MethodsStudy area: Our initial study area included the entire globe. We began with a seamless grid of cells with a resolution of 0.5 degrees (i.e., ~50 km at the equator). Next, we created polylines representing coastlines using SRTM (Shuttle Radar Topographic Mission) v4.1 global digital elevation model data at a resolution of 250 m (Reuter et al. 2007). We used these coastline polylines to identify and retain cells that intersected the coast. We excluded 192,227 cells that did not intersect the coast. To avoid cells with minimal potential coastal wetland habitat, we used the coastline data to remove an additional 1,056 coastal cells that contained less than or equal to 5% coverage of land. We also removed 176 cells which did not have suitable climate data; most of these cells were removed because they either did not have minimum air temperature data or they had unrealistic low or high minimum air temperature data relative to their neighboring cells. Collectively, these steps produced a grid (hereafter, study grid) that contained a total of 4,908 cells at a resolution of 0.5 degrees. Biogeographic zone and range limit assignmentsFor biogeographic zone and range limit-specific analyses, we assigned various identification codes to each study grid cell. Biogeographic zone assignments included either Atlantic East Pacific (AEP) or Indo West Pacific (IWP) (sensu Duke et al. 1998). Range limits, defined as areas where mangroves abruptly become absent from coastlines, were assigned individually using a combination of climate data, mangrove presence data, and descriptions in the literature. We conducted a literature review to develop hypotheses regarding the climatic and non-climatic factors that control each range limit (Table 1). We created polygons for 14 focal range limits (Fig. 2), and used these polygons to assign study grid cells to a particular range limit. All range limits spanned a mangrove presence-absence transition. For range limits that were expected to be controlled, at least in part, by winter temperatures, we created polygons that spanned the cold-to-hot transition zone. Where possible, this zone extended from a minimum temperature of -20 °C (cold) up to a maximum temperature of 20 °C (hot). However, due to various constraints, most of these transitions covered smaller temperature gradients. For range limits that were expected to be controlled, at least in part, by precipitation, we created polygons that spanned the wet-to-dry transition zone, as determined via the mean annual precipitation data.Climate dataPrior studies in North America have identified the importance of using air temperature extremes in mangrove distribution and abundance models (Osland et al. 2013, Cavanaugh et al. 2014). For all cells within the study grid, we sought to identify the absolute coldest daily air temperature that occurred across a recent multi-decadal period. Although monthly-based mean minimum air temperature data are readily available, daily minimum air temperature data have historically been more difficult to obtain at the global scale (Donat et al. 2013). Due to the absence of a consistent and seamless global dataset of daily air temperature minima, we used a combination of three different gridded daily minimum air temperature data sources. For cells in the United States, we used 2.5-arcminute resolution data created by the PRISM Climate Group (Oregon State University; http://prism.oregonstate.edu) (Daly et al. 2008), for the period extending from 1981-2010. For all continental cells outside of the United States (i.e., coastal cell connected to large bodies of land on all continents except for the United States), we used 1-degree resolution data created by Sheffield et al. (2006), for the same time period. For most islands, we used 0.5-degree resolution data created by Maurer et al. (2009), for the period extending from 1971-2000. From these three data sources, we created a minimum temperature (MINT) data set for the study grid cells to represent the absolute coldest air temperature that occurred across a recent three to four decade period, depending upon the source. For each study grid cell, we also obtained 30-second resolution mean annual precipitation (MAP) data from the WorldClim Global Climate Data (Hijmans et al. 2005), for the period extending from 1950-2000. We also obtained 5-arcminute resolution global gridded mean annual sea surface temperature data from a dataset produced by UNEP-WCMC (2015), for the period extending from 2009-2013. In addition to the gridded climate data, we obtained station-based air temperature data. For 13 of the 14 focal range limits, we identified a proximate station with a long-term record of daily air temperatures. For each of these stations, we obtained daily minimum air temperature data for the 30-year period extending from 1981-2010. From these data, we calculated: (1) the absolute coldest temperature during the 30-year record (MINT); (2) the annual minimum temperature (i.e., the coldest temperature of each year); and (3) annual mean winter minimum temperature (i.e., the mean of the daily minima for the coldest quarter of each year). Mangrove dataTo determine mangrove presence, we used two global mangrove distribution data sources (Spalding et al. 2010, Giri et al. 2011), and assigned a binary code to each study grid cell denoting presence or absence. For most of the world, mangrove presence was assigned to a cell only when both of these sources deemed that mangroves were present. For Myanmar, however, the two mangrove distribution sources were not in agreement, and the Giri et al. (2011) data were deemed more reliable and used to assign mangrove presence for those cells. The two sources were also not in agreement for the coasts of Gabon, Congo, and the Cabinda Province of Angola, and the Spalding et al. (2011) data were deemed more reliable and used to assign mangrove presence for those cells. To determine mangrove species richness within each cell, we used data produced by Polidoro et al. (2010). For each cell where mangroves were deemed to be present, we used the sum of the species-specific mangrove distributional range data to determine the total number of mangrove species potentially present within a cell. To determine mangrove abundance within each cell, we used the 30-m resolution global mangrove distribution data produced by Giri et al. (2011).