Precipitation comparison for the CFSR, MERRA, TRMM3B42 and Combined Scheme datasets in Bolivia

Published in Atmospheric Research, v. 163,:17-131 
Authors

Blacutt, L.A., Herdies, D.L., de Gonçalves, L.G., Vila, D.A. and Andrade, M.F.
Publication year 2015
DOI https://doi.org/10.1016/j.atmosres.2015.02.002
Affiliations
  • Centro de Previsão de Tempo e Estudos Climáticos/Instituto Nacional de Pesquisas Espaciais (CPTEC/INPE), Cachoeira Paulista, SP 12630-000, Brazil
  • Laboratorio de Física de la Atmósfera &ndash Instituto de Investigaciones Físicas &ndash Universidad Mayor de San Andrés (LFA-IIF-UMSA) Campus Universitario Cota Cota calle 28 Edificio Facultativo p.2 Casilla 8635, Bolivia

 
IAI Program

CRN3
IAI Project CRN3035
Keywords
Precipitation
Precipitation estimations
Data assimilation
Statistical analysis

Abstract

An overwhelming number of applications depend on reliable precipitation estimations. However, over complex terrain in regions such as the Andes or the southwestern Amazon, the spatial coverage of rain gauges is scarce. Two reanalysis datasets, a satellite algorithm and a scheme that combines surface observations with satellite estimations were selected for studying rainfall in the following areas of Bolivia: the central Andes, Altiplano, southwestern Amazonia, and Chaco. These Bolivian regions can be divided into three main basins: the Altiplano, La Plata, and Amazon. The selected reanalyses were the Modern-Era Retrospective Analysis for Research and Applications, which has a horizontal resolution (~ 50 km) conducive for studying rainfall in relatively small precipitation systems, and the Climate Forecast System Reanalysis and Reforecast, which features an improved horizontal resolution (~ 38 km). The third dataset was the seventh version of the Tropical Rainfall Measurement Mission 3B42 algorithm, which is conducive for studying rainfall at an ~ 25 km horizontal resolution. The fourth dataset utilizes a new technique known as the Combined Scheme, which successfully removes satellite bias. All four of these datasets were aggregated to a coarser resolution. Additionally, the daily totals were calculated to match the cumulative daily values of the ground observations. This research aimed to describe and compare precipitations in the two reanalysis datasets, the satellite-algorithm dataset, and the Combined Scheme with ground observations. Two seasons were selected for studying the precipitation estimates: the rainy season (December-February) and the dry season (June-August). The average, bias, standard deviation, correlation coefficient, and root mean square error were calculated. Moreover, a contingency table was generated to calculate the accuracy, bias frequency, probability of detection, false alarm ratio, and equitable threat score.

All four datasets correctly depicted the spatial rainfall pattern. However, CFSR and MERRA overestimated precipitation along the Andes' eastern-facing slopes and exhibited a dry bias over the eastern Amazon TRMM3B42 and the Combined Scheme depicted a more realistic rainfall distribution over both the Amazon and the Andes. When separating the precipitation into classes, MERRA and CFSR overestimated light to moderate precipitation (1-20 mm/day) and underestimated very heavy precipitation (> 50 mm/day). TRMM3B42 and CoSch depicted behaviors similar to the surface observations however, CoSch underestimated the precipitation in very intense systems (> 50 mm/day).

The statistical variables indicated that CoSch's correlation coefficient was highest for every season and basin. Additionally, the bias and RMSE values suggested that CoSch closely represented the surface observations.