A consistent gauge database for daily rainfall analysis over the Legal Brazilian Amazon

Abstract

Reliable precipitation information provided by rain gauge networks is invaluable for applications ranging from climatological studies to validation of remote sensing rainfall estimates. Daily precipitation data from 235 tipping bucket gauges in the Legal Brazilian Amazon are quality controlled for the period 1998-2009 by comparing precipitation time series from individual rain gauges with neighboring stations. Within a geostatistical framework, optimal distances for the comparison between stations are identified while considering the density of the network and the spatial structure of the precipitation. The number of rain gauges used as neighboring rain gauges varies between 2 and 57. Each gauge station is compared to nearby gauge stations using fourteen climate and statistical indicators. A proposed variability index allows for the detection of ‘suspicious’ precipitation data before more detailed analysis of the nature of the issue and the time period of concern in the time series. The internal consistency of the database is examined through the use of variograms and cross validation. The sensitivity of the effectiveness of the quality control procedure to the comparison distance is examined. These results are positive and this demonstrates that the spatial structure of daily rainfall is better sampled. The cross validation confirms that the quality control procedures improved the consistency of the database. Finally, the spatial consistency of the rainfall database is assessed through a classification of the seasonal regimes in the Legal Brazilian Amazon.