Drivers of urban biodiversity in Mexico and joint risks from future urban expansion, climate change, and urban heat island effect

worst (WRT) scenarios [27]. This difference represents the projected shift in urbanization in each municipality. We estimated the urban heat island effect (UHI) following [68]. Karl et al. [68] developed a methodology to detect the effect of heat island intensity based on the population size of cities in the United States. Specifically, they derived a regression equation where the urban-to-rural temperature differences were estimated using the urban population. We used the same equa- tion but with population data for Mexican cities from a recent demography census (2020) from the Instituto Nacional de Estadı´stica y Geografı´a INEGI (https://www.inegi.org.mx). The UHI is a simple approximation based on empirical relationships of the form: UHI ¼ a ∗ UrbanTotalpopulationb The two fixed parameters, a and b, were calibrated for annual temperatures for US cities with a population exceeding the 100,000 inhabitants (see Table 5 in [68]). We used b = 0.45 and a central estimate of a = 1.74x10−3. Although many studies quantify UHI based on urban- to-rural temperature differences [69, 70], this is difficult to implement in many countries where meteorological stations are scarce and have a lot of missing data. However, empirical methods, as we use here, were developed to estimate the effect of urbanization on annual mean temperatures, removing the effects of the increasing population (i.e., urbanization) across large scales [68, 71, 72]. We standardized the projected shifts in temperature, urbanization, and current UHI

between 0 and 1 and summed these in a single metric of multivariate risk. This multivariate risk metric allows us to highlight which municipalities will be more exposed to a combination of impacts from urban expansion, climate change, and UHI in the future (Table D in S1 Data). We evaluate how regional clusters of Mexican cities are related to this multivariate risk metric and discuss how this can potentially affect urban biodiversity. However, we caution here that increases in urban populations are likely related to additional risks beyond UHI, including poor air quality, noise pollution, increase in artificial lights, waste disposal problems, and reduction in green urban spaces. Results SAR models We found strong linearity between species richness and urban area across municipalities with log-log transformation (Fig 1). In addition, we found notable differences in SAR shapes between urban and rural environments, with different intercepts but relatively similar slopes (Fig 1). We found that the slope for amphibians was similar between rural and urban areas, and intercept larger in rural areas. However, amphibians are a poorly sampled group in Mexico and likely are underrepresented in urban settings. The SAR model fitting approach revealed that urban species richness and urban area exhibit a non-linear relationship across municipalities, and convex-up, linear, convex-down, or sigmoidal SAR shapes were recovered across the four taxa (Table 2). Random forest models The RF models showed a relatively good predictive capacity although a relatively low explana-

tory power for urban species richness of terrestrial vertebrates (Fig E in S1 Appendix). The estimates of the effect sizes revealed that urban area, size of the network of urban protected areas, the number of ecoregions, and GDP were the most important positive determinants of urban total species richness (Fig 2a). By contrast, the percentage of people working in the PLOS ONE Urban biodiversity and anthropogenic pressures PLOS ONE | https://doi.org/10.1371/journal.pone.0308522 October 4, 2024 7 / 19