10.1007/s00360-018-1171-4 (2018).Article 
    CAS 

    Google Scholar 
    Grolemund, G. & Wickham, H. Dates and times made easy with lubridate. J Stat. Softw. 40(3), 1–25 (2011).Article 

    Google Scholar 
    Wickham, H., François, R., Henry, L. & Müller, K. RStudio. dplyr: A Grammar of Data Manipulation (1.0. 7) (2021).Zeileis, A. & Grothendieck, G. zoo: S3 infrastructure for regular and irregular time series. J. Stat. Softw. 14(6), 1–27. https://doi.org/10.18637/jss.v014.i06 (2005).Article 

    Google Scholar 
    Sarkar, D. Lattice: Multivariate Data Visualization with R (Springer Science & Business Media, New York, 2008).Book 
    MATH 

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

    Google Scholar 
    Kuznetsova, A., Brockhoff, P. B. & Christensen, R. H. lmerTest package: Tests in linear mixed effects models. J. Stat. Softw. 82(13), 1–26 (2017).Article 

    Google Scholar 
    Wickham, H. ggplot2: Elegant graphics for data analysis (Springer, 2016).Book 
    MATH 

    Google Scholar 
    Fox, J. Effect displays in R for generalised linear models. J. Stat. Softw. 8(15), 1–27 (2003).Article 

    Google Scholar 
    Garamszegi, L. Z. et al. Changing philosophies and tools for statistical inferences in behavioral ecology. Behav. Ecol. 20(6), 1363–1375. https://doi.org/10.1093/beheco/arp137 (2009).Article 

    Google Scholar 
    Symonds, M. R. E. & Moussalli, A. A brief guide to model selection, multimodel inference and model averaging in behavioural ecology using Akaike’s information criterion. Behav. Ecol. Sociobiol. 65(1), 13–21. https://doi.org/10.1007/s00265-010-1037-6 (2010).Article 

    Google Scholar 
    Whittingham, M. J., Stephens, P. A., Bradbury, R. B. & Freckleton, R. P. Why do we still use stepwise modelling in ecology and behaviour?. J. Anim. Ecol. 75(5), 1182–1189. https://doi.org/10.1111/j.1365-2656.2006.01141.x (2006).Article 

    Google Scholar 
    Barton, K. & Barton, M. K. MuMIn: Multi-Model Inference. R package version 1.43.17; https://CRAN.R-project.org/package=MuMIn (2020).Zuur, A., Ieno, E. N., Walker, N., Saveliev, A. A. & Smith, G. M. Mixed effects models and extensions in ecology with R ( Springer Science & Business Media 2009).Burnham, K. P. & Anderson, D. R. Multimodel inference: Understanding AIC and BIC in model selection. Soc. Method. Res. 33(2), 261–304 (2004).Article 

    Google Scholar 
    Johnson, J. B. & Omland, K. S. Model selection in ecology and evolution. Trends Ecol. Evol. 19(2), 101–108. https://doi.org/10.1016/j.tree.2003.10.013 (2004).Article 

    Google Scholar 
    Lorah, J. Effect size measures for multilevel models: Definition, interpretation, and TIMSS example. Large-scale Assess. Educ. 6(1), 8. https://doi.org/10.1186/s40536-018-0061-2 (2018).Article 

    Google Scholar 
    Selya, A. S., Rose, J. S., Dierker, L. C., Hedeker, D. & Mermelstein, R. J. A practical guide to calculating cohen’s f2, a measure of local effect size, from PROC MIXED. Front. Psychol. 3, 111–111. https://doi.org/10.3389/fpsyg.2012.00111 (2012).Article 

    Google Scholar 
    Lüdecke, D. sjPlot: Data visualization for statistics in social science. R package version 2.8.5 2020; https://CRAN.R-project.org/package=sjPlot (2020).Nakagawa, S., Johnson, P. C. & Schielzeth, H. The coefficient of determination R2 and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded. J. R. Soc. Interface 14(134), 20170213 (2017).Article 

    Google Scholar 
    Nakagawa, S. & Schielzeth, H. Repeatability for Gaussian and non-Gaussian data: A practical guide for biologists. Biol. Rev. Camb. Philos. Soc. 85(4), 935–956. https://doi.org/10.1111/j.1469-185X.2010.00141.x (2010).Article 

    Google Scholar 
    Stoffel, M. A., Nakagawa, S. & Schielzeth, H. rptR: Repeatability estimation and variance decomposition by generalized linear mixed-effects models. Methods Ecol. Evol. 8(11), 1639–1644. https://doi.org/10.1111/2041-210X.12797 (2017).Article 

    Google Scholar  More

  • in

    Response of cyanobacterial mats to ambient phosphate fluctuations: phosphorus cycling, polyphosphate accumulation and stoichiometric flexibility

    Our findings highlight the critical role of polyP in Sodalinema stali-formed cyanobacterial mats, as it was dynamically accumulated and recycled during acclimation to P fluctuations.Cellular response to progressive P starvationAnalogous to planktonic cyanobacteria, growth under low P availability could be sustained by recycling polyP, which acted as a primary P source (Fig. 2a) [16, 23, 24]. We further attribute the rapid reduction of easily dispensable cellular P-containing compounds to the substitution of cellular phospholipids with S- or N-containing membrane lipids to maintain growth at the onset of P stress (Fig. 2a) [15, 23]. However, the exhaustion of this easily dispensable P pool limited proliferation in Phase 2, and the metabolic strategy switched from a focus on growth towards maintenance (Fig. 5). The interpretation of prevailing cellular processes based on our results is graphically summarized and explained in detail below (Fig. 5).Fig. 5: Schematic interpretation of cellular phosphorus (P) cycling in a cyanobacterial mat, based on significant changes of the monitored parameters (arbitrary units).a At low P availability, initially contained polyphosphate (polyP) was recycled simultaneously with phosphate uptake to sustain growth at constant C:N:P ratios. Further proliferation at the onset of P stress in Phase 1 was sustained by mobilization of cellular P, e.g. phospholipids, which led to rapidly increasing C:N:P ratios. Severe P stress in Phase 2, indicated by increasing APase activity, prevented proliferation and photosynthesis, indicated by a loss of green chlorophyll pigments. PolyP accumulation by deficiency response occurs with severely increasing P stress, whereby globular DNA accumulation indicates the allocation of P contained in DNA into polyP. P re-addition to the P-stressed cells in Phase 3 triggered overplus uptake and narrow C:N:P ratios, transitioning to luxury uptake at higher C:N:P ratios following polyP recycling. b At high P availability, polyP in Phase 1 was accumulated by overplus uptake at narrow C:N:P ratios, transitioning to luxury uptake at higher C:N:P ratios during polyP recycling in Phase 2. P-deprivation in Phase 3 did not affect the cells, which we attributed to a sufficient amount of phosphate in the residual medium or within the biofilm matrix. Arrows indicate phosphorus transformation processes, whereby arrows pointing towards DNA represent cell growth. Yellow granules = polyP, blue granules = globular DNA spheres, P = phospholipids, S = substitute lipids.Full size imageSevere P stress in Phase 2 was indicated by the colour change from green towards yellow-green (Fig. S1) and increasing APase activity (Fig. 2a). The colour change suggested the loss of photosynthetic pigments [40], but we could not clarify whether this occurred through active cellular pigment reduction or degradation of available chlorophyll e.g., by oxidation. The increasing APase activity (Fig. 2a) suggested that Sodalinema stali is capable of hydrolysing organic P [14]. Even though APase expression did not trigger proliferation, it likely hydrolysed a potentially available organic P pool, as increasing DIC, NH4 and decreasing pH indicated progressive decay and remineralisation of organic matter (Fig. 1a). This suggests that in analogous oligotrophic environments with often fluctuating conditions, the strategy has to be maximizing the utilization of external P sources contained in organic and inorganic sediment particles that get trapped in the EPS [41]. The sediment can contain large amounts of organic P [42] and the fluctuating physico-chemical gradients in the EPS matrix due to high daytime pH and low oxygen conditions at night, facilitate P desorption from metal oxides, leading to higher dissolved phosphate concentrations within the mat, compared to the overlying water body [3]. However, alternating redox conditions at the SWI could also trigger polyP release from benthic microorganisms to the sediments, where it could act as a P source for the benthic food-chain, or ultimately trigger the formation of mineral P phases [32], to sustainably remove P from the aquatic cycle. Either way, we suggest that polyP-containing cyanobacterial mats critically impact P fluxes at the SWI.With persisting severe P stress and increasing APase activity in Phase 2, polyP accumulation as a deficiency response was observed (Fig. 2a), which has been reported from planktonic cyanobacteria of different habitats [24, 29, 23], as well as stream periphyton [28]. However, the reasons causing this deficiency response remain unresolved. In marine phytoplankton of the oligotrophic Sargasso Sea, Martin et al. [23] excluded that polyP-rich cells were in a perpetual overplus state with ‘undetectable’ pulses driving this state and suggested that polyP accumulation occurred as a cellular stress response. In other studies, reduced biosynthesis of P-rich rRNA coincided with deficiency responses [26, 28] and led to the suggestion that polyP accumulation at P concentrations below a certain threshold required for growth occurs because of P allocation changes away from growth and towards storage. Further, APase can hydrolytically cleave phosphate groups from nucleic acids and convert DNA-lipid-P to DNA-lipids, which were shown to self-assemble into globular lipid-based DNA micelles [43]. These preferentially anchor on cell membranes [44], and indeed, such DNA spheres were found to accumulate at the cell’s polar membranes in our experiments adjacent to polyP during deficiency response (Fig. 4a: Phase 2,c). Therefore, we suggest that intracellular P recovery by cleavage from P-rich DNA and reallocation to polyP, and potentially reduced rRNA synthesis [31], is also a strategy in benthic mats of Sodalinema stali as a response to severe P stress when P availability is too low to sustain growth. This supports the theory of a reallocation of resources away from growth towards flexibly available P and energy storage. Such direct intracellular P cycling could be beneficial to help retain P within the cyanobacterial population; while external P moieties such as dissolved organic P within the matrix can act as an additional P source, they are also likely to be subject to nutrient competition between cyanobacteria and other organisms inhabiting the matrix.Such effects of potential interactions in terms of nutrient competition or provision between cyanobacteria and mutualistic microorganisms contained within the same EPS matrix are difficult to assess and we cannot exclude some potential effects on our results. However, mutualistic microorganisms that are naturally contained in many cyanobacterial or algal cultures are often critical for metacommunity functioning and hence, working with axenic mat-forming strains may even further falsify any obtained results. Furthermore, microscopic analyses revealed that Sodalinema always dominated the biomass and hence, it is here considered reasonable to work with a non-axenic culture.Cellular response to a simulated P pulseIn P-deficient cells, the affinity of the P uptake system is typically increased to maximize P uptake for future pulses [13, 45]. The simulated P pulse to the P-stressed cells in Phase 3 led to a rapid increase of the cellular P content by 1260% relative to C within 3 days (Fig. 2a), whereby P was accumulated to a significant part as polyP, which is characteristic for overplus uptake [25]. Many different types of oligotrophic aquatic habitats experience only temporal P pulses, e.g., from redox changes at the benthic interface leading to P release from the sediment [32], storm run-off [28], upwelling [46], or excretions of aquatic animals [47]. The capability of microorganisms to immediately take up, store, and efficiently re-use this P by overplus uptake is hence of critical importance for a population to sustain a potential subsequent period of low P availability. Overplus uptake is typically accompanied by the overall slow growth of the population and cellular recovery from P starvation, including ultrastructural organization and recovery of the photosynthetic apparatus [48]. This took one week after re-feeding of P-starved Nostoc sp. PCC 7118 cells [48]—a timeframe very similar to the delayed onset of photosynthesis observed in our study, indicated by the elevated pH at day 9 (Fig. 1a). Regarding overplus-triggering mechanisms following P pulses, Solovochenko et al. [48] suggested that overplus uptake occurs due to a delayed down-regulation of high-affinity Pi transporters, which are active during P starvation, and emphasized the simultaneous advantage of osmotically inert polyP accumulation as a response to dramatically high phosphate concentrations in the cells. Even though APase levels declined following our experimental P re-addition, they were significantly elevated for at least 9 days (Fig. 2a). As our experimental design involved replacing the medium with APase-free, BG11 + medium after Phase 2, we assume that the APase detected in Phase 3 was actively produced, and we conclude that previously relevant, low-P response mechanisms are slowly disengaged with some sort of lag, even when ambient P is repleted. Following cellular recovery, Sodalinema now recycled stored polyP instead of further accumulating it during the transition from overplus-to luxury uptake, which was reflected in the increasing C:N:P molar ratios and decreasing polyP levels without significant additional phosphate uptake (Figs. 1a, 2a, 5).Qualitative observations on polyP distributionMost methods applied to analyse polyP in microorganisms are quantitative and do not contain information on its spatial distribution within a population. The here observed variable distribution of polyP between the cells during luxury uptake and deficiency response, as well as the retention of polyP in few individual filaments during polyP recycling in Phase 1 of the low P experiment (Fig. 4) suggests strategies of either slow growth with a retention of polyP, or of high growth with polyP recycling. This was also suggested for cells of a unicellular Synechocystis sp. PCC 6803 population during overplus uptake [47]. In contrast, polyP in our experiment was distributed homogeneously between all cyanobacterial cells during overplus uptake (Fig. 4a: Phase 3, Fig. 4b: Phase 1). Yet, we are unaware of any polyP distribution study in multicellular or mat-forming cyanobacteria and hence, further mechanisms of interactions, e.g., cell-to-cell communication [49, 50], might also contribute to purposeful differentiation of cells or filaments within a common matrix.In summary, our study shows that the mat-forming Sodalinema stali (1) is capable of luxury uptake, overplus uptake and deficiency response with a heterogenous polyP distribution during polyP recycling, luxury uptake and deficiency response, while (2) dynamically adjusting cellular P content to changing phosphate concentrations. (3) Proliferation is sustained under the expense of polyP, followed by P acquisition from other easily dispensable cellular P-containing compounds under the onset of P stress. (4) Further, biosynthetic allocation changes away from growth towards maintenance with relative polyP accumulation at the expense of P-rich DNA are conducted under severe P stress. Our findings demonstrate the extraordinary capabilities of mat-forming cyanobacteria to adapt their P acquisition strategies to strong P fluctuations. While lasting proliferation under P limitation requires the mobilization of additional P sources through regeneration of P from particulate matter, the transition to net P accumulation under excess ambient P is rapid and effective. Since current projections of climate and land use change include intensified pulses of P load to aquatic ecosystems [50], e.g., through external input from surplus of agriculture fertilizer, inefficient wastewater treatment plants, and internal loads via the mobilization of legacy P, these P ‘bioaccumulators’ could form an important component in P remediation by temporarily accumulating P within the mat, and synthesizing polyP that could ultimately stimulate the formation of mineral P phases to sustainably remove P from the aquatic cycle. More

  • in

    Formation of necromass-derived soil organic carbon determined by microbial death pathways

    Bradford, M. A. et al. Soil carbon science for policy and practice. Nat. Sustain. 2, 1070–1072 (2019).Article 

    Google Scholar 
    Lehmann, J. & Kleber, M. The contentious nature of soil organic matter. Nature 528, 60–68 (2015).Article 

    Google Scholar 
    Liang, C., Schimel, J. P. & Jastrow, J. D. The importance of anabolism in microbial control over soil carbon storage. Nat. Microbiol. 2, 17105 (2017).Article 

    Google Scholar 
    Liang, C., Amelung, W., Lehmann, J. & Kästner, M. Quantitative assessment of microbial necromass contribution to soil organic matter. Glob. Change Biol. 25, 3578–3590 (2019).Article 

    Google Scholar 
    Wang, B. R., An, S. S., Liang, C., Liu, Y. & Kuzyakov, Y. Microbial necromass as the source of soil organic carbon in global ecosystems. Soil Biol. Biochem. 162, 108422 (2021).Article 

    Google Scholar 
    Kästner, M. & Miltner, A. in The Future of Soil Carbon (eds Garcia, C. et al.) Ch. 5 (Academic Press, 2018).Buckeridge, K. M. et al. Sticky dead microbes: rapid abiotic retention of microbial necromass in soil. Soil Biol. Biochem. 149, 107929 (2020).Article 

    Google Scholar 
    Kallenbach, C. M., Grandy, A. S., Frey, S. D. & Diefendorf, A. F. Microbial physiology and necromass regulate agricultural soil carbon accumulation. Soil Biol. Biochem. 91, 279–290 (2015).Article 

    Google Scholar 
    Kallenbach, C. M., Frey, S. D. & Grandy, A. S. Direct evidence for microbial-derived soil organic matter formation and its ecophysiological controls. Nat. Commun. 7, 13630 (2016).Article 

    Google Scholar 
    Emerson, J. B. et al. Schrödinger’s microbes: tools for distinguishing the living from the dead in microbial ecosystems. Microbiome 5, 86 (2017).Article 

    Google Scholar 
    Zhang, Y. et al. Simulating measurable ecosystem carbon and nitrogen dynamics with the mechanistically defined MEMS 2.0 model. Biogeosciences 18, 3147–3171 (2021).Article 

    Google Scholar 
    Ackermann, M., Stearns Stephen, C. & Jenal, U. Senescence in a bacterium with asymmetric division. Science 300, 1920–1920 (2003).Article 

    Google Scholar 
    Aguilaniu, H., Gustafsson, L., Rigoulet, M. & Nyström, T. Asymmetric inheritance of oxidatively damaged proteins during cytokinesis. Science 299, 1751–1753 (2003).Article 

    Google Scholar 
    Maheshwari, R. & Navaraj, A. Senescence in fungi: the view from Neurospora. FEMS Microbiol. Lett. 280, 135–143 (2008).Article 

    Google Scholar 
    See, C. R. et al. Hyphae move matter and microbes to mineral microsites: integrating the hyphosphere into conceptual models of soil organic matter stabilization. Glob. Change Biol. 28, 2527–2540 (2022).Article 

    Google Scholar 
    Pusztahelyi, T. et al. Comparative studies of differential expression of chitinolytic enzymes encoded by chiA, chiB, chiC and nagA genes in Aspergillus nidulans. Folia Microbiologica 51, 547–554 (2006).Article 

    Google Scholar 
    Bartoszewska, M. & Kiel, J. A. The role of macroautophagy in development of filamentous fungi. Antioxid. Redox Signal. 14, 2271–2287 (2011).Article 

    Google Scholar 
    Josefsen, L. et al. Autophagy provides nutrients for nonassimilating fungal structures and is necessary for plant colonization but not for infection in the necrotrophic plant pathogen Fusarium graminearum. Autophagy 8, 326–337 (2012).Article 

    Google Scholar 
    Heaton, L. L., Jones, N. S. & Fricker, M. D. Energetic constraints on fungal growth. Am. Nat. 187, E27–E40 (2016).Article 

    Google Scholar 
    Taiz, L. & Zeiger, E. Plant Physiology 4th edn (Spektrum Akademischer Verlag, 2008).Bowman, E. J. & Bowman, B. J. in Cellular and Molecular Biology of Filamentous Fungi (eds Borkovich, K. & Ebbole, D.) 179–190 (ASM Press, 2010).Voigt, O. & Pöggeler, S. Self-eating to grow and kill: autophagy in filamentous ascomycetes. Appl. Microbiol. Biotechnol. 97, 9277–9290 (2013).Article 

    Google Scholar 
    Grimmett, I. J., Shipp, K. N., Macneil, A. & Barlocher, F. Does the growth rate hypothesis apply to aquatic hyphomycetes? Fungal Ecol. 6, 493–500 (2013).Article 

    Google Scholar 
    Camenzind, T., Philipp Grenz, K., Lehmann, J. & Rillig, M. C. Soil fungal mycelia have unexpectedly flexible stoichiometric C:N and C:P ratios. Ecol. Lett. 24, 208–218 (2021).Article 

    Google Scholar 
    Mason-Jones, K., Robinson, S. L., Veen, G. F., Manzoni, S. & van der Putten, W. H. Microbial storage and its implications for soil ecology. ISME J. 16, 617–629 (2022).Article 

    Google Scholar 
    Gow, N. A. R., Latge, J. P. & Munro, C. A. The fungal cell wall: structure, biosynthesis, and function. Microbiol. Spectr. 5, FUNK-0035–2016 (2017).Article 

    Google Scholar 
    Steiner, U. K. Senescence in bacteria and its underlying mechanisms. Front. Cell Dev. Biol. 9, 668915 (2021).Article 

    Google Scholar 
    Allocati, N., Masulli, M., Di Ilio, C. & De Laurenzi, V. Die for the community: an overview of programmed cell death in bacteria. Cell Death Dis. 6, e1609 (2015).Article 

    Google Scholar 
    Peeters, S. H. & de Jonge, M. I. For the greater good: programmed cell death in bacterial communities. Microbiol. Res. 207, 161–169 (2018).Article 

    Google Scholar 
    Wang, J. & Bayles, K. W. Programmed cell death in plants: lessons from bacteria? Trends Plant Sci. 18, 133–139 (2013).Article 

    Google Scholar 
    Nagamalleswari, E., Rao, S., Vasu, K. & Nagaraja, V. Restriction endonuclease triggered bacterial apoptosis as a mechanism for long time survival. Nucleic Acids Res. 45, 8423–8434 (2017).Article 

    Google Scholar 
    Kysela, D. T., Brown, P. J. B., Huang, K. C. & Brun, Y. V. Biological consequences and advantages of asymmetric bacterial growth. Annu. Rev. Microbiol. 67, 417–435 (2013).Article 

    Google Scholar 
    Bayles, K. W. Bacterial programmed cell death: making sense of a paradox. Nat. Rev. Microbiol. 12, 63–69 (2014).Article 

    Google Scholar 
    Flemming, H.-C. & Wuertz, S. Bacteria and archaea on Earth and their abundance in biofilms. Nat. Rev. Microbiol. 17, 247–260 (2019).Article 

    Google Scholar 
    Coleman, D. C. & Wall, D. H. in Soil Microbiology, Ecology and Biochemistry 4th edn (ed. Paul, E. A.) Ch. 5 (Academic Press, 2015).Hungate, B. A. et al. The functional significance of bacterial predators. mBio 12, e00466-21 (2021).Article 

    Google Scholar 
    Kuzyakov, Y. & Mason-Jones, K. Viruses in soil: nano-scale undead drivers of microbial life, biogeochemical turnover and ecosystem functions. Soil Biol. Biochem. 127, 305–317 (2018).Article 

    Google Scholar 
    Williamson, K. E., Fuhrmann, J. J., Wommack, K. E. & Radosevich, M. Viruses in soil ecosystems: an unknown quantity within an unexplored territory. Annu. Rev. Virol. 4, 201–219 (2017).Article 

    Google Scholar 
    Sokol, N. W. et al. Life and death in the soil microbiome: how ecological processes influence biogeochemistry. Nat. Rev. Microbiol. 20, 415–430 (2022).Article 

    Google Scholar 
    Bonkowski, M. & Clarholm, M. J. A. P. Stimulation of plant growth through interactions of bacteria and protozoa: testing the auxiliary microbial loop hypothesis. Acta Protozool. 51, 237–247 (2012).
    Google Scholar 
    Potapov, A. M., Pollierer, M. M., Salmon, S., Šustr, V. & Chen, T.-W. Multidimensional trophic niche revealed by complementary approaches: gut content, digestive enzymes, fatty acids and stable isotopes in Collembola. J. Anim. Ecol. 90, 1919–1933 (2021).Article 

    Google Scholar 
    Esteban, G. F. & Fenchel, T. M. in Ecology of Protozoa: The Biology of Free-living Phagotrophic Protists (eds Esteban, G. F. & Fenchel, T. M.) 33–54 (Springer, 2020).Koksharova, O. A. Bacteria and phenoptosis. Biochemistry 78, 963–970 (2013).
    Google Scholar 
    Tilman, D. Resource Competition and Community Structure (Princeton Univ. Press, 1982).Boddy, L. Interspecific combative interactions between wood-decaying basidiomycetes. FEMS Microbiol. Ecol. 31, 185–194 (2000).Article 

    Google Scholar 
    Hibbing, M. E., Fuqua, C., Parsek, M. R. & Peterson, S. B. Bacterial competition: surviving and thriving in the microbial jungle. Nat. Rev. Microbiol. 8, 15–25 (2010).Article 

    Google Scholar 
    Müller, S. et al. Predation by Myxococcus xanthus induces Bacillus subtilis to form spore-filled megastructures. Appl. Environ. Microbiol. 81, 203–210 (2015).Article 

    Google Scholar 
    Laskowska, E. & Kuczynska-Wisnik, D. New insight into the mechanisms protecting bacteria during desiccation. Curr. Genet. 66, 313–318 (2020).Article 

    Google Scholar 
    Rillig, M. C., Ryo, M. & Lehmann, A. Classifying human influences on terrestrial ecosystems. Glob. Change Biol. 27, 2273–2278 (2021).Article 

    Google Scholar 
    Dörr, T., Moynihan, P. J. & Mayer, C. Bacterial cell wall structure and dynamics. Front. Microbiol. 10, 02051 (2019).Article 

    Google Scholar 
    Corredor, B., Lang, B. & Russell, D. Effects of nitrogen fertilization on soil fauna in a global meta-analysis. Preprint at Res. Sq. https://doi.org/10.21203/rs.3.rs-1438491/v1 (2022).Blankinship, J. C., Niklaus, P. A. & Hungate, B. A. A meta-analysis of responses of soil biota to global change. Oecologia 165, 553–565 (2011).Article 

    Google Scholar 
    Manzoni, S., Chakrawal, A., Spohn, M. & Lindahl, B. D. Modeling microbial adaptations to nutrient limitation during litter decomposition. Front. For. Glob. Change 4, 686945 (2021).Article 

    Google Scholar 
    Frank, D. et al. Effects of climate extremes on the terrestrial carbon cycle: concepts, processes and potential future impacts. Glob. Change Biol. 21, 2861–2880 (2015).Article 

    Google Scholar 
    Gunina, A. & Kuzyakov, Y. From energy to (soil organic) matter. Glob. Change Biol. 28, 2169–2182 (2022).Article 

    Google Scholar 
    Fernandez, C. W. & Koide, R. T. Initial melanin and nitrogen concentrations control the decomposition of ectomycorrhizal fungal litter. Soil Biol. Biochem. 77, 150–157 (2014).Article 

    Google Scholar 
    Kästner, M., Miltner, A., Thiele-Bruhn, S. & Liang, C. Microbial necromass in soils—linking microbes to soil processes and carbon turnover. Front. Environ. Sci. 9, 756378 (2021).Article 

    Google Scholar 
    Buckeridge, K. M., Creamer, C. & Whitaker, J. Deconstructing the microbial necromass continuum to inform soil carbon sequestration. Funct. Ecol. 36, 1396–1410 (2022).Article 

    Google Scholar 
    Lehmann, J. et al. Persistence of soil organic carbon caused by functional complexity. Nat. Geosci. 13, 529–534 (2020).Article 

    Google Scholar 
    Blazewicz, S. J. et al. Taxon-specific microbial growth and mortality patterns reveal distinct temporal population responses to rewetting in a California grassland soil. ISME J. 14, 1520–1532 (2020).Article 

    Google Scholar 
    Kallenbach, C. M., Wallenstein, M. D., Schipanksi, M. E. & Grandy, A. S. Managing agroecosystems for soil microbial carbon use efficiency: ecological unknowns, potential outcomes, and a path forward. Front. Microbiol. 10, 1146 (2019).Article 

    Google Scholar 
    Liang, C. Soil microbial carbon pump: mechanism and appraisal. Soil Ecol. Lett. 2, 241–254 (2020).Article 

    Google Scholar 
    Sinsabaugh, R. L., Manzoni, S., Moorhead, D. L. & Richter, A. Carbon use efficiency of microbial communities: stoichiometry, methodology and modelling. Ecol. Lett. 16, 930–939 (2013).Article 

    Google Scholar 
    van Groenigen, J. W. et al. Sequestering soil organic carbon: a nitrogen dilemma. Environ. Sci. Technol. 51, 4738–4739 (2017).Article 

    Google Scholar 
    Greenlon, A. et al. Quantitative stable-isotope probing (qSIP) with metagenomics links microbial physiology and activity to soil moisture in Mediterranean-climate grassland ecosystems (in the press).Mafla-Endara, P. M. et al. Microfluidic chips provide visual access to in situ soil ecology. Commun. Biol. 4, 889 (2021).Article 

    Google Scholar 
    Schaible, G. A., Kohtz, A. J., Cliff, J. & Hatzenpichler, R. Correlative SIP-FISH-Raman-SEM-NanoSIMS links identity, morphology, biochemistry, and physiology of environmental microbes. ISME Commun. 2, 52 (2022).Article 

    Google Scholar 
    See, C. R. et al. Distinct carbon fractions drive a generalisable two-pool model of fungal necromass decomposition. Funct. Ecol. 35, 796–806 (2021).Article 

    Google Scholar 
    Wang, C. et al. Stabilization of microbial residues in soil organic matter after two years of decomposition. Soil Biol. Biochem. 141, 107687 (2020).Article 

    Google Scholar 
    Veresoglou, S. D., Halley, J. M. & Rillig, M. C. Extinction risk of soil biota. Nat. Commun. 6, 8862 (2015).Article 

    Google Scholar 
    Potapov, A. M. et al. Feeding habits and multifunctional classification of soil-associated consumers from protists to vertebrates. Biol. Rev. 97, 1057–1117 (2022).Article 

    Google Scholar 
    Trap, J., Bonkowski, M., Plassard, C., Villenave, C. & Blanchart, E. Ecological importance of soil bacterivores for ecosystem functions. Plant Soil 398, 1–24 (2016).Article 

    Google Scholar 
    Dooley, S. R. & Treseder, K. K. The effect of fire on microbial biomass: a meta-analysis of field studies. Biogeochemistry 109, 49–61 (2012).Article 

    Google Scholar 
    Muñoz-Leoz, B., Ruiz-Romera, E., Antigüedad, I. & Garbisu, C. Tebuconazole application decreases soil microbial biomass and activity. Soil Biol. Biochem. 43, 2176–2183 (2011).Article 

    Google Scholar 
    Meyer, M., Diehl, D., Schaumann, G. E. & Muñoz, K. Agricultural mulching and fungicides—impacts on fungal biomass, mycotoxin occurrence, and soil organic matter decomposition. Environ. Sci. Pollut. Res. 28, 36535–36550 (2021).Article 

    Google Scholar 
    Thiery, S. & Kaimer, C. The predation strategy of Myxococcus xanthus. Front. Microbiol. 11, 2 (2020).Article 

    Google Scholar 
    Laloux, G. Shedding light on the cell biology of the predatory bacterium Bdellovibrio bacteriovorus. Front. Microbiol. 10, 3136 (2020).Article 

    Google Scholar  More