Multi-decadal variability of tropical rainfall
Title | Multi-decadal variability of tropical rainfall
|
Technical memorandum
|
|
Date Published |
12/2024
|
Secondary Title |
ECMWF Technical Memoranda
|
Number |
922
|
Author | |
Publisher |
ECMWF
|
Abstract | We present results from a statistical procedure aimed at finding temporally coherent signals in tropical rainfall from different datasets. Specifically, we compare rainfall from the ERA5 re-analysis in the period 1950-2022 with a dataset obtained by merging GPCP v3.2 data for 1983-2022 with 1950-1982 rainfall estimated by a regression of GPCP data against rainfall indices from the CERA20C re-analysis. For each sub-sample of 3 consecutive months, a canonical correlation analysis (CCA) is performed on the first 21 principal components (PCs) of each dataset; from the CCA, two sets of rotated PCs (one for each dataset) are found, which are orthogonal within each set and highly correlated with the corresponding rotated PC (RPC) of the other dataset. The associated spatial patterns are computed taking the covariance of the RPCs with the original rainfall anomalies in each grid point. Filtered anomalies are then reconstructed from the first 16 RPCs and the associated spatial patterns (REOFs). The purpose of this procedure is to filter out variability components and trends which are poorly correlated across the two datasets, and which may therefore originate from changes in the assimilation methodology or in the availability of data sources across the 72-year period. The two leading RPCs and REOFs are representative of rainfall variability associated with El Niño - Southern Oscillation (ENSO), and show a striking similarity across the two datasets. Subsequent modes, representative of rainfall variability over the Indian Ocean, are also well correlated but show differences in regional features. Looking at correlations of reconstructed grid-point data from the merged-GPCP and ERA5 data, it is found that the CCA-filtered anomalies provide a reliable estimate of rainfall variability in the ocean regions with large rainfall variability, but are unable to improve the poor correlation between the merged-GPCP and the ERA5 data in some dry regions such as the Sahara and upwelling regions on the western side of Africa and South America. Finally, examples of tropical and extratropical teleconnection patterns computed from the CCA-filtered datasets in different period are presented, which show the potential for uncertainty estimation in diagnostics of inter-annual and inter-decadal variability of tropical rainfall and its remote impacts. |
URL | https://www.ecmwf.int/en/elibrary/81636-multi-decadal-variability-tropical-rainfall |
DOI | 10.21957/5952e84bb7 |
Download citation |