GoKawiilThis study integrates daily gauge observations, CMIP6 model outputs and socioeconomic data to evaluate global precipitation monitoring gaps and identify global priority regions for new gauge deployment. The methodology consists of four main components: (1) compiling and harmonizing a multi-source precipitation gauge dataset to map current global observational coverage; (2) constructing a physiography-based reference map and quantifying network deficiencies; (3) linking historical gauge placement priorities to existing gauge density and precipitation spatial variability; and (4) projecting future priority areas under various climate and socioeconomic scenarios using a target-based simulation framework. To ensure the robustness of model outputs and main conclusions, we evaluate the results through sensitivity analysis using multiple spatial windows and validate the priority sites for new gauges.
Input data
Precipitation gauge network
We compiled a global precipitation gauge network comprising 245,368 stations from 15 daily gauge-based precipitation datasets (Supplementary Table 1; for details, see Supplementary Text 1). Among these, 14,505 gauges were relocated several times. To ensure data accuracy and eliminate duplication, we consolidated and deduplicated stations based on location (0.001°), resulting in a final dataset of 221,483 unique stations. A subset of 38,203 LR stations was identified for a duration exceeding 30 years and having less than 10% missing data over the entire observed record1. Although precipitation data were collected from 1850 onwards, our analysis mainly focused on the spatial and temporal distributions of all stations and LR stations during the period 1900–2022 because of high data gaps in the nineteenth century. Station densities were calculated on a 1° × 1° grid following the WMO standards (Supplementary Table 3).
CMIP6 model data
We obtained monthly climate variables from 13 Coupled Model Intercomparison Project Phase 6 (CMIP6) models (Supplementary Table 4) because they provided output with the necessary variables for the historical period (1850–2014) and future projections (2015–2100) under various scenarios. Five key variables include: precipitation flux (pr), near-surface (2 m) air temperature (tas), water evapotranspiration flux (evspsbl), relative humidity (hur) and total emission rate of black carbon aerosol mass (emibc). We used historical simulations for the period 1970–1999 and future projections for 2025–2100 under two scenarios: a low-emission pathway (SSP1-2.6) and a high-emission (SSP5-8.5) pathway. All variables were uniformly reprocessed to a standardized 1° × 1° grid using bilinear interpolation.
Future economic and population data
We used global gridded population data (person per km2) with a 30-arc-second resolution from ref. 81, covering 2025–2100 under the Sustainability pathway (SSP1) and Fossil fuelled development (SSP5) scenarios. Global GDP data (in 2005 US dollars, purchasing power parity) from ref. 82 were similarly gridded and time-matched. Both datasets were aggregated into a 1° × 1° grid by average.
Mapping physiographical regions
The global physiographic map (Supplementary Fig. 3) includes seven regions: Interior Plains (labelled as Plains in figures), Hilly/Undulating, Mountains, Coastal Zones (Coastal), Small Islands (Islands), Urban Areas (Urban) and Polar/Arid regions. We used the 2015 Global Ecological Land Units database83 (GELU) with a 250-m resolution for most regions, supplemented by the Global, Self-consistent, Hierarchical, High-resolution Shoreline database84 (GSHHS) and the Global Islands Database85 (GID). We then aggregated the original resolution of the physiographic map to 1° × 1° by maintaining the dominant class within each grid cell. With special consideration given to the Urban region, where the grids containing more than 2.5% urban region were classified as belonging to the Urban region (details in Supplementary Text 2 and Supplementary Fig. 21). This resulted in 15,390 grid cells, excluding Antarctica, which formed the basis for our analyses.
... continue reading