Similarity assessment and adjustment of integrated temperature databases. Cuyo region, Argentina

Abstract

Given the frequent spatial-temporal limitations and deficiencies of instrumental meteorological records, the use of alternative information sources, such as integrated databases, are important for analyses and studies of diverse nature. The research aim was to evaluate the accuracy of integrated databases of monthly temperature, belonging to Climate Research Unit, University of Delaware and Global Historical Climatology Network, gridded with a pixel size of 3,098.01 km2 (0.5º x 0.5º), surface area of 151,802.5 km2 and temporary length of 22 years (1993-2014), through the modified structural similarity index (mSSIM). The study area is located in central-western Argentina (between 30º and 35º S, and 71º and 66º W). The University of Delaware grid showed the best fit of the data series from 10 weather stations located in the study area. Therefore, a proposal was presented to increase similarity indices, especially for those cells without instrumental reference information. The study determined that by applying this modification, the gridded datasets increases the similarity of the measured data, especially in mountainous areas, where originally there were differences of more than 7.5 ºC between the gridded data and observed one. The proposal decreases these differences to average values below 1 ºC. The use and subsequent adjustment of these integrated databases, allows access to information in areas without meteorological records.