Effect of climate and geography on worldwide fine resolution economic activity

Data The underlying meteorological data are provided by the ERA-Interim reanalysis. This reanaly- sis is produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) and is described in [23]. Here we use data from 1979 to 2016, i.e. most of the ERA-Interim available period. Its main features are a horizontal resolution of 0.75˚ by 0.75˚, and a temporal resolu- tion which varies between 3 hours and 6 hours, depending on the variable (see Table 1). Also, several derivatives of the meteorological variables considered, and listed in Table 1, are used. The globally gridded GCP data used here comes from the Global Gridded Geographically Based Economic Data (G-Econ, https://gecon.yale.edu/data-and-documentation-g-econ- project), Version 4 [21] (see also an early version, for 1990 only, in [6]). This dataset contains derived one-degree grid cells of GCP data for the years 1990, 1995, 2000 and 2005. To allow for an easier direct comparison across the globe, we consider GCP at Purchasing Power Parity and per capita. Per capita GCP (hereafter referred to as GCP-PC) is computed by dividing GCP by the population in each cell, also provided with the G-Econ dataset. While the main use of the G-Econ is the gridded GCP data, this dataset also includes geo- graphical parameters such as area of grid cells, distance to coast, elevation, vegetation and soil types, distance to rivers, which are also used in our analysis, along with the climate variables (see Table 2). All geography parameters are fixed for the period covered by this study–possible

local variations particularly in vegetation or soil types, e.g. around urban centres, do not appear to be critical and they vary on a timescale slower than those of climate variables. To ensure geographical overlap with G-Econ, ERA-Interim is retrieved at 1˚, namely at a slightly lower resolution than its original 0.75˚. Assumptions and methodology The main focus of this work is the investigation of the casual relationship between the main features of meteorological variables over few decades and the corresponding GCP, rather than the (concurrent) correlation between meteorological variables and GCP. These two objectives require different approaches. Specifically, in the first case statistical properties such as seasonal variations of meteorological variables are used. It is this type of features, namely the changes in variables, that we want to analyse in addition to the more standard statistics such as the mean (of e.g. temperature). Accordingly, the main assumption here is that the statistics of both the Table 1. Meteorological variables used in this study, as derived from the ERA-Interim reanalysis. The following statistics have been computed for all variables: Mean, 1st Quartile (bottom Q), Median, 3rd Quartile (top Q), Standard Deviation of the original time series (SD), Standard Deviation of monthly means (SD S, representing intra-annual vari- ations) and temporal variations (representing short-term ‘gradients’). The latter are computed according to daily or

6-hourly steps (see column Step). For 6-hourly variables, increments are increased by 10% at each subsequent steps (out to 5 steps, i.e. 30 hours) and by 15% for daily steps (out to 5 days). For air temperature (at 2 m height), also daily excursions are calculated using two additional variables, Tmin and Tmax, available at 6-hour intervals. Climate Variable Unit ‘gradient’ Step Mean Sea Level Pressure (MSLP) hPa 750 6-hr Wind Speed at 10m (UV10) m s-1 2.5 6-hr Air Temperature (T2) ˚C 5 Daily Air Temperature daily excursion (DT) ˚C 2.5 Daily Dew Point Temperature (D2) ˚C 7 Daily Precipitation (TP) mm 5 Daily Relative Humidity (RH) % 20 Daily Solar Radiation (SR) W m-2 100 Daily Sunshine Duration (SUND) hr 3 Daily https://doi.org/10.1371/journal.pone.0229243.t001 PLOS ONE Effect of geography on fine resolution economy PLOS ONE | https://doi.org/10.1371/journal.pone.0229243 March 2, 2020 4 / 20