Spatio-Temporal Modelling of Reference Crop Evapotranspiration Across Karnataka State, India

Abstract

Reference crop evapotranspiration (ET0 ) is a key variable required in the computa- tion of crop water and irrigation water requirements and also as an input in hydro- logical modelling. Due to the non-availability of direct ET0 measurements, ET0 is usually estimated using regularly recorded climate data. The physically-based FAO- 56 Penman-Monteith (PM) equation has been identified as a reliable method for this purpose but suffers from the disadvantage that it requires input data pertaining to a large number of climate variables. Therefore, in data short situations, simpler empir- ical to semi-empirical temperature and radiation-based methods may have to be used to estimate ET0 . However, their performances in diverse climates and the effect of local calibration on their accuracies have not been extensively studied. Therefore, the present study was taken up with the following objectives: 1) Assess- ment of spatio-temporal variations of climatic variables associated with the calcula- tion of reference crop evapotranspiration (ET0 ) over Karnataka State, India for the historical period 2006-2016 2) Assessment of spatio-temporal variations of Penman- Monteith (PM) ET0 for the study area for the same historical period 3) Performance evaluation of simpler alternative ET0 equations relative to PM ET0 with and without local calibration of parameters and 4) Development of a gridded PM ET0 product for Karnataka State. For this purpose, historical climate data from the period 2006-2016 for 67 stations located in various agro-climatic zones of Karnataka State, India was used. Spatio-temporal variability assessment was carried out on the historical data of eight climatic variables from 67 stations using statistical indices, box-whisker plots and spatial maps. The dataset consisted of daily values of six climatic variables, namely maximum air temperature (Tmax ), minimum air temperature (Tmin ), maximum rela- tive humidity (RHmax ), minimum relative humidity (RHmin ), actual hours of sunshine (n) and wind speed (u). The analysis clearly demonstrated the prevalence of distinct climatic regimes ranging from humid to arid. The overall characterization of climate iiiABSTRACT variables indicates the northern agroclimatic zones (ND, NED, and NT) exhibited high temperature, low humidity, and high sunshine values. Similarly, the southern region is characterized by low temperature and high humidity. Since the climate of Karnataka State is largely influenced by the monsoon phenomenon, distinct varia- tions in climatic variables arise on its account. A similar influence of the monsoon was seen in the intra-annual variations of mean ET0 estimates. It was seen that ET0 values peak during the month of April (4.5-5 mm d−1 ) and with the onset of the mon- soon rains in May-June and begin to decrease to around 3.0-4 mm d−1 . The spatial variability of ET0 across the Karnataka State indicates lower values in the south- ern, coastal, and hilly regions in comparison to northern regions. This difference is caused probably on account of the topographical effect created by the Western Ghat mountains which are located in the western part of the State. The primary objective of the present research was to develop a comprehensive method- ology for the evaluation of the performances of simpler ET0 equations with and without local calibration of parameters at a large number of climate stations located in Karnataka State, India. To achieve this objective, five simpler popular alter- native ET0 equations namely, temperature-based Blaney-Criddle (BC) and Harg- reaves (HG), radiation-based FAO-24 Radiation (RAD), Priestly-Taylor (PT) and Turc (TC) equations were tested for their performance against the physically-based combination-based FAO-56 PM method. Initially, the accuracies of monthly ET0 estimates obtained by the five alternative methods for the period of record were as- sessed relative to PM ET0 estimates using the coefficient of determination (R2 ), root mean square error (RMSE) and mean bias error (MBE) as performance measures. Overall, results indicate that the original PT equation performed the best among the 4 alternative simpler equations in terms of R2 and RMSE in most of the agroclimatic zones of the study area. However, the equation consistently over-estimated ET0 val- ues relative to the PM equation in several of the zones. Subsequently, numerical coefficients in the original forms of the alternative equations were treated as unknown parameters and optimized for each station using PM ET0 estimates as a reference. Results from local calibration of the five alternative ET0 ivABSTRACT methods indicated significant improvement in their prediction accuracies. Mean val- ues of model parameters for different agro-climatic zones in the study area and also maps showing spatial variability of the parameters were developed for the benefit of practitioners who wish to obtain estimates of ET0 comparable to the PM method us- ing limited climate data. However, replacing the Tmean term (with equal weightage for Tmax and Tmin ) with an alternative Teff with variable weightages for these vari- ables did not lead to any substantial improvement in the accuracy of ET0 estimates by the alternative equations. Furthermore, to provide ET0 estimates at large spatial scales in the State the study de- veloped a gridded daily PM ET0 product for a period of 2006-2016 at 0.25◦ ×0.25◦ res- olution using spatial interpolation techniques. Three interpolation techniques namely, inverse distance weightage (IDW), Kriging and P-BSHADE were chosen as possible candidates. Prior to application, the three spatial interpolation methods were evalu- ated for their prediction accuracies using a limited sample of climate stations. Ac- cordingly, the conceptually and computationally simpler IDW method was selected since the prediction accuracies were more or less the same for all the methods. Using daily PM ET0 estimates obtained at the 67 climate stations for the period 1st January 2006 to 31st December 2016, the IDW method was implemented to derive daily ET0 estimates at the 260 grid points. The accuracy of the gridded data product devel- oped in this study was compared with three other global ET0 data products available in the public domain. Results indicated that the gridded product developed in this study provided the most accurate estimates of ET0 in all the agroclimatic zones of Karnataka State. Web links for the gridded data product have been created in an ef- fort to share the data on ET0 which is a critical input in a variety of studies in earth sciences. It is hoped that researchers and practitioners will benefit from this unique data product.