Ecology

Both in Polistes biglumis and P. gallicus in most of the inhabited nests all types of brood were present: eggs, larvae and pupae (Table S1), with the exception of one foundress nest of P. biglumis with only one egg. The size of thermographed nests was quite variable in both species, the number of cells ranging from 18 to 99 in P. biglumis (mean: 61.6 cells), and from 19 to 381 in P. gallicus (mean: 101.7 cells) (Table S1). The mean number of wasps on the thermographed nests was higher in P. gallicus (12.6 wasps) than in P. biglumis (7.1 wasps). All nests of Polistes biglumis we observed in this study were built on stone substrate or walls (Figs. 1c, 2a). Only recently we found one nest built on a pile of wood. The choice of the nest substrate was more diverse in P. gallicus (Figs. 1d, 2b). They chose stone, concrete, walls, window grilles, and metal of fences or doorframes.Figure 2Examples of nests and fieldwork set-up in Obergail (a) and Sesto Fiorentino (b). 1 = thermocouple wire; 2 = global radiation sensor, 3 = Peltier-element IR reference source.Full size imageDaily nest temperature coursePolistes biglumisFigure 3 shows a sequence of thermograms of a P. biglumis nest taken from dawn to dusk. Before sunrise the temperatures of the nest and of the wasps on it were quite low (mean ~ 15 °C) and uniform (~ 12 to 17.5 °C; Fig. 3a). The temperature of the stone substrate where the nest was built on was considerably higher (~ 20 °C). After sunrise (Fig. 3b,c) the nest temperature began to rise quickly. It only needed 13 min of sunshine (radiation) to heat the nest from ~ 17 to ~ 25 °C. Within one hour, temperature differences of almost 20 °C were measured within the nest. At 6:50, when the highest temperature on the nest was already at 36.2 °C, fast movements of the adults with inspections of the cells were observed (Fig. 3c). Soon afterwards the increasing temperature induced the wasps to start fanning (arrow in Fig. 3d). The wasps also began to gather water and spread it on and inside cells to cool the nest by evaporation (Fig. 3d,e). Towards late morning, some parts of the nest reached temperatures as high as 46 °C (Fig. 3e)! As soon as the nest was shaded by the substrate (~ 13:00) the nest temperature decreased according to the decrease in ambient temperature (Fig. 3f,g), reaching ~ 21 °C on average after dusk (Fig. 3h). At that time the substrate temperature (~ 25 °C) was still about 4 °C higher than the nest temperature.Figure 3Thermograms of a P. biglumis nest during a whole day (19.07.2017). (a) Before sunrise at 6:20; (b) during sunrise (06:33); (c) nest temperature increasing fast in sunshine; (d) with a fanner for convective nest cooling (arrow; see also Fig. S4); (e) with water drops for evaporative cooling when sunshine increased part of the nest to temperatures  > 45 °C; (f,g) after sunset (nest now in shade) in the afternoon; (h) at dusk with wasps sitting motionless on the nest. Time = CEST = UTC + 2 h.Full size imageThe nest and body temperatures of a complete 24 h cycle of a different nest are shown in Fig. 4a. At night the nest temperature and the wasps’ thorax temperature decreased slowly according to the decrease of the air temperature. The substrate temperature was always higher than the mean nest temperature, which surely helped to keep the nest temperature higher than the temperature of the surrounding air (Tanest). Variation of within-nest temperature (max–min) was low at night. As soon as solar radiation increased in early morning, the nest temperature and the body temperature of the wasps on it increased rapidly, and the variation of nest temperature (max–min) increased (see also Fig. 3b). Though the maximum nest temperature reached values as high as 46.9 °C, cooling measures of the wasps (fanning and spreading of water drops, see below) kept the mean nest temperature always below 38.5 °C. Cooling of the nest after sunset (at the nest) was much slower than the increase in the morning, following the decrease of ambient and substrate temperature (Fig. 4a,b).Figure 4Examples of daily temperature changes of nests and wasps of P. biglumis (a,b) and P. gallicus (c,d). Tthorax = mean thorax surface temperature of up to five adult individuals per time of measurement; gray ribbon: total range of nest temperatures (Tmax:Tmin) with mean; Tsubstrate = temperature beside the nest (see Fig. S1c,d); Tanest = ambient air temperature directly at the nest. Ta = ambient air temperature in shade 1–3 m away from nest; Radiation = global radiation hitting the nest; black bars = fanning events at the time of thermographic measurements: actually, many more fanning events were observed. (c) Fanning was never observed! See also Fig. S2 for another example of a P. gallicus nest in shade. Time = CEST = UTC + 2 h.Full size imagePolistes gallicusMost P. gallicus nests were built in locations with no or only little direct sunshine (Figs. 2b, 4c, Fig. S2). In their habitats temperatures in midsummer are often already quite high in the morning, and may increase to values higher than 40 °C during the day (Fig. 4d). Mean temperatures of the nest and of the imagines on it were usually higher than the air temperature close to the nest (Tanest). In most nests variation of within-nest temperature (max–min) remained small throughout the day. On hot days (Tanest  > 40 °C), however, maximum temperatures of empty cells in the nest margin sometimes reached values as high as 49.9 °C even in shade. Body temperature of the adults was mostly similar to the mean nest temperature (Fig. 4c, Fig. S2). At night, the nest temperature decreased according to the decrease of Tanest, similar to P. biglumis but at a higher level (Fig. 4d).The situation was different in one large nest which had been built in a location exposed to the morning sun (Figs. 4d, 5). On a hot day when Tanest increased to values higher than 42 °C, the body temperature of the adults increased to values up to 5 °C higher than the mean nest temperature. Nevertheless, though the combined effects of high air temperature and intense insolation increased part of the nest to a temperature of ~ 58 °C (Fig. 4d), mean nest temperature was kept below 41 °C. This was accomplished by cooling with many water droplets in the cells (dark spots in Fig. 5), and by the occurrence of fanning during the period when the sun was shining on the nest (Fig. 4d; see arrows in Fig. 5c). Fanning, however, was quite rare in all the other observed nests, even during the hottest time of the day! Water droplets were carried onto this nest until evening (Fig. 5h), as at that time the nest temperature was still at about 35–38 °C.Figure 5Thermograms of a large P. gallicus nest during a whole day (01.08.2017). Thermograms are rotated 90° clockwise (the upper part is on the right). (a) Before sunrise (6:36); (b) during sunrise (06:46) with the first water drops visible (dark spots); (c) with two fanners for convective nest cooling (arrows, see also Fig. 4d); (d) with more cooling drops; (e) after sunset at the nest site (nest now in shade); (f–h) after sunset in the afternoon and evening. Time = CEST = UTC + 2 h. For temperature evaluation see Fig. 4d.Full size imageBody and nest temperaturesFigure 6 shows a comparison of the dependence of body and nest temperatures on ambient air temperature and insolation between the two species. In the lower ranges of air temperature, usually at night, body temperature followed Tanest closely in both species. The exposition of the P. biglumis nests to the morning sun at ESE (Fig. 7) increased the wasp body temperature to values of often more than 15 °C higher than the surrounding air. However, body temperatures remained always below 40 °C (Fig. 6a). In P. gallicus, by contrast, the body temperature of the wasps increased considerably above 40 °C, to maximum values of about 46 °C, especially (but not exclusively) during intense insolation in the nest exposed to the morning sun (Fig. 6b).Figure 6Surface temperature of the thorax of adult wasps, of different stages of brood and of water drops of P. biglumis (left) and P. gallicus (right), in dependence on ambient air temperature close to the nest (Tanest) and global radiation (color scale). Egg f.n. = single egg on a foundress nest; diagonal lines = isolines. Regressions were calculated for shaded conditions (Radiation = 0–100 W/m2; black or gray solid lines) and sunshine (Radiation  > 100 W/m2; pink broken lines); P  More

Field collectionEndophytic fungi and insects were assessed from dormant twig samples from 155 tree species at 51 locations in 32 countries. Sampled tree species belonged to genera that are native to, and occur widely across, either the northern or southern hemisphere, since very few tree genera occur naturally in both hemispheres (e.g., in our study only Podocarpus appears in both hemispheres but has a limited distribution in the northern hemisphere). We sampled largely in botanical gardens and arboreta, which allowed us to sample native and non-native, congeneric and confamiliar, tree species at each location. At each location, one native and one to three non-native congeneric or confamiliar tree species were sampled.At each location, twenty 50-cm long asymptomatic twigs were collected from 1–5 individual trees per species, from different branches and different parts of the crown (Fig. 1). The number of individual trees per species depended on the number of trees available in the specific botanical garden or arboretum, which was often low (Table 1). All twigs per tree species and location were pooled and analysed as a single sample. On some occasions two samples of the same tree species at the same location are considered. Sampling was conducted in the month with the shortest day-length in the year (end of December 2017 in the Northern hemisphere, end of June 2018 in the Southern hemisphere). Samples originating from a tropical region (eleven samples from Tanzania) were collected in June 2018. Trees were sampled in winter to align with the timing of trade, i.e. most woody plants are traded in winter or early spring, as plants will be planted in the following spring, and to reduce the risk of introducing foliar pests in deciduous trees. Evergreen gymnosperm and angiosperm tree species, which were also considered, do not lose foliage during winter, and are thus sold with leaves/needles.Table 1 Site information for sampling locations included in this study.Full size tableFungal endophytesTo assess fungal communities, a total of 352 samples from 145 native and non-native tree species, belonging to nine families of angiosperms and gymnosperms, were collected. Sampling was done at 44 locations in 28 countries on five continents (Fig. 1, Table 1).From each twig in a sample, one bud, one needle/leaf and one 1 cm long twig segment were taken (Fig. 1). Needles from gymnosperms, and leaves from evergreen angiosperms were sampled to accurately assess the risk of trading these species. Twig segments were cut from the twig bases. The selected plant parts were surface sterilized by immersion in 75% ethanol for 1 min, 4% NaOCl for 5 min, and 75% ethanol for 30 s26. After air drying on a sterile bench, the following material from each of 20 twigs per sample was pooled: half of one bud, a 0.5 cm long piece of a needle (from gymnosperms) or a 0.25 cm2 leaf (for evergreen angiosperms) and a 0.5 cm long piece of twig.DNA extraction, PCR amplification and Illumina sequencingTotal genomic DNA was extracted from 50 mg of pooled, surface sterilized, and ground tissue (Fig. 1) using DNeasy PowerPlant Pro Kit (Qiagen, Hilden, Germany), following the manufacturer’s instructions. For a total of 31 out of 352 samples, DNA was extracted from different tissues separately, and DNA extracts were then pooled. DNA concentrations were quantified using the Qubit dsDNA BR Assay Kit (Thermo Fisher Scientific, Waltham, USA) on a Qubit 3.0 Fluorometer (Thermo Fisher Scientific) and DNA was diluted to 5 ng/μl. Samples that yielded less than 5 ng/μl were not diluted. The ITS2 region was amplified with the 5.8S-Fung and ITS4-Fung primers27. PCR amplifications were carried out in 20 μl reaction volumes containing 25 ng of DNA template, 1 mg/ml BSA, 1 mM of MgCl2, 0.4 μM of each primer, and 0.76 × JumpStart REDTaq ReadyMix Reaction Mix (Sigma-Aldrich, Steinheim, Germany). PCR was performed using Veriti 96-Well Thermal Cycler (Applied Biosystems, Foster City, CA, USA) as described in Franić et al. (2019). Each sample was amplified in triplicates and successful PCR amplification confirmed by visualization of the PCR products, before and after pooling the triplicates, on 1.5% (w/v) agarose gel with ethidium bromide staining. Pooled amplicons were sent to the Génome Québec Innovation Center at McGill University (Montréal, Quebec, Canada) for barcoding using Fluidigm Access Array technology (Fluidigm, South San Francisco, CA, USA) and paired-end sequencing on the Illumina MiSeq v3 platform (Illumina Inc., San Diego, CA, USA). Raw sequences obtained in this study are deposited at the NCBI Sequence Read Archive under BioProject accession number PRJNA70814822.Bioinformatics and taxonomical classification of ASVsQuality filtering and delineation into ASVs were done with a customized pipeline28 largely based on VSEARCH29, as described by Herzog et al.30. The output data available on Figshare show the abundances of fungal ASVs in the samples24. Taxonomic classification of ASVs was conducted using Sintax31 implemented in VSEARCH against the UNITE v.7.2 database32 with a bootstrap support of 80%. The data on the taxonomic classification of fungal ASVs is deposited in Figshare24.Quality filtering, delineation into ASVs, and taxonomical assignments were done on a larger data set (total of 474 samples), which increased the confidence in the selected centroid sequences. This data set consisted of (1) sequences obtained from 352 samples of pooled tree tissues that are presented here22, (2) sequences obtained from 33 samples of pooled tree tissues which were not included in this manuscript due to violation of the common protocol, (3) sequences from 21 contaminated samples (positive DNA extraction controls), including sequences from the two control samples (not presented here), and (4) sequences obtained from 66 samples of non-pooled tree tissues of Pinus sylvestris and Quercus robur that were collected from the subset of locations considered in this study, but for a different study, and are thus not presented here.Herbivorous insectsInsects were assessed from 227 samples of 109 tree species, collected at 31 locations and in 18 countries (Fig. 1, Table 1).The collected twigs (twenty 50 cm twigs per species per location) were brought to a laboratory close to each sampling location and inspected for the presence of insects that overwinter as adults. Twigs were kept at room temperature with the cut ends immersed in water to induce budding and to allow the development of insects that overwinter as larvae, pupae or eggs. Twigs from each sample were protected with gauze bags to prevent insects moving between samples (Fig. 1). Twigs were inspected for the presence of insects daily for 4 weeks and all collected insects were stored in 95% ethanol for further examination.Morphological and molecular identificationInsects were inspected using a stereo microscope and sorted to taxonomic orders and feeding guilds (i.e. herbivores, predators, parasitoids and other). The abundance of the different feeding guilds and taxonomic orders in the samples is presented in a file deposited on Figshare24. Herbivorous insects were further sorted into morphospecies and at least one specimen per morphospecies was stored at −20 °C for molecular analysis. The abundance of the different morphospecies in each sample is presented in a file deposited on Figshare24. Specimens for molecular analysis were photographed with a Leica DVM6 digital microscope and the Leica Application Suite X (LAS X). Depending on the size of the insects, the whole individual or parts (e.g. legs, head) were used for molecular analysis. Genomic DNA was extracted with a KingFisher (Thermo Fisher Scientific) extraction protocol suitable for insects (35 min incubation at RT, 30 min wash at RT with 3 different washing buffers, 13 min elution at 60 °C) in a 96-well plate. PCR for the COI was carried out in 25 µl reaction volume with 2 µl diluted DNA (1:10), 0.5 µM of each of the primers LCO1490 and HCO219833 and 1 x REDTaq ReadyMix Reaction Mix (Sigma-Aldrich) using a Veriti 96-Well Thermal Cycler (Applied Biosystems) with the following setting: 2 min at 94 °C, five cycles of 30 s at 94 °C, 40 s at 45 °C, and 1 min at 72 °C, 35 cycles of 30 s at 94 °C, 50 s at 51 °C, and 1 min at 72 °C, and a final extension step at 72 °C for 10 min. The success of amplification was verified by electrophoresis of the PCR products in 1.5% (w/v) agarose gel at 90 V for 30 min with ethidium bromide staining. A standard Sanger sequencing of the PCR products in both directions with the same primers was done at Macrogen Europe, Amsterdam, Netherlands. Sequences were assembled and edited with CLC Workbench (Version 7.6.2, Quiagen) and compared to reference sequences in BOLD34. If no conclusive results were found, sequences were compared to reference sequences in the National Centre for Biotechnology Information (NCBI) GenBank databases35. Specimens were assigned to species if the query sequence showed less than 1% divergence from the reference sequence. If two or more taxa matched within the same range, the assignment was ranked down to the next taxonomic level (i.e., genus). When no species match was obtained based on the above criteria, a genus was assigned with a divergence of less than 3%. For lower taxonomic groups the 100 nearest sequences were inspected on the Blast Tree (Fast Minimum Evolution Method) and the taxonomic relationship was evaluated based on that tree. If none of the approaches above revealed a conclusive taxonomic assignment, the morphological identification was taken as reference. The results of morphological and molecular identification of insect specimens are presented in a file deposited on Figshare24. Insect sequences are deposited in GenBank database under accession numbers MW441337-MW44176725.Sample metadataPairwise geographic distances (Euclidean distances) between sampling locations were calculated based on the geographic coordinates of the locations, with function “dist” in the R statistical programme36.Climate data, including mean annual temperature, mean annual precipitation, and temperature seasonality were obtained from the WorldClim database37, at a resolution of 2.5 min, and represent averages between 1970 and 2000.A host-tree phylogeny was constructed with the phylomatic function from the package brranching38 in R using the “zanne2014” reference tree39. One Eucalyptus sample collected in Argentina and two Eucalyptus samples collected in Tunisia were not identified to species. To place them in the phylogeny, we assigned them to different congeneric species that were not sampled in this study and that we considered as representative samples of phylogenetic diversity from across Eucalyptus genus (E. viminalis, E. robusta and E. radiata). Pairwise phylogenetic distances between study tree species were calculated using the “cophenetic” function in R36.The described sample metadata are available in a file on Figshare24. More

Pure coordination gameIn this section we study the Pure Coordination Game (PCG) (also known as doorway game, or driving game) in which (R=1), (S=0), (T=0), and (P=1), resulting in a symmetric payoff matrix with respect to the two strategies:$$begin{gathered} begin{array}{*{20}c} {} & {quad ; {text{A}}} &; {text{B}} \ end{array}hfill \ begin{array}{*{20}c} {text{A}} \ {text{B}} \ end{array} left( {begin{array}{*{20}c} 1 & 0 \ 0 & 1 \ end{array} } right) hfill \ end{gathered}$$
(2)
There are two equivalent equilibria for both players coordinating at the strategy A or B (a third Nash equilibrium exists for players using a mix strategy of 50% A and 50% B). As the absolute values of the payoff matrix are irrelevant and the dynamics is defined by ratios between payoffs from different strategies, the payoff matrix (2) represents all games for which the relation (R=P >S=T) is fulfilled.In the PCG the dilemma of choosing between safety and benefit does not exist, because there is no distinction between risk-dominant and payoff-dominant equilibrium. Both strategies yield equal payoffs when players coordinate on them and both have the same punishment (no payoff) when players fail to coordinate. Therefore, the PCG is the simplest framework to test when coordination is possible and which factors influence it and how. It is in every player’s interest to use the same strategy as others. Two strategies, however, are present in the system at the beginning of the simulation in equal amounts. From the symmetry of the game we can expect no difference in frequency of each strategy being played, when averaged over many realisations. Still, the problem of when the system reaches full coordination in one of the strategies is not trivial. We address this question here.Figure 1Time evolution of the coordination rate (alpha) (in MC steps) in individual realisations for different values of the degree k in a random regular network of (N=1000) nodes, using (a) the replicator dynamics, (b) the best response, and (c) the unconditional imitation update rule.Full size imageFigure 2Coordination rate (alpha) and interface density (rho) vs degree k of a random regular network for (N=1000) using (a) the replicator dynamics, (b) the best response, and (c) the unconditional imitation update rule. Each green circle represents one of 500 realisations for each value of the degree k and the average value is plotted with a solid line, separately for (alpha >0.5) and (alpha le 0.5). Results are compared to the ER random network ((alpha _{ER})) with the same average degree.Full size imageFirst, we look at single trajectories as presented in Fig. 1. Some of them quickly reach (alpha =0) or 1, or stop in a frozen state without obtaining global coordination. Other trajectories take much longer and extend beyond the time scale showed in the figure. What we can already tell is that the process of reaching coordination is slower in the replicator dynamics where it usually takes more time than in the best response and unconditional imitation to reach a frozen configuration. For all update rules the qualitative effect of the connectivity is similar—for bigger degree it is more likely to obtain full coordination and it happens faster. For the UI, however, larger values of degree than for the RD and BR are required to observe coordination. For example, in the case of (k=10) or 20 the system stops in a frozen disorder when using UI, while for the RD and BR it quickly reaches a coordinated state of (alpha =0) or 1.To confirm the conclusions from observation of trajectories, we present the average outcome of the system’s evolution in the Fig. 2. The first thing to notice is that all plots are symmetrical with respect to the horizontal line of (alpha = 0.5). It indicates that the strategies are indeed equivalent as expected. In all cases there is a minimal connectivity required to obtain global coordination. For the RD and BR update rules this minimum value is (k=4), although in the case of BR the system fails to coordinate for small odd values of k due to regular character of the graph. This oscillating behaviour does not exist in Erdős–Rényi random networks. When nodes choose their strategies following the UI rule much larger values of k are required to obtain full coordination. Single realisations can result in (alpha = 0), or 1 already for (k=15). However, even for (k=60) there is still a possibility of reaching a frozen uncoordinated configuration.The important conclusion is that there is no coordination without a sufficient level of connectivity. In order to confirm that this is not a mere artefact of the random regular graphs we compare our results with those obtained for Erdős–Rényi (ER) random networks76,77 (black dashed line in Fig. 2). The level of coordination starts to increase earlier for the three update rules, but the general trend is the same. The only qualitative difference can be found in the BR. The oscillating level of coordination disappears and it doesn’t matter if the degree is odd or even. This shows that different behaviour for odd values of k is due to topological traps in random regular graphs78. Our results for the UI update rule are also consistent with previous work reporting coordination for a complete graph but failure of global coordination in sparse networks40.Figure 3Examples of frozen configuration reached under the UI update rule for small values of the average degree k in random regular networks (top row) and Erdős–Rényi networks (bottom row) with 150 nodes. Red colour indicates a player choosing the strategy A, blue colour the strategy B. Note the topological differences between random regular and ER networks when they are sparse. For (k=1) a random regular graph consists of pairs of connected nodes, while an ER network has some slightly larger components and many loose nodes. For (k=2) a random regular graph is a chain (sometimes 2–4 separate chains), while an ER network has one large component and many disconnected nodes. For (k=3) and (k=4) a random regular graph is always composed of one component, while an ER network has still a few disconnected nodes.Full size imageSince agents using the RD and BR update rule do not achieve coordination for small values of degree, one might suspect that the network is just not sufficiently connected for these values of the degree, i.e. there are separate components. This is only partially true. In Fig. 3, we can see the structures generated by random regular graph and by ER random graph algorithms. Indeed, for (k=1) and 2 the topology is trivial and a large (infinite for (k=1)) average path length23 can be the underlying feature stopping the system to reach coordination. For (k=3), however, the network is well connected with one giant component and the system still does not reach the global coordination when using RD or BR. For the UI update rule coordination arrives even for larger values of k. Looking at the strategies used by players in Fig. 3 we can see how frozen configuration without coordination can be achieved. There are various types of topological traps where nodes with different strategies are connected, but none of them is willing to change the strategy in the given update rule.We next consider the question of how the two strategies are distributed in the situations in which full coordination is not reached. Looking at the trajectories in Fig. 1 we can see that there are only few successful strategy updates in such scenario and the value of (alpha) remains close to 0.5 until arriving at a frozen state for (k=2) (also (k=7) for UI). This suggests that there is not enough time, in the sense of the number of updates, to cluster the different strategies in the network. Therefore, one might expect that they are well mixed as at the end of each simulation. However, an analysis of the density of active links in the final state of the dynamics, presented in Fig. 2, shows a slightly more complex behaviour. When the two strategies are randomly distributed (i.e. well mixed) in a network, the interface density takes the value (rho =0.5). When the two strategies are spatially clustered in the network there are only few links connecting them and therefore the interface density takes small values. Looking at the dependence of (rho) on k, we find that for the replicator dynamics the active link density starts at 0.5 for (k=1), then drops below 0.2 for (k=2) and 3 indicating good clustering between strategies, to fall to zero for (k=4) where full coordination is already obtained. When using the best response update rule the situation is quite different. For (k=1) there are no active links, (rho =0), and hardly any for (k=2). There is a slight increase of the active link density for (k=3), to drop to zero again for (k=4) due to full coordination. Because of the oscillatory level of coordination there are still active links for odd values of (kP) (otherwise we can rename the strategies and shuffle the columns and rows). What defines the outcome of a game are the greater than and smaller than relations among the payoffs. Therefore we can add/subtract any value from all payoffs, or multiply them by a factor grater than zero, without changing the game. Thus, the payoff matrix (1) can be rewritten as:$$begin{gathered} begin{array}{*{20}c} {} & {qquad {text{A}}} & {quad quad {text{B}}} \ end{array} ;; hfill \ begin{array}{*{20}c} {text{A}} \ {text{B}} \ end{array} left( {begin{array}{*{20}c} 1 & {frac{{S – P}}{{R – P}}} \ {frac{{T – P}}{{R – P}}} & 0 \ end{array} } right) hfill \ end{gathered}$$
(3)
which, after substituting (S’=frac{S-P}{R-P}) and (T’=frac{T-P}{R-P}), is equivalent to the matrix: $$begin{gathered} begin{array}{*{20}c} {} &quad ;;{text{A}} &; {text{B}} \ end{array} ;quad quad quad quad quad quad begin{array}{*{20}c} {} & quad; {text{A}} & ;{text{B}} \ end{array} hfill \ begin{array}{*{20}c} {text{A}} \ {text{B}} \ end{array} left( {begin{array}{*{20}c} 1 & {S^{prime}} \ {T^{prime}} & 0 \ end{array} } right)xrightarrow[{{text{apostrophes}}}]{{{text{skipping}}}}begin{array}{*{20}c} {text{A}} \ {text{B}} \ end{array} left( {begin{array}{*{20}c} 1 & S \ T & 0 \ end{array} } right) hfill \ end{gathered}$$
(4)
From now on we omit the apostrophes and simply refer to parameters S and T. This payoff matrix can represent many games, including e.g. the prisoner’s dilemma14,46 (for (T >1) and (S More

IPCC Climate Change 2014: Synthesis Report. In Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (eds Core Writing Team, Pachauri, R. K. & Meyer L. A.) 151 (IPCC, Geneva, Switzerland, 2014).Grizzetti, B., Bouraoui, F. & Aloe, A. Changes of nitrogen and phosphorus loads to European seas. Glob. Change Biol. 18, 769–782 (2012).
Google Scholar 
Hoegh-Guldberg, O. & Bruno, J. F. The impact of climate change on the world’s marine ecosystems. Science 328, 1523–1528 (2010).CAS 
PubMed 

Google Scholar 
Duarte, C. M. Global change and the future ocean: a grand challenge for marine sciences. Front. Mar. Sci. 1, 1–16 (2014).
Google Scholar 
Richardson, A. J. & Schoeman, D. S. Climate impact on plankton ecosystems in the Northeast Atlantic. Science 305, 1609–1612 (2004).CAS 
PubMed 

Google Scholar 
Rose, J. M. et al. Effects of increased pCO2 and temperature on the North Atlantic spring bloom. II. Microzooplankton abundance and grazing. Mar. Ecol. Prog. Ser. 388, 27–40 (2009).CAS 

Google Scholar 
Sommer, U., Paul, C. & Moustaka-Gouni, M. Warming and ocean acidification effects on phytoplankton—from species shifts to size shifts within species in a mesocosm experiment. PLoS ONE 10, 1–17 (2015).
Google Scholar 
Garzke, J., Hansen, T., Ismar, S. M. H. & Sommer, U. Combined effects of ocean warming and acidification on copepod abundance, body size and fatty acid content. PLoS ONE 11, 1–22 (2016).
Google Scholar 
Horn, H. G., Boersma, M., Garzke, J., Sommer, U. & Aberle, N. High CO2 and warming affect microzooplankton food web dynamics in a Baltic Sea summer plankton community. Mar. Biol. 167, 1–17 (2020).
Google Scholar 
Boyd, P. W. et al. Experimental strategies to assess the biological ramifications of multiple drivers of global ocean change—a review. Glob. Change Biol. 24, 2239–2261 (2018).
Google Scholar 
Stewart, R. I. A. et al. Mesocosm experiments as a tool for ecological provided for ecological climate-change research. In Advances in Ecological Research/Guy Woodward (ed. O’Gorman, E. J.) 71–181 (Academic Press, 2013).Rost, B. & Riebesell, U. Coccolithophores and the biological pump: responses to environmental changes. In Coccolithophores: From Molecular Processes to Global Impact (eds Thierstein, H. R. & Young, J. R.) 99–125 (Springer, 2004).Peter, K. H. & Sommer, U. Phytoplankton cell size reduction in response to warming mediated by nutrient limitation. PLoS ONE 8, 1–6 (2013).
Google Scholar 
Bermúdez, J. R., Riebesell, U., Larsen, A. & Winder, M. Ocean acidification reduces transfer of essential biomolecules in a natural plankton community. Sci. Rep. 6, 1–8 (2016).
Google Scholar 
Peter, K. H. & Sommer, U. Interactive effect of warming, nitrogen and phosphorus limitation on phytoplankton cell size. Ecol. Evolution 5, 1011–1024 (2015).
Google Scholar 
Alvarez-Fernandez, S. et al. Plankton responses to ocean acidification: the role of nutrient limitation. Prog. Oceanogr. 165, 11–18 (2018).
Google Scholar 
Stramski, D., Sciandra, A. & Claustre, H. Effects of temperature, nitrogen, and light limitation on the optical properties of the marine diatom Thalassiosira pseudonana. Limnol. Oceanogr. 47, 392–403 (2002).CAS 

Google Scholar 
Marañón, E. Cell size as a key determinant of phytoplankton metabolism and community structure. Annu. Rev. Mar. Sci. 7, 241–264 (2015).
Google Scholar 
Peñuelas, J., Sardans, J., Rivas‐Ubach, A. & Janssens, I. A. The human-induced imbalance between C, N and P in Earth’s life system. Glob. Change Biol. 18, 3–6 (2011).
Google Scholar 
Azam, F. et al. The ecological role of water-column microbes in the sea. Mar. Ecol. Prog. Ser. 10, 257–63. (1983).
Google Scholar 
Legendre, L. & Le Fèvre, J. Microbial food webs and the export of biogenic carbon in oceans. Aquat. Microb. Ecol. 9, 69–77 (1995).
Google Scholar 
Beaufort, L. et al. Sensitivity of coccolithophores to carbonate chemistry and ocean acidification. Nature 476, 80–83 (2011).CAS 
PubMed 

Google Scholar 
Langer, G., Nehrke, G., Probert, I., Ly, J. & Ziveri, P. Strain-specific responses of Emiliania huxleyi to changing seawater carbonate chemistry. Biogeosciences 6, 2637–2646 (2009).CAS 

Google Scholar 
Winter, A., Henderiks, J., Beaufort, L., Rickaby, R. E. M. & Brown, C. W. Poleward expansion of the coccolithophore Emiliania huxleyi. J. Plankton Res. 36, 316–325 (2014).CAS 

Google Scholar 
Hopkins, J., Henson, S. A., Painter, S. C., Tyrrell, T. & Poulton, A. J. Phenological characteristics of global coccolithophore blooms. Glob. Biogeochemical Cycles 29, 239–253 (2015).CAS 

Google Scholar 
León, P. et al. Seasonal variability of the carbonate system and coccolithophore Emiliania huxleyi at a Scottish Coastal Observatory monitoring site. Estuar., Coast. Shelf Sci. 202, 302–314 (2018).
Google Scholar 
Rivero-Calle, S., Gnanadesikan, A., Del Castillo, C. E., Balch, W. M. & Guikema, S. D. Multidecadal increase in North Atlantic coccolithophores and the potential role of rising CO2. Science 350, 1533–1537 (2015).CAS 
PubMed 

Google Scholar 
Purdie, D. A. & Finch, M. S. Impact of a coccolithophorid bloom on dissolved carbon dioxide in sea water enclosures in a Norwegian fjord. Sarsia 79, 379–387 (1994).
Google Scholar 
Nejstgaard, J. C., Gismervik, I. & Solberg, P. T. Feeding and reproduction by Calanus finmarchicus, and microzooplankton grazing during mesocosm blooms of diatoms and the coccolithophore Emiliania huxleyi. Mar. Ecol. Prog. Ser. 147, 197–217 (1997).
Google Scholar 
Leblanc, K. et al. Distribution of calcifying and silicifying phytoplankton in relation to environmental and biogeochemical parameters during the late stages of the 2005 North East Atlantic Spring Bloom. Biogeosciences 6, 2155–2179 (2009).CAS 

Google Scholar 
Sett, S. et al. Temperature modulates coccolithophorid sensitivity of growth, photosynthesis and calcification to increasing seawater pCO2. PLoS ONE 9, e88308 (2014).PubMed 
PubMed Central 

Google Scholar 
Benner, I. et al. Emiliania huxleyi increases calcification but not expression of calcification-related genes in long-term exposure to elevated temperature and pCO2. Philos. Trans. R. Soc. B 368, 20130049 (2013).
Google Scholar 
Borchard, C., Borges, A. V., Händel, N. & Engel, A. Biogeochemical response of Emiliania huxleyi (PML B92/11) to elevated CO2 and temperature under phosphorous limitation: a chemostat study. J. Exp. Mar. Biol. Ecol. 410, 61–71 (2011).CAS 

Google Scholar 
Harrison, P. J. et al. Geographical distribution of red and green Noctiluca scintillans. Chin. J. Oceanol. Limnol. 29, 807–831 (2011).
Google Scholar 
Johns, D. G., Edwards, M., Greve, W. & SJohn, A. W. G. Increasing prevelance of the marine cladoceran Penilia avirostris (Dana, 1852) in the North Sea. Helgol. Mar. Res. 59, 215–218 (2005).
Google Scholar 
O’Connor, M. I. O., Piehler, M. F., Leech, D. M., Anton, A. & Bruno, J. F. Warming and resource availability shift food web structure and metabolism. PLoS Biol. 7, 1–6 (2009).
Google Scholar 
Cross, W. F., Hood, J. M., Benstead, J. P., Huryn, A. D. & Nelson, D. Interactions between temperature and nutrients across levels of ecological organization. Glob. change Biol. 21, 1025–1040 (2015).
Google Scholar 
Boersma, M. et al. Temperature driven changes in the diet preference of omnivorous copepods: no more meat when it’s hot? Ecol. Lett. 19, 45–53 (2016).PubMed 

Google Scholar 
Anderson, T. R., Hessen, D. O., Boersma, M., Urabe, J. & Mayor, D. J. Will invertebrates require increasingly carbon-rich food in a warming world? Am. Naturalist 190, 725–742 (2017).
Google Scholar 
Kirchner, M., Sahling, G., Uhlig, G., Gunkel, W. & Klings, K.-W. Does the red tide-forming dinoflagellate Noctiluca scintillans feed on bacteria? Sarsia 81, 45–55 (2015).
Google Scholar 
Elbrächter, M. & Qi, Y. Aspects of Noctiluca (Dinophyceae) population dynamics. In Physiological Ecology of Harmful Algal Blooms (ed. Anderson, M. D.) 315–335 (Springer-Verlag, 1998).Atienza, D., Saiz, E. & Calbet, A. Feeding ecology of the marine cladoceran Penilia avirostris: natural diet, prey selectivity and daily ration. Mar. Ecol. Prog. Ser. 315, 211–220 (2006).
Google Scholar 
Zhang, S., Liu, H., Chen, B. & Chih-Jung, W. Effects of diet nutritional quality on the growth and grazing of Noctiluca scintillans. Sci. Rep. 527, 73–85 (2015).CAS 

Google Scholar 
Reid, P. C., Borges, M. F. & Svendsen, E. A regime shift in the North Sea circa 1988 linked to changes in the North Sea horse mackerel fishery. Fish. Res. 50, 163–171 (2001).
Google Scholar 
Beaugrand, G., Brander, K. M., Lindley, J. A., Souissi, S. & Reid, P. C. Plankton effect on cod recruitment in the North Sea. Nature 426, 661–664 (2003).CAS 
PubMed 

Google Scholar 
Payne, M. R. et al. Recruitment in a changing environment: the 2000s North Sea herring recruitment failure. ICES J. Mar. Sci. 66, 272–277 (2009).
Google Scholar 
Perälä, T., Olsen, E. M. & Hutchings, J. A. Disentangling conditional effects of multiple regime shifts on Atlantic cod productivity. PLoS ONE 15, e0237414 (2020).PubMed 
PubMed Central 

Google Scholar 
Behrenfeld, M. J., Boss, E. S. & Halsey, K. H. Phytoplankton community structuring and succession in a competition-neutral resource landscape. ISME COMMUN. 1, 1–8 (2021).Monteiro, F. M. et al. Why marine phytoplankton calcify. Sci. Adv. 2, 1–14 (2016).
Google Scholar 
Mayers, K. M. J. et al. The possession of coccoliths fails to deter microzooplankton grazers. Front. Mar. Sci. 7, 976 (2020).
Google Scholar 
Zhao, Y. et al. Grazing by microzooplankton and copepods on the microbial food web in spring in the southern Yellow Sea, China. Mar. Life Sci. Technol. 2, 442–455 (2020).
Google Scholar 
Aberle, N. et al. High tolerance of microzooplankton to ocean acidification in an Arctic coastal plankton community. Biogeosciences 10, 1471–1481 (2013).
Google Scholar 
Horn, H. G. et al. Low CO2 sensitivity of Microzooplankton communities in the Gullmar Fjord, Skagerrak: evidence from a long-term Mesocosm Study. PLoS ONE 11, 1–24 (2016).
Google Scholar 
Chen, B., Landry, M. R., Huang, B. & Liu, H. Does warming enhance the effect of microzooplankton grazing on marine phytoplankton in the ocean? Limnol. Oceanogr. 57, 519–526 (2012).CAS 

Google Scholar 
Vázquez-Domínguez, E., Vaqué, D. & Gasol, J. M. Temperature effects on the heterotrophic bacteria, heterotrophic nanoflagellates, and microbial top predators of the NW Mediterranean. Aquat. Microb. Ecol. 67, 107–121 (2012).
Google Scholar 
Lara, E. et al. Experimental evaluation of the warming effect on viral, bacterial and protistan communities in two contrasting Arctic systems. Aquat. Microb. Ecol. 70, 17–32 (2013).
Google Scholar 
Olson, M. B., Solem, K. & Love, B. Microzooplankton grazing responds to simulated ocean acidification indirectly through changes in prey cellular characteristics. Mar. Ecol. Prog. Ser. 604, 83–97 (2018).CAS 

Google Scholar 
Sherr, E. B. & Sherr, B. F. Bacterivory and herbivory: key roles of phagotrophic protists in pelagic food webs. Microb. Ecol. 28, 223–235 (1994).CAS 
PubMed 

Google Scholar 
Brander, K. & Kiørboe, T. Decreasing phytoplankton size adversely affects ocean food chains. Glob. Change Biol. 26, 5356–5357 (2020).
Google Scholar 
Irigoien, X. et al. A high frequency time series at weathership M, Norwegian Sea, during the 1997 spring bloom: feeding of adult female Calanus finmarchicus. Mar. Ecol. Prog. Ser. 172, 127–137 (1998).
Google Scholar 
Fenchel, T. The microbial loop—25 years later. J. Exp. Mar. Biol. Ecol. 366, 99–103 (2008).
Google Scholar 
Aberle, N., Malzahn, A. M., Lewandowska, A. M. & Sommer, U. Some like it hot: the protozooplankton— copepod link in a warming ocean. Mar. Ecol. Prog. Ser. 519, 103–113 (2015).
Google Scholar 
Berglund, J., Müren, U., Båmstedt, U. & Andersson, A. Efficiency of a phytoplankton-based and a bacteria-based food web in a pelagic marine system. Limnol. Oceanogr. 52, 121–131 (2007).CAS 

Google Scholar 
Sherr, E. B. & Sherr, B. F. Heterotrophic dinoflagellates: a significant component of microzooplankton biomass and major grazers of diatoms in the sea. Mar. Ecol. Prog. Ser. 352, 187–197 (2007).
Google Scholar 
Gifford, D. J. The protozoan-metazoan trophic link in pelagic ecosystems. J. Protozool. 38, 81–86 (1991).
Google Scholar 
Rollwagen-Bollens, G. & Gifford, S. The role of protistan microzooplankton in the upper San Francisco estuary planktonic food web: source or sink? Estuaries Coasts 34, 1026–1038 (2011).CAS 

Google Scholar 
Anjusha, A. et al. Trophic efficiency of plankton food webs: observations from the Gulf of Mannar and the Palk Bay, Southeast Coast of India. J. Mar. Syst. 115, 40–61 (2013).
Google Scholar 
IPCC. Global Warming of 1.5 °C. An IPCC Special Report on the impacts of global warming of 1.5 °C above pre-industrial levels and related global greenhouse gas emission pathways. In The Context of Strengthening the Global Response to The Threat of Climate Change, Sustainable Development, and Efforts to Eradicate Poverty (Masson-Delmotte, V. et al (eds.) 616 (IPCC, Geneva, Switzerland, 2018).Pansch, A., Winde, V., Asmus, R. & Asmus, H. Tidal benthic mesocosms simulating future climate change scenarios in the field of marine ecology. Limnol. Oceanogr.: Methods 14, 257–267 (2016).
Google Scholar 
van Leeuwen, S., Tett, P., Mills, D. & van der Molen, J. Stratified and nonstratified areas in the North Sea: long-term variability and biological and policy implications. J. Geophys. Res.: Oceans 120, 4670–4686 (2015).
Google Scholar 
Grasshoff, K., Kremling, K. & Ehrhardt, M. (eds). Methods of Seawater Analysis, 3rd edn. (Wiley-VCH, Weinheim, 1999).Dickson, A. G. An exact definition of total alkalinity and a procedure for the estimation of alkalinity and total inorganic carbon from titration data. Deep-Sea Res. 28, 609–623 (1981).CAS 

Google Scholar 
Pierrot, D. E., Lewis, E. & Wallace, D. W. R. MS Excel program developed for CO2 system calculations. ORNL/CDIAC-105a. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tennessee https://doi.org/10.3334/CDIAC/otg.CO2SYS_XLS_CDIAC105a (2006).Dickson, A. G. & Millero, F. J. A comparison of the equilibrium constants for the dissociation of carbonic acid in seawater media. Deep-Sea Res. 34, 1733–1743 (1987).CAS 

Google Scholar 
Arrigo, K. R. et al. Phytoplankton community structure and the drawdown of nutrients and CO2 in the Southern Ocean. Science 283, 365–368 (1999).CAS 
PubMed 

Google Scholar 
Utermöhl, H. Zur Vervollkommnung der quantitativen Phytoplankton- Methodik. Int. Ver. für. Theoretische und Angew. Limnologie: Mitteilungen 9, 1–38 (1958).
Google Scholar 
McEwen, G. F., Johnson, M. W. & Folsom, T. R. A statistical analysis of the performance of the Folsom plankton sample splitter, based upon test observations. Archiv für. Archiv Meteorologie, Geophysik und Bioklimatologie, Ser. A 7, 502–527 (1954).
Google Scholar 
Sell, D. W. & Evans, M. S. A statistical analysis of subsampling and an evaluation of the Folsom plankton splitter. Hydrobiologia 94, 223–230 (1982).
Google Scholar 
Boersma, M., Wiltshire, K. H., Kong, S., Greve, W. & Renz, J. Long-term change in the copepod community in the southern German Bight. J. Sea Res. 101, 41–50 (2015).
Google Scholar 
Marie, D., Simon, N. & Vaulot, D. Phytoplankton cell counting by flow cytometry. Algal Culturing Tech. 1, 253–267 (2005).
Google Scholar 
Hillebrand, H., Dürselen, C., Kirschtel, D., Pollingher, U. & Zohary, T. Biovolume calculation for pelagic and benthic microalgae. J. Phycol. 35, 403–424 (1999).
Google Scholar 
Menden-Deuer, S. & Lessard, E. J. Carbon to volume relationships for dinoflagellates, diatoms, and other protist plankton. Limnol. Oceanogr. 45, 569–579 (2000).CAS 

Google Scholar 
Putt, M. & Stoecker, D. K. An experimentally determined carbon: volume ratio for marine “oligotrichous” ciliates from estuarine and coastal waters. Limnol. Oceanogr. 34, 1097–1103 (1989).
Google Scholar 
Beran, A. et al. Carbon content and biovolume of the heterotrophic dinoflagellate Noctiluca scintillans from the Northern Adriatic Sea. In Proceedings of the CESUM-BS 2003, Varna. 28 (Book of Abstracts, UNESCO, Paris, 2003).Lee, S. & Fuhrman, J. A. Relationships between biovolume and biomass of naturally derived marine bacterioplankton. Appl. Environ. Microbiol. 53, 1298–1303 (1987).CAS 
PubMed 
PubMed Central 

Google Scholar 
Kraberg, A., Baumann, M. & Dürselen, C. Coastal Phytoplankton: Photo Guide for Northern European Seas (Dr. Friedrich Pfeil, München, 2010).R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2021). More

We used publicly available, daily, county-level COVID-19 cases and deaths from the Pennsylvania Department of Health (PA DOH) (https://www.health.pa.gov/topics/disease/coronavirus/pages/cases.aspx)13,14 for Centre County and the six neighboring counties with which it shares borders: Blair, Clearfield, Clinton, Huntingdon, Mifflin, and Union (Table 1, Fig. 1). Official COVID-19 reporting for these counties began on March 1, 2020 and is ongoing.Table 1 Summary statistics. COVID-19 reporting, census data, SafeGraph mobile-device derived data.Full size tableFigure 1(a) The cumulative COVID-19 case trajectory for Centre County minus the student cases (red line) has the same shape as the outbreak for the neighboring counties. When looking at student cases only (blue line), the curve leads other counties. Centre County cumulative cases including the university (purple line) take on the shape of an early increase because of the student cases. (b) When aggregating cases from students and non-students, Centre County (purple dot) reported about the number of cases expected for its population size, relative to the neighboring counties (black dots). When the university-reported student cases are separated from the non-student residents of the county, cases reported in Centre County non-students (red dots show possible range of total cases) fall below the number of cases we would expect for the population size. Student cases only (blue dot) are high for the student population size.Full size imageWithin Centre County, PSU provided COVID-19 testing for UP students from August 7, 2020 onward and reported anonymized weekly (2020) and daily (2021) confirmed cases, negative test results, and total tests completed for each campus in a public dashboard (Figs. 1a, S1) (https://virusinfo.psu.edu/covid-19-dashboard/)8. Two types of testing were conducted: students who were enrolled in on-campus classes were randomly selected for surveillance testing and all students could use on-demand testing. Through March 23, 2021, a total of 45,092 random tests were conducted for surveillance, of which 462, or 1.0%, were infected. Surveillance testing efforts ranged from 2440 to 4020 weekly tests through the Fall 2020 semester and were designed to consistently test approximately 1% of students throughout the school year.During the same time period, 75,436 on-demand tests were conducted, of which 6093, or 8.1%, were infected. Students living in both on-campus dorms and off-campus apartments had equal access to university-provided testing. Both on-campus and off-campus residences are within Centre County so positive and negative tests results were also included in the overall Centre County reports of COVID-19 cases.Pre-arrival testing was required for students returning to campus from transmission hotspots. Students with positive tests from pre-arrival testing were required to isolate for 10–14 days after their positive test before arriving on campus. Results from pre-arrival testing for students returning to campus in the Fall of 2020 are not included in these data.At the county level, PA DOH reports the total positive, probable, and negative tests for each county. Because PSU is within Centre County, we estimated the number of total positive and negative tests for non-student Centre County residents by subtracting the PSU estimates (from the PSU dashboard) from the Centre County estimates provided by PA DOH. However, not all student tests were reported to DOH. A portion of the on-demand tests conducted for PSU UP students were completed by a third-party vendor, which required student registration. At the time of student registration, an estimated 0–25% of students registered with an address for a family home that did not reflect their residence in Centre County. Their test results were reported to the county of their registered address. This impacts a maximum of 1,166 positive student test results and 10,760 negative student tests.We conducted a sensitivity analysis to assess the uncertainty in reporting around the negative and positive students tests that may have been misallocated due to the reported residence of student tests. We have calculated the minimum and maximum number of affected positive and negative student tests. This uncertainty from student tests impacts non-student values, which are calculated by subtracting student values from county level reports. The calculations are based on a range of a possible 0–1166 positive student tests misallocated to other counties and up to 10,760 misallocated negative student tests. We have used the ranges of misallocated student tests to calculate, for non-student Centre County residents, the full possible range of (1) total cases, (2) reported cases per capita, and (3) tests per capita (Table 1, Fig. 1b). As a result, our estimates of cases and per capita testing among non-student residents in Centre County are imprecise (Table 1).We also used publicly available data from PA DOH data and PSU to calculate COVID-19 deaths per 100,000 for Centre County, the six neighboring counties, and PSU UP.We acquired county-level data on median household income, population size, and college enrollment status from the 2019 United States Census Bureau’s American Community Survey (ACS) 5-year data (https://www.census.gov/data/developers/data-sets/acs-5year.html) for all seven previously mentioned counties in central PA15.We divide the census block groups (CBG) of Centre County into two categories. We first designated ‘student-dominated CBGs’ as CBGs where  > 50% of ACS responses report enrollment as undergraduate students. We consider data from the 19 student-dominated CBGs in Centre County to be representative of the student population in Centre County. In addition to off-campus locations, the 19 student-dominated CBGs include all on-campus dorms. These 19 CBGs are either on or adjacent to PSU’s UP campus and occupy exactly 6 census tracts. The remaining 25 county census tracts were designated as non-student dominated areas.SafeGraph16 receives geolocation data from anonymized mobile devices collected from numerous applications. We analyzed SafeGraph’s mobile device-derived daily visit counts to points of interest (POI), which are fixed locations, such as businesses or attractions. SafeGraph data provide daily counts for total numbers of visits by mobile devices while using at least one application that provides geolocation data to SafeGraph. A “visit” indicates that the device entered the building or spatial perimeter designated as a POI. We acquired daily visit counts for POIs in the seven previously mentioned counties in central PA from January 1, 2019 forward (Table 1) and within Centre County grouped counts into student-dominated CBGs and non-student dominated CBGs. From January 1, 2020 forward, we used SafeGraph data on the median daily minutes that devices spent outside of their home in each county and the student- and non-student dominated CBG divisions in Centre County. The “home location” of each device is defined by its location overnight. Finally, we used SafeGraph’s weekly calculated number of devices residing in each county and the CBGs of Centre County for 2019 to measure SafeGraph’s data representation across the seven counties and the CBGs of Centre County.No administrative permissions were required to obtain these data. Academic researchers can register to receive access to SafeGraph data at no charge for non-commercial purposes only. See Data Availability statement below for details. More

Knapp, A. K. & Smith, M. D. Variation among biomes in temporal dynamics of aboveground primary production. Science 291, 481–484 (2001).CAS 
PubMed 
ADS 

Google Scholar 
Collins, S. L. et al. Stability of tallgrass prairie during a 19-year increase in growing season precipitation. Funct. Ecol. 26, 1450–1459 (2012).
Google Scholar 
Maurer, G. E., Hallmark, A. J., Brown, R. F., Sala, O. E. & Collins, S. L. Sensitivity of primary production to precipitation across the United States. Ecol. Lett. 23, 527–536 (2020).PubMed 

Google Scholar 
IPCC. IPCC. (Cambridge University Press, 2013) https://doi.org/10.1017/cbo9781107415324.Knapp, A. K. et al. Differential sensitivity to regional-scale drought in six central US grasslands. Oecologia 177, 949–957 (2015).PubMed 
ADS 

Google Scholar 
Smith, M. D. An ecological perspective on extreme climatic events: A synthetic definition and framework to guide future research. J. Ecol. 99, 656–663 (2011).
Google Scholar 
Zeppel, M. J. B., Wilks, J. V. & Lewis, J. D. Impacts of extreme precipitation and seasonal changes in precipitation on plants. Biogeosciences 11, 3083–3093 (2014).ADS 

Google Scholar 
Frank, D. A. Drought effects on above- and belowground production of a grazed temperate grassland ecosystem. Oecologia 152, 131–139 (2007).PubMed 
ADS 

Google Scholar 
Skinner, R. H., Hanson, J. D., Hutchinson, G. L. & Schuman, G. E. Response of C3 and C4 grasses to supplemental summer precipitation. J. Range Manag. 55, 517–522 (2002).
Google Scholar 
Shi, Z. et al. Dual mechanisms regulate ecosystem stability under decade-long warming and hay harvest. Nat. Commun. 7, 1–6 (2016).ADS 

Google Scholar 
Zavaleta, E. S. et al. Grassland responses to three years of elevated temperature, CO2, precipitation, and N deposition. Ecol. Monogr. 73, 585–604 (2003).
Google Scholar 
Prather, R. M., Castillioni, K., Welti, E. A. R., Kaspari, M. & Souza, L. Abiotic factors and plant biomass, not plant diversity, strongly shape grassland arthropods under drought conditions. Ecology 101, 1–7 (2020).
Google Scholar 
Nippert, J. B., Knapp, A. K. & Briggs, J. M. Intra-annual rainfall variability and grassland productivity: Can the past predict the future?. Plant Ecol. 184, 65–74 (2006).
Google Scholar 
La Pierre, K. J. et al. Explaining temporal variation in above-ground productivity in a mesic grassland: The role of climate and flowering. J. Ecol. 99, 1250–1262 (2011).
Google Scholar 
Cleland, E. E. et al. Sensitivity of grassland plant community composition to spatial vs. temporal variation in precipitation. Ecology 94, 1687–1696 (2013).PubMed 

Google Scholar 
Grant, K., Kreyling, J., Heilmeier, H., Beierkuhnlein, C. & Jentsch, A. Extreme weather events and plant–plant interactions: Shifts between competition and facilitation among grassland species in the face of drought and heavy rainfall. Ecol. Res. 29, 991–1001 (2014).
Google Scholar 
Brooker, R. W. et al. Facilitation in plant communities: The past, the present, and the future. J. Ecol. 96, 18–34 (2008).MathSciNet 

Google Scholar 
Schöb, C., Armas, C. & Pugnaire, F. I. Direct and indirect interactions co-determine species composition in nurse plant systems. Oikos 122, 1371–1379 (2013).
Google Scholar 
Gross, N., Börger, L., Duncan, R. P. & Hulme, P. E. Functional differences between alien and native species: Do biotic interactions determine the functional structure of highly invaded grasslands?. Funct. Ecol. 27, 1262–1272 (2013).
Google Scholar 
van der Merwe, S., Greve, M., Olivier, B. & le Roux, P. C. Testing the role of functional trait expression in plant–plant facilitation. Funct. Ecol. https://doi.org/10.1111/1365-2435.13681 (2020).Article 

Google Scholar 
Tremmel, D. C. & Bazzaz, F. A. How neighbor canopy architecture affects target plant performance. Ecology 74, 2114–2124 (1993).
Google Scholar 
Weiher, E. & Keddy, P. A. In Ecological Assembly Rules: Perspective, Advances, Retreats. (eds. Weiher, E. & Keddy, P. A.) (Cambridge University Press, 2001).Anten, N. P. R. & Hirose, T. Interspecific differences in above-ground growth patterns result in spatial and temporal partitioning of light among species in a tall-grass meadow. J. Ecol. 87, 583–597 (1999).
Google Scholar 
Yann, H., Pascal, A. & Niklaus, A. H. Competition for light causes plant. Science 324, 636–638 (2009).
Google Scholar 
Walker, B., Kinzig, A. & Langridge, J. Plant attribute diversity, resilience, and ecosystem function: The nature and significance of dominant and minor species. Ecosystems 2, 95–113 (1999).
Google Scholar 
Brooker, R. W. Plant–plant interactions and environmental change. New Phytol. 171, 271–284 (2006).PubMed 

Google Scholar 
Michalet, R. & Pugnaire, F. I. Facilitation in communities: Underlying mechanisms, community and ecosystem implications. Funct. Ecol. 30, 3–9 (2016).
Google Scholar 
Maestre, F. T., Callaway, R. M., Valladares, F. & Lortie, C. J. Refining the stress-gradient hypothesis for competition and facilitation in plant communities. J. Ecol. 97, 199–205 (2009).
Google Scholar 
Saccone, P., Delzon, S., Jean-Philippe, P., Brun, J. J. & Michalet, R. The role of biotic interactions in altering tree seedling responses to an extreme climatic event. J. Veg. Sci. 20, 403–414 (2009).
Google Scholar 
Smith, M. D., Knapp, A. K. & Collins, S. L. A framework for assessing ecosystem dynamics in response to chronic resource alterations induced by global change. Ecology 90, 3279–3289 (2009).PubMed 

Google Scholar 
Borer, E. T., Seabloom, E. W., Gruner, D. S., Harpole, W. S. & Hillebrand, H. Herbivores and nutrients control grassland plant diversity via light limitation. Nature 508, 517–520 (2014).CAS 
PubMed 
PubMed Central 
ADS 

Google Scholar 
de Sassi, C. & Tylianakis, J. M. Climate change disproportionately increases herbivore over plant or parasitoid biomass. PLoS One 7, e40557 (2012).PubMed 
PubMed Central 

Google Scholar 
Strauss, S. Y. & Ivalú Cacho, N. Nowhere to run, nowhere to hide: The importance of enemies and apparency in adaptation to harsh soil environments. Am. Nat. 182, E1 (2013).PubMed 

Google Scholar 
Brady, K. U., Kruckeberg, A. R. & Bradshaw, H. D. Evolutionary ecology of plant adaptation to serpentine soils. Annu. Rev. Ecol. Evol. Syst. 36, 243–266 (2005).
Google Scholar 
Moran, M. S. et al. Soil evaporation response to Lehmann lovegrass (Eragrostis lehmanniana) invasion in a semiarid watershed. Agric. For. Meteorol. 149, 2133–2142 (2009).ADS 

Google Scholar 
Pérez-Harguindeguy, N. et al. New handbook for standardised measurement of plant functional traits worldwide. Aust. J. Bot. 61, 167–234 (2013).
Google Scholar 
Díaz, S. et al. The global spectrum of plant form and function. Nature 529, 167–171 (2016).PubMed 
ADS 

Google Scholar 
Gross, N., Suding, K. N. & Lavorel, S. Leaf dry matter content and lateral spread predict response to land use change for six subalpine grassland species. J. Veg. Sci. 18, 289–300 (2007).
Google Scholar 
Quiroga, R., Golluscio, R., Blanco, L. & Fernandez, R. Aridity and grazing as convergent selective forces: An experiment with an Arid Chaco bunchgrass. Ecol. Appl. https://doi.org/10.1890/09-0641 (2010).Article 
PubMed 

Google Scholar 
Blumenthal, D. M. et al. Traits link drought resistance with herbivore defence and plant economics in semi-arid grasslands: The central roles of phenology and leaf dry matter content. J. Ecol. 108, 2336–2351 (2020).
Google Scholar 
Taylor, S. H. et al. Ecophysiological traits in C3 and C4 grasses: A phylogenetically controlled screening experiment. New Phytol. 185, 780–791 (2010).CAS 
PubMed 

Google Scholar 
N’Guessan, M. & Hartnett, D. C. Differential responses to defoliation frequency in little bluestem (Schizachyrium scoparium) in tallgrass prairie: Implications for herbivory tolerance and avoidance. Plant Ecol. 212, 1275–1285 (2011).
Google Scholar 
Castillioni, K. et al. Drought mildly reduces plant dominance in a temperate prairie ecosystem across years. Ecol. Evol. 10, 6702–6713 (2020).PubMed 
PubMed Central 

Google Scholar 
Ivalú Cacho, N. & Strauss, S. Y. Occupation of bare habitats, an evolutionary precursor to soil specialization in plants. Proc. Natl. Acad. Sci. U. S. A. 111, 15132–15137 (2014).PubMed 
PubMed Central 
ADS 

Google Scholar 
Cottingham, K. L., Lennon, J. T. & Brown, B. L. Knowing when to draw the line: Designing more informative ecological experiments. Front. Ecol. Environ. 3, 145–152 (2005).
Google Scholar 
Xu, X., Sherry, R. A., Niu, S., Li, D. & Luo, Y. Net primary productivity and rain-use efficiency as affected by warming, altered precipitation, and clipping in a mixed-grass prairie. Glob. Change Biol. 19, 2753–2764 (2013).ADS 

Google Scholar 
Braun-Blanquet, J. Plant Sociology: The Study of Plant Communities. (1932).Shipley, B. The AIC model selection method applied to path analytic models compared using ad-separation test. Ecology 94, 560–564 (2013).PubMed 

Google Scholar 
Lefcheck, J. S. piecewiseSEM: Piecewise structural equation modelling in r for ecology, evolution, and systematics. Methods Ecol. Evol. 7, 573–579 (2016).
Google Scholar 
Grace, J. B. In Structural Equation Modeling and Natural Systems. (Cambridge University Press, 2006). https://doi.org/10.1017/CBO9780511617799.Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D. & Team, R.C. nlme: Linear and nonlinear mixed effects models. R package version 3.1 111 (2013).Bates, D., Mächler, M., Bolker, B. M. & Walker, S. C. Fitting linear mixed-effects models using lme4. Ecol. Austral 67, 1–48 (2015).
Google Scholar 
Pearson, D. E., Ortega, Y. K. & Maron, J. L. The tortoise and the hare: reducing resource availability shifts competitive balance between plant species. J. Ecol. 105, 999–1009 (2017).CAS 

Google Scholar 
Maron, J. L. & Crone, E. Herbivory: Effects on plant abundance, distribution and population growth. Proc. R. Soc. B Biol. Sci. 273, 2575–2584 (2006).
Google Scholar 
Bertness, M. & Callaway, R. M. Positive interactions in communities. Trends Ecol. Evol. 9, 191–193 (1994).CAS 
PubMed 

Google Scholar 
Ploughe, L. W. et al. Community Response to Extreme Drought (CRED): A framework for drought-induced shifts in plant–plant interactions. New Phytol. 222, 52–69 (2019).PubMed 

Google Scholar 
Klanderud, K., Vandvik, V. & Goldberg, D. The importance of Biotic vs. Abiotic drivers of local plant community composition along regional bioclimatic gradients. PLoS One 10, 1–15 (2015).
Google Scholar 
Maricle, B. R., Caudle, K. L. & Adler, P. B. Influence of water availability on photosynthesis, water potential, leaf δ 13 C, and phenology in dominant C 4 grasses in Kansas, USA. Trans. Kans. Acad. Sci. 118, 173–193 (2015).
Google Scholar 
Collins, S. L., Knapp, A. K., Briggs, J. M., Blair, J. M. & Steinauer, E. M. Modulation of diversity by grazing and mowing in native tallgrass prairie. Science 280, 745–747 (1998).CAS 
PubMed 
ADS 

Google Scholar 
Gornish, E. S. & Tylianakis, J. Community shifts under climate change: Mechanisms at multiple scales. Am. J. Bot. 100, 1422–1434 (2013).PubMed 

Google Scholar  More

Download PDF

I study the Chinese giant salamander (Andrias davidianus), which is native to the Yangtze River Basin of central China. This particular species is critically endangered in the wild owing to habitat loss and overcatching — a particular problem is their use in traditional Chinese medicine. My research focuses on the salamander’s conservation biology and evolutionary ecology.In this photo, I am releasing a Chinese giant salamander at the Golden Whip River in Zhangjiajie National Forest Park on an early morning in September 2021. My team and I caught the salamander the night before, to measure its size and collect tissue samples for genetic analyses.My interest in aquatic animals started as a child. I grew up in a rural village in Hunan province, and I remember spending most of my childhood playing and fishing near my home. Because of this, I knew where each fish species lived in nearby rivers and lakes, and it sparked my interest in river ecology.I’m employed as an associate professor at Jishou University, where I lead a team dedicated to researching this species of salamander. Wild salamanders are quiet, nocturnal animals that live in remote areas. This makes studying them challenging. My team tried many creative ways to track down the animals, including walking along riverbanks with torches and photographing salamanders under water — but these techniques didn’t work as well as we needed them to. We eventually found that the best way to trap wild salamanders is to use small live fish and chicken livers as bait. The research is challenging, but we’ve learnt to be patient and celebrate every small success we have.Studying Chinese giant salamanders has also taught me an important life lesson: adapt to thrive. When food is abundant, the salamanders grow rapidly; when food is scarce, they can go up to 11 months without feeding. In my personal life and work, I have experienced successes and failures, and taking on that lesson has been useful.

Nature 603, 194 (2022)
doi: https://doi.org/10.1038/d41586-022-00564-y

Related Articles

Close-up with a parasite that can blind

Handling snakes for science

Broaden your scientific audience with video animation

Managing up: how to communicate effectively with your PhD adviser

Subjects

Careers

Conservation biology

Ecology

Latest on:

Careers

Smaller science company? Tailor your CV for a manager, not HR
Career Column 25 FEB 22

Female scientists in Africa are changing the face of their continent
Editorial 22 FEB 22

African scientists engage with the public to tackle local challenges
Career Feature 15 FEB 22

Ecology

How colonialism fed the flames of Australia’s catastrophic wildfires
Research Highlight 24 FEB 22

Apply Singapore Index on Cities’ Biodiversity at scale
Correspondence 22 FEB 22

Marching in the streets for climate-crisis action
Career Q&A 22 FEB 22

Jobs

POST-DOC POSITION IN ELECTROPHYSIOLOGY OF FUNGAL NETWORKS

VU University Amsterdam
Amsterdam, Netherlands

Director, Division of Receipt and Referral Center for Scientific Review National Institutes of Health (NIH) Department of Health and Human Services (DHHS)

National Institutes of Health (NIH)
Bethesda, MD, United States

Postdoctoral Fellow

NIH National Heart, Lung, and Blood Institute (NHLBI)
Bethesda, MD, United States

Two postdoctorial researchers in structural biology, with focus on artificial intelligence

University of Gothenburg (GU)
Uppsala, Sweden More

The full-length sequences of PacBio SMRT sequencingBased on PacBio SMRT sequencing, 3,751,089, 3,434,452, 3,900,180, 8,535,019, and 4,435,846 subreads were generated for root, stem, leaf, flower, and seed, with a N50 of 3040, 3367, 2611, 2198, and 4584 bp, respectively (Table S1; Fig. S1). Subreads were processed to generate circular consensus sequences (CCSs). By detecting the primers and poly(A) tail, 238,196, 232,290, 211,535, 257,905, and 231,877 full-length non-chimeric (FLNC) reads were identified for root, stem, leaf, flower, and seed, with a mean length of 2633, 3070, 2561, 1746, and 3762 bp, respectively (Table S2; Fig. S2). After Iterative Clustering for Error Correction (ICE) clustering, polishing, base correction, de-redundancy, and non-plant sequences filtering, 37,789, 34,034, 38,100, 54,937, and 53,906 unigenes were retained for root, stem, leaf, flower, and seed, respectively, with an average unigene length of 1802–3786 bp and N50 of 2238–4707 bp (Table S2). The length of most unigenes from five organs exceeded 2000 bp, accounting for 68.88% of the total number (Table S3; Fig. 1A). Based on Benchmarking Universal Single-Copy Orthologs (BUSCO) assessment, about 88.1% (single-copy: 353; duplicated: 916) of the 1440 core embryophyte genes were found to be complete (90.6% were present when counting fragmented genes), suggesting the high integrity of the M. micrantha transcriptome (Fig. S3).Figure 1Length distribution of unigenes from PacBio SMRT sequencing (A) and Illumina RNA-Seq (B) across five organs.Full size imageDe novo assembly of Illumina RNA-Seq dataBased on Illumina RNA-Seq, 43.23, 40.27, 41.01, 65.85, and 41.09 million clean reads were obtained for root, stem, leaf, flower, and seed, respectively, with Q20 exceeding 96.72%. Using Trinity software, clean reads were de novo assembled into 124,238, 60,232, 63,370, 93,229, and 66,411 unigenes for root, stem, leaf, flower, and seed. After filtering non-plant sequences, 124,233, 60,232, 63,370, 93,228, and 66,410 unigenes were finally retained for the five organs, respectively (Table S4). The length of most unigenes (84.70%) was shorter than 2000 bp (Table S3). In addition, the average length and N50 of unigenes generated by Illumina RNA-Seq were 1067–1312 bp and 1336–1685 bp, respectively, which were shorter than that from PacBio SMRT sequencing (Table S4; Fig. 1B).Functional annotationTo obtain a comprehensive functional annotation of M. micrantha transcriptome, unigenes generated by PacBio SMRT sequencing were annotated in seven public databases, including NCBI non-redundant nucleotide sequences (NT), NCBI non-redundant protein sequences (NR), Gene Ontology (GO), Eukaryotic Orthologous Groups (KOG), Kyoto Encyclopedia of Genes and Genomes (KEGG), Swiss-Prot, and Pfam protein families. For root, stem, leaf, flower, and seed, 35,714 (94.51%), 32,614 (95.83%), 36,134 (94.84%), 49,197 (89.55%), and 50,962 (94.54%) unigenes were annotated to at least one database, respectively, suggesting that our transcriptome is well annotated and that most of unigenes may be functional (Table 1).Table 1 Statistics of annotation of full-length transcripts from five M. micrantha organs in seven databases.Full size tableBased on NR database annotation, the top three homologous species for the five organs were Cynara cardunculus, Vitis vinifera, and Daucus carota (Fig. S4). The top homologous species was a plant of the Asteraceae family. For the GO function annotation, “binding”, “catalytic activities”, “metabolic process”, “cellular process”, “cell”, and “cell part” were functional categories with the most abundant unigenes (Fig. S5). In addition, numerous unigenes were assigned to “response to stimulus”, “response to biotic stimulus”, and “response to oxidative stress” category (Table S5). Positive response to stress stimuli is an important strategy for invasive plants to adapt to the environment. In the KEGG annotation, the top two pathways with the most abundant unigenes were “carbohydrate metabolism” and “translation”. Furthermore, “energy metabolism” and “environmental adaptation” were also worthy of attention, which are important pathways responsible for energy supply and stress responses (Fig. S6).TFs identification and AS analysisUsing the iTAK pipeline, 1776 (root), 1293 (stem), 1627 (leaf), 2529 (flower), and 1733 (seed) unigenes were identified as TFs, which were classified into 68 families (Table S6). C3H (884), C2H2 (525), and bHLH (501) were the most abundant TF families (Fig. S7A). In addition, MYB (333) TFs were also found in the five organs. The differential expression levels of the top 15 TF families were further characterized. We found that the top 15 TF families had a certain amount of expression in the five organs of M. micrantha (Fig. S7B).For root, stem, leaf, flower, and seed, 3300, 2324, 3219, 4730, and 3740 unique transcript models (UniTransModels) were constructed, among which the UniTransModels containing two isoforms were the most common (Fig. S8A). There were 329, 270, 358, 336, and 537 AS events identified in root, stem, leaf, flower, and seed, respectively. Retained introns (RIs) were detected as the most abundant AS event in all five organs, followed by alternative 3′ splice sites (A3) and alternative 5′ splice sites (A5). Mutually exclusive exons (MX) were the least frequent event (Fig. S8B).Gene expression analysisThe number of unigenes in different expression level intervals was similar across the five organs (Fig. 2A). Using FPKM  > 0.3 as the threshold for unigene expression, the total number of unigenes expressed in the five organs was 102,464 (Fig. 2B). Among them, 39,227 unigenes were co-expressed in all five organs. The information of differentially expressed genes (DEGs) identified in pairwise comparisons among the five organs is listed in Table S7. In total, 21,161 DEGs were identified among the five organs (Fig. S9). The number of DEGs between the five organs varied from 3469 (root vs stem) to 10,716 (leaf vs seed) (Fig. 2C). Notably, 933, 428, 1410, 1018, and 1292 DEGs showed significant higher expression in root, stem, leaf, flower, and seed, respectively (Figs. S10 and S11).Figure 2Gene expression patterns in five M. micrantha organs. (A) The FPKM interval distribution in the five organs. (B) Venn diagram of the number of unigenes expressed in five organs. (C) Number of differentially expressed genes in each pairwise comparison of five organs.Full size imageKEGG enrichment of unigenes with higher expression in each organAccording to the KEGG enrichment analysis results, there were obvious differences in enriched pathways in the five organs (Table S8; Fig. 3). The unigenes with higher expression in root were mainly enriched to defense response and protein processing pathways, such as “plant-pathogen interaction” and “protein processing in endoplasmic reticulum”. In stem, unigenes with higher expression were predominantly enriched to pathways related to the secondary metabolite, sugar, and terpenoid biosynthesis, such as “phenylpropanoid biosynthesis”, “starch and sucrose metabolism”, and “diterpenoid biosynthesis”. In flower, unigenes with higher expression were mainly related to “starch and sucrose metabolism”, “phenylpropanoid biosynthesis”, and “cutin, suberine, and wax biosynthesis”. The unigenes with higher expression in seed were mainly enriched in three fatty acid and sugar metabolism pathways, namely “biosynthesis of unsaturated fatty acids”, “galactose metabolism”, and “amino sugar and nucleotide sugar metabolism”. The unigenes with higher expression in leaf were significantly enriched in photosynthesis pathways, including “photosynthesis-antenna proteins”, “photosynthesis”, “porphyrin and chlorophyll metabolism”, and “carbon fixation in photosynthetic organisms”, which are important for the photosynthesis of M. micrantha.Figure 3The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of unigenes with higher expression in each organ. The significantly enriched pathways with corrected p-value (q value)  More