Discrepancies with satellite observations in the spatial structure of global precipitation as derived from global climate models. (5th November 2018)

• Record Type: Journal Article
• Authors: Tapiador, Francisco J.; Navarro, Andrés; Jiménez, Alfonso; Moreno, Raúl; García‐Ortega, Eduardo
• Is Part Of: Quarterly journal of the Royal Meteorological Society. Volume 144(2018)Supplement 1
• Page Start: 419
• Page End: 435
• Publication Date: 2018-11-05
• Subjects: climatologies -- model validation -- precipitation -- remote sensing; Meteorology -- Periodicals; 551.5
• DOI: 10.1002/qj.3289
• Languages: English
• ISSNs: 0035-9009

Abstract

Abstract : One of the applications of satellite‐derived precipitation datasets, such as Global Precipitation Climatology Project (GPCP) or Climate Prediction Centre Merged Analysis of Precipitation (CMAP), is to validate output from numerical models, either numerical weather prediction (NWP) models, Regional Climate Models (RCMs), Global Climate Models (GCMs) or Earth System Models (ESMs). A qualitative comparison of total annual precipitation and climatology is the first step in detecting model deficiencies and thus improving our understanding of the world's climate. However, spatial (or association) analysis of the precipitation fields offers new insights into model performance and their ability to provide realistic predictions of rain and snowfall in both current and future climates. Here we analyse the spatial structure of precipitation according to 40 GCMs for 20 years (January 1980 to December 1999), quantitatively comparing the modelled precipitation against five observational datasets: three land‐only (CRU, PRECL and GPCC) and two global (GPCP and CMAP). We found discrepancies between the GCMs' predictions and the observational datasets and noted that satellite‐derived datasets are essential for pinpointing areas that require attention. The analyses also revealed a consistent trend towards less spatially correlated fields. This trend is not apparent in aggregated, traditional validation exercises but arises when spatial association indices are applied. So long as the trend is not an artefact in the observational datasets, then we suggest the tendency could be attributed to changes in stratiform/convective partitioning or the result of increasingly intense convection. Abstract : Numerical models are routinely compared with observations of precipitation in order to pinpoint areas that require attention or improvement. "Spatial structure metrics" can help identifying processes that are hidden to standard validation methods such as the analyses of aggregated mean climatological values. Here, drawing on 39 Global Climate Models and 5 observational databases, it was found that a process of spatial decorrelation of the global precipitation field features in the global satellite‐based observational data but not in the models.