Calibration and Validation of a Statistical Method to Downscale Daily Temperatures in the Central Pampas region.

BETTOLLI, MARIA LAURA; PENALBA, OLGA C; RAGGIO, GABRIELA

Santo Domingo

Workshop; WCRP VAMOS/CORDEX Workshop on Latin-America and Caribbean. CORDEX LAC: Phase II - The Caribbean.; 2014

WCRP VAMOS/CORDEX Workshop on Latin-America and Caribbean. CORDEX LAC: Phase II - The Caribbean.

The central Pampas region is extending itself towards the south of the La Plata basin. It is the most successful agricultural region of Argentina, producing great quantities of cereal crops and vegetables each year. The agricultural activity involves a broad set of decision making in which a variety of factors have significant influence. Climate is a source of variability and risk, specially taking into account the projected changes for the 21st century. In this context, the exploration and development of statistical downscaling techniques are of special interest for the region. The purpose of this work is to calibrate and validate a statistical method to downscale daily maximum and minimum temperatures in the Central Pampas. This work represents an extension to the advances in statistical downscaling addressed in a previous study submitted to the CORDEX-LAC I Workshop. Daily mean fields of the NCEP-NCAR Reanalysis 1 in the area bounded by 60°S, 15°S, 90°W and 44.5°W were used as predictor variables. Daily maximum and minimum temperatures from 25 meteorological stations located in the Central Pampas region were also used for the period 1979-2010. This information was provided by the Argentina?s National Weather Service. The statistical downscaling is based on the analogue method which consists in finding, for each day in the record, the most similar days (called analogue fields). This similarity between the fields was assessed by the Euclidean distance. The first ten analogues were retained and five different types of estimations were performed: 1) the average, 2) the median, 3) the first analogue field, 4) the weighted average giving more importance to first analogues and 5) the temperature value of a random analogue. The statistical model was calibrated in the period from 1979 to 2000 and validated in the period from 2001to 2010. Sea level pressure, air temperature at 850 hPa, geopotential height at 500 hPa, specific humidity at 850 hPa and the different combinations among them were tested as atmospheric predictors. For the two variables and for each type of estimation, different validation measures were considered. The bias, the absolute error and the mean square error were analyzed. The correlation between the estimated and the observed time series was also performed. The analysis was focused on evaluating the ability of the method in representing spatial and temporal variability of the variables as well as the annual cycle. Amongst the atmospheric predictors evaluated, air temperature at 850 hPa, was found to have the most skill. The method has a good performance at estimating spatial distributions, seasonal cycles and temporal variability. Both variables and all estimations showed annual cycle correlations greater than 0.98, though fall and winter temperatures tend to be overestimated. In general, the estimations of Tmin were better than for Tmax. The model skill in representing interannual variability also depends on the season. Interannual correlations were mostly greater than 0.59 for Tmax and 0.64 for Tmin. For both temperatures, errors show values smaller than the observed standard deviation for each season. In all cases, the spatial distribution of estimated fields is in agreement with the observed fields. The statistical model reproduces the spatial structure of the mean fields.