engleski

Monthly temperature and precipitation maps and derived climate monitoring products

Spatial analysis of climate variables measured at dense national observation network is key to understand the climate of a region. The monthly temporal scales of the gridded data that are available for a long observing period of several decades are essential for monitoring and understanding climate variability and change. The presented maps of temperature and precipitation amounts for each month starting from the year 1981 to present are a part of a new set of spatial data of the Croatian meteorological and hydrological service. Regression kriging (RK) was used to produce monthly maps at 1 km spatial resolution, preceded by an analysis of data quality. This analysis included a homogeneity analysis and filling of the missing monthly data. Leave-one-out cross-validation was applied for the evaluation of the RK model. The quality of the maps is compared with other similar products and prediction variance maps are calculated. Trends of measured and homogenized data are compared. For monthly precipitation, the homogeneity breaks were detected at 150 out of 459 time series. Most of the summer rainfall series show the tendency of drying. RMSE shows seasonality course, it is lower from March to August (12-14 mm). RMSE is higher in months with more precipitation, 16 mm in January and February, 19-22 mm from September to December. The accuracy of the RK model was from 0.75 in August to 0.82 in October and November which is considered to be an accurate estimate. Average RMSE for spatial prediction of the minimum, mean and maximum temperature was from 0.4°C 0.8°C, for maximum temperature with pronounced seasonality. Accuracy was higher than 92% for all months and temperature variables. Producing such gridded data of high quality is an important goal for climatologists, as well as for users from other natural sciences and society.

monthly temperature and precipitation maps, homogeneity, regression kriging

nije evidentirano