Estimation of reference evapotranspiration and crop water requirment using artificial neural network for semiarid
Loading...
Files
Date
2018-04-20
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Vasantrao Naik Marathwada Krishi Vidyapeeth, Parbhani
Abstract
Evapotranspiration is a major and important component of hydrological process consisting of evaporation from land surface and transpiration of water from plants. Accurate estmation of reference crop evapotranspiration (ETo) in the water balance or irrigation scheduling allows to improve water resources utilization pattern in general and crop water management practices in particular for the region. Several methods including the most intricate energy balance methods requiring detail climatic data to simpler methods requiring less data were developed and modified worldover to estimate reference crop evapotranspiration (ETo) under various climatic conditions. Amongst them Penman-Monteith FAO-56 (PM 56) method has been accepted as the standardized method for precise estimation of ETo. For several resons the input data on climatic variables required for PM 56 method may not be easily available at every location because of either unavailability of nearby metereological station or difficulties in collecting accurate data on all necessary climatic variables at available station, especially in developing countries. Under such circumastance one may be forced to use data from the station which is far away, with completely different hydrometereological settings which manytimes restricts application of Penman-Monteith FAO-56 method. This directs towards development of simple alternative techniques like artificial neural network for accurate estimation of ETo for situations where values of some of the potential variables are not available.
Marathwada, the semiarid region always remains under the threats of frequesnt droughts and subsequent crop failure for want of water. Identifying the most suitable method or advance technique for accurate estimation of ETo, its forecasting for water resources planning and computing the crop water requirement of major cropping system of the region is therefore a major challenge. With this major goal, attempts have been made to analyse the daily weather data of two metereological stations of Marathwada region viz. Parbhani and Aurangabad to estimate the reference crop evapotranspiration (ETo) using different ETo estimation methods and developing inter-relationship between them alonwith estimation and forecasting of ETo using ANN models. The comparative performance of different indirect ETo estimation methods and ANN models in estimating ETo was also tested. Attempts were also made to develop modified crop coefficient of major crops of this region to determine crop water requirement for planning and management of water resources. The weekly reference crop evapotranspiration (ETo) estimated by indirect methods viz., FAO-24 Blaney-Criddle, Hargreaves Samani method, Priestly-Taylor, Turc, FAO- 24 Pan Evaporation, 1982 Kimberly Penman, FAO-24 Penman (c=1) were compared with ETo estimated by the standard FAO 56 Penman Monteith Method using daily climatic data for a period of 46 and 11 years, respectively for Parbhani and Aurangabad stations using DSS_ET Software.
The Trainlmfunction in MATLAB software (The MathWorks, 2009) through NN tool was used for ANN modelling. For forecasting and generation of evapotranspiration model, feed-forward neural network (FFNN) along with Levenberg-Marquardt back propagation algorithms, Tansigmoid transfer function and single layer network architecture with supervised training were used. Model performances were judged on the basis of various statistical parameters. The crop coefficients for Bt. Cotton, Soybean, Piegen pea, Gram, Safflower and Wheat were derived and modified for local conditions by using procedure suggested in FAO 56 and crop water requirements were estimetd based on average climatic parameters for Parbhani and Aurangabad conditions.
Results of the investigation revealed that 1982 Kimberly-Penman and FAO-24 Penman (c=1) method over-predicted ETo whereas Hargreaves, Priestley-Taylor and Turc radiation methods under-predicted ETo. Amongst the indirect methods, the temperature based Hargreaves method ranked first based on its statistical performance and showed closer agreement to ETo estimats of PM 56 method under limited data availability in semiarid region. The fifth order polynomials non linear inter-relationship between PM 56 and other indirect method for estimation of ETo was found to be the best fitted. Results showed that Hargreaves method was the best suitable method for Parbhani and Aurangabad condition hence it can be recommended as the best alternative to PM 56 method for accurate estimation of ETo. The ANN Model-A with 2-5-1 architecture using minimum and maximum temperature and ANN Model-B with 3-6-1 architecture using minimum temperature, maximum temperature and wind speed were found to estimate much closer ETo for Prabhani station whereas ANN Model-A and ANN Model-B having 2-11-1 and 3-9-1 architectures gave accurate ETo estimates for Aurangabad. Both these models require less input parameters. However in case of unavailability of wind speed data ANN Model-A with 2-5-1 and 2-11-1 architectures for Parbhani and Aurangabad, respectively may be used for ETo estimation. The study revelaed that all ANN models recorded superior performance than any of the indirect ETo estimation methods tested. For forecasting of ETo at Parbhani the forecasting Model-1 which uses only two inputs of previous one week’s ETo and previous year’s same week’s ETo of respective metereological week may be preferred over Model-3 requiring six inputs whereas for Aurangabad ANN forecasting Model 3 showed closer ETo estimates when compared with PM 56 ETo estimates. The results also indicated that there is a need of more than 10 years length of data sets for ETo forecasting by ANN.
The modified crop coefficients derived for Parbhani and Aurangabad in semiarid region are found to be higher which significantly differ from those suggested in FAO 56. The polynomial equations developed for crop coefficients for Bt. cotton, soybean, pigeonpea, chickpea, safflower and wheat can be used to determine daily or weekly crop coefficients which also can help to schedule irrigation even through micro-irrigation. Results also indicate that the kharif crops under Parbhani conditions have higher total seasonal water requirements than Aurangabad conditions whereas in Rabi season the total seasonal water requirements of crops at Aurangabad were higher than Parbhani. The modified crop coefficients developed for Bt. cotton, soybean, pigeonpea, chickpea, safflower and wheat are recommended to determine crop water requirements and water resources planning in the region.
Description
Keywords
null