Performance evaluation and comparison of observed and reanalysis gridded precipitation datasets over Pakistan

The present study evaluates five observed gridded precipitation datasets [Global Precipitation Climatology Centre (GPCC), Climate Prediction Centre (CPC), Climatic Research Unit (CRU), Cressman Interpolated High-resolution Gauge-based Gridded Observations (CIHGGO), and ERA-Interim data Merged and Bias-corrected for ISIMIP (EWEMBI)] and five reanalysis products [ERA-Interim and ERA5 of European Centre for Medium-Range Weather Forecasts, Twentieth Century Reanalysis (20CR), Japanese 55-year Reanalysis (JRA55), and the Modern-Era Retrospective Analysis for Research and Applications version 2 (MERRA2)] against surface precipitation gauge data as reference for the period 1981–2015 over Pakistan. The performance of the above gridded datasets is assessed using six statistical metrics: correlation coefficient, relative bias, root mean square error, mean absolute error, wet/dry years, and precipitation centroid on monthly, seasonal [summer (June–September), and winter (December–March)], and annual timescales. Our results show that GPCC has overall much better performance across an entire country (avg. correlation > 0.95) in terms of all timescales and statistical metrics used. EWEMBI1 displays comparable results to GPCC with higher correlation and lower error values and thus can be ranked as the second-best performing observed gridded precipitation dataset. On the other hand, reanalysis products are found relatively weak in approximating the spatiotemporal distribution of precipitation, especially over complex northern areas of Pakistan. However, ERA5 exhibits a comparatively good positive linear relationship with surface precipitation gauge data at monthly (0.92), seasonal [0.89 (summer) to 0.98 (winter)], and annual (0.87) timescales, which may be attributed to an advanced data assimilation technique and model dynamics employed in the generation of the data.