2.2. Precipitation data
As a proxy for water availability in the habitat, we used the mean precipitation rate at the time and site in which the ant-plant interactions were sampled in each study (Konapala et al. 2020). This metric allowed us to estimate the precipitation rate experienced by the plants with EFNs and the ants when the interactions were observed, leading to more refined results about the influence of the local climatic conditions on the patterns of ant-plant interactions. We obtained this information from TRMM_3B43 dataset (TRMM from now on), available at the NASA Giovanni web database, from NASA’S EOSDIS (NASA’S Earth Observing System Data Information System) (TRMM, 2001). In this database, the precipitation rate is measured as the amount of rainfall per area per unit of time (TRMM, 2001). TRMM compiles global climatic information for the period between 01/01/1998 to 31/12/2019, covering the period when most of the empirical networks in our database were collected (N = 21).
To retrieve the precipitation data, we extracted from each paper the period and the geographical coordinates of the study site in which each network was sampled. This information was available in most of the papers included in our database. In two studies, the authors did not report the period or site in which the interactions were observed. Then, we excluded these studies from our dataset, which reduced our dataset to 19 studies. Among all networks described in these papers, only one network reported interactions collected over a period not covered by TRMM dataset (before 1998). In this case, we excluded this single network from our dataset but included the other two networks described in this study and collected after 1998 (Díaz-Castelazo et al. 2013).
In studies where the ant-plant interactions were sampled in an area equal to or inferior to one minute latitudinal and longitudinal scale, we used a one-minute scale as our standard spatial unit on TRMM (N = 43 networks, 68.25%). Only two studies sampled the ant-plant networks in an area larger than a one-minute scale (N = 20 networks, 31.75%). In these cases, we selected an area proportional to one minute at the center of the sampling area reported in the original paper. After setting the area, we informed the period of interest and extracted the precipitation information as a “time series of area-averaged values”. By doing that, we could obtain the monthly averaged mean precipitation rate in millimeters for the period in which each network included in our dataset was sampled.