Climate controls the amount of energy available for plants, which in turn determines the quantity of resources available for animals. It follows that when climate changes, so should trophic communities. Using a novel modeling approach, we investigate how bird and mammal trophic communities might disassemble and reassemble under 21^st century climate changes. We show that trophic structures are expected to undergo profound changes globally, chiefly in the tropics and across high latitudes in the northern hemisphere. This trophic reorganization of communities is characterized by shifts in species richness within trophic guilds. While some guilds might face population collapses, others are projected to find new opportunities to maintain stable populations in previously inhospitable areas. The proposed models offer a tool for projecting and understanding the trophic ramifications of climate change, highlighting their potential in guiding future research and conservation efforts.

Data

Geographical data were garnered from two primary sources and subsequently plotted on a global terrestrial grid, each cell measuring 1 × 1°: The global distribution ranges of terrestrial mammals and non-marine birds; Predictor bioclimatic variables potentially pertinent to these species. The distributions of species, specifically 9,993 non-marine birds and 5,272 terrestrial mammals, totalling 15,265 species, were informed by the IUCN Global Assessment's data on native ranges(38). To enable analysis, a presence/absence matrix was created. In this matrix, the species were aligned as columns, each named, against 18,418 terrestrial grid cells, each cell measuring 1 × 1°, as rows. These include all the non-coastal cells of the world, excluding Antarctica and some northern regions, such as most of Greenland, for which some data are lacking. This approach provided a clear, granular view of species distribution across the globe.

The bioclimatic variables were divided into two datasets: historical (1961-2018) and future (2021-2100). Historical bioclimatic variables were not obtained directly but derived from three monthly meteorological variables: mean minimum temperature (°C), mean maximum temperature (°C), and total precipitation (mm). These variables were downscaled from CRU-TS-4.03(39) with WorldClim 2.1(40) for bias correction. The nineteen WorldClim variables were calculated from these three-monthly meteorological variables using the "biovars" function of the R dismo package(41). Unlike the historical data, pre-processed bioclimatic variables into the future could be accessed directly. We used a multimodel ensemble approach, which tends to perform better than any individual model(42, 43). The ensemble integrates mean outputs from 25 global climate models (GCMs) corresponding to an array of twelve different future climate change scenarios(39, 40). These scenarios emerge from the interplay of four specific timeframes (2021-2040, 2041-2060, 2061-2080, and 2081-2100) and three Shared Socio-economic Pathways (ssp2-45, ssp3-70, and ssp5-85). Each pathway provides a speculative lens into the potential shifts in global socio-economic landscapes, demographics, and economic structures in the forthcoming century(17).

Lastly, the feeding habits of bird and mammal species were obtained from the global species-level compilation of key trophic attributes, known as Elton traits 1.0(44).

Step 1: Obtaining the trophic profile of the cells.

We assigned each of the 15,265 terrestrial mammal and non-marine bird species to one of 9 trophic guilds and then counted the number of species in each guild within each cell, following a previous analysis(5). The result is a matrix with the 9 trophic guilds as columns, 14,520 cells as rows, and values representing numbers of species. The community trophic structure of every cell is thus a point in a 9-dimensional trophic space defined by the number of species from each trophic guild (a vector of dimension 9).

Step 2: Identifying community trophic structures.

From the initial set of 14,520 terrestrial grid cells, each measuring 1°×1°, a specific subset of 6,610 continental cells was selected. The selection criterion was the presence of some form of protected area. This subset was crucial for two analytical steps: first, to decipher the community trophic structures; and second, to model the interaction between the prevailing climate and the trophic structure. Given the nature of these cells — designated as "continental protected area cells" — we assume they experience reduced human activity compared to the surrounding matrix; an assumption that may not align with reality globally, considering evidence of reduced effectiveness of protected areas in ensuring tangible protection in various parts of the tropics(45).  Nevertheless, a working assumption is made that the trophic structures displayed within these areas likely present a closer reflection of what might be expected from an undisturbed, stable energy network(5). Additionally, to ensure the integrity and consistency of the analysis, we employed AMD analysis(5). This was performed to examine whether the six foundational trophic structures identified in previous analyses (ranging from TS1 to TS6) were indeed manifesting in this curated subset of cells.

Step 3: Climate modelling of community trophic structures.

We modelled the relationship between climate and trophic structures, utilizing 19 predictors derived from historical bioclimatic data encompassing the years 1961-1990. Denoted as pre-1990 period, this phase marks a time before the significant uptick in temperatures attributable to human-induced greenhouse gas emissions. The trophic profile data, systematically assembled from faunal lists gathered over numerous decades, also hail from an era prior to this pronounced temperature increase. Therefore, these records present a fitting basis for examining the interplay between the trophic structure and the climatic conditions prevalent during the pre-1990 period.

We used the Extreme Gradient Boosting algorithm (Xgboost)(46), using the package xgboost(47). Xgboost to predict trophic structures in response to climate variables. This is a recently developed technique of machine learning that has demonstrated superior performance compared to random forests. In our analysis, the target variable was the trophic structure, identified in the second step and spanning from TS1 to TS6, applicable to the 6610 grid cells. By projecting the outcomes of the Xgboost algorithm onto the 14503 1°×1° terrestrial grid cells, as visualized in Figure 1, we were able to infer the anticipated trophic structure based on their specific climate conditions. This inference remains insulated from the influences of more contemporary climate shifts, given that the algorithm's training was grounded in pre-1990 data. Additionally, the model is free from the myriad impacts of human activity, due to the exclusive use of continental protected area cells in our analysis.

Step 4: Inferring the expected processes of local extinction and colonization in course.

Projecting the Xgboost algorithm—trained on data from the pre-1990 era—onto post-1990 climatic conditions enable us to infer potential shifts from one trophic structure to another. These shifts can have important implications for understanding future species range trajectories within communities. For instance, if the ensuing trophic structure supports a reduced count of species for a particular guild, some of these species will potentially be lost from the region. As such, alterations in the trophic structure could lead to diminished guild species carrying capacities for specific trophic guilds, culminating in the local extinction of certain species. Conversely, if the evolved structure accommodates a greater number of species for some guilds, the species-carrying capacity for those guilds will likely augment. In these circumstances, it is anticipated that previously absent species from such guilds might establish stable local populations.

Correlating transitions between different trophic structures (like TS6 transitioning to TS5), we can quantify potential variations in the species-carrying capacity for each guild. This is achieved by contrasting the mean species count of the original structure with that of the projected one.

When the Xgboost algorithm, which is rooted in pre-1990 data, is superimposed on the latest available decade (2009-2018)—subsequently termed as the recent past climate data (illustrated in Fig.1b)—it helps in understanding the dynamic of trophic structures. For example, changes from TS6 to TS5 enable us to approximate, for each cell, the shifts anticipated in the species-carrying capacity of each trophic guild during that period. Given the anticipated delay between shifts in climate and community trophic reorganization, these projections are also likely indicative of current or imminent local extinction and colonization processes. The collective alteration (either a decrease or an increase) in the species-carrying capacity across all guilds equates to the total change in species-carrying capacity. Furthermore, aggregating the projected changes in species-carrying capacity across all grid cells provides an estimate, for a stipulated time frame, of the number of species-cell units affected post-1990. This metric can also be delineated for individual trophic guilds or comprehensively for all combined.

Step 5: Inferring local extinction and colonization processes expected for this century.

The Xgboost algorithm was further applied to various climate change predictions spanning the 21st century (specifically, 2021-2040, 2041-2060, 2061-2080, and 2081-2100). The resulting projection was intended to discern potential shifts in community trophic structures, contingent on the Shared Socioeconomic Pathway (SSP) the global community adopts(17). Initially, three emission scenarios were contemplated: SSP2-45, SSP3-70, and SSP5-85(48). The SSP2-45 scenario aligns closely with the objectives of the Paris Agreement, seeking to restrain global temperature rise to beneath 2°C, while simultaneously making efforts to cap the increase at 1.5°C. However, the growing consensus among scientists(49), deems the realization of this target under current policy trajectories as increasingly unlikely. Thus, our analysis used the more moderate SSP3-70 scenario and the more extreme SSP5-85 scenario, so to span a wide range of uncertainties. Findings related to the SSP2-45 scenario are detailed in the supplementary section. By assessing deviations in the trophic structure relative to pre-1990 data, we were able to approximate changes in species-carrying capacity—both in terms of decline and growth—for individual trophic guilds, as well as cumulatively. The collective projected decrease in species-carrying capacity of guilds across all grid cells allowed us to estimate the anticipated number of species-cell units impacted since 1990, after each time window. This was evaluated both at the level of individual trophic guilds and in a comprehensive manner, encompassing all guilds.