Comparison of Gridded Weather Datasets with Point Datasets over Three Locations of India

Abstract

Climate information of any place is of prime importance for understanding the dynamics of natural resources in general and agriculture in particular. But availability of good quality, long-term weather data over large region remains a challenge. Hence, in this study a comparative analysis of temperature (maximum and minimum temperature) and rainfall was conducted between the manual observatory datasets with that of the gridded datasets of India Meteorological Department (IMD) over three places of Indo-Gangetic plain of India. In this analysis the mean monthly values of the parameters as well as the daily extremes were compared. The agro-meteorological stations data used in this study are from three districts namely Ludhiana, Hisar and Kanpur over north India. Results showed that, in case of monthly based comparison, the gridded datasets could explain almost all the variability (R2 >0.95) in the manual observatory/point datasets while could explain about 65% to 86% variability in the rainfall amount. The higher values of rainfall were seen to be underestimated by the gridded dataset. The root mean square error (RMSE) for maximum temperature varied from 0.71°C to 0.84°C while for minimum temperature it was higher in the tune of 1.25°C to 1.36°C. For, daily values of extreme temperatures gridded data could also explain almost the whole range of variations in the extremes of station data of all the three locations as indicated by very high value of coefficient of determination (R2). The nRMSE for maximum temperature varied between 2.1% to 2.5% while for minimum temperature the nRMSE ranged between 4.2% and 8.7%. Hence, it can be inferred that gridded temperature datasets may be directly used if meteorological station datasets are not available/accessible for any particular place.