Estimating the Hourly Reference Evapotranspiration with Fuzzy Inference Systems and Limited Weather Data

Abstract

Evapotranspiration is the most important part of the hydrological cycle, which plays<br />a key role in water resource management, crop yield simulation, and irrigation<br />scheduling. Therefore, developing a cost-effective and precise model is essential for<br />estimating hourly grass crop reference evapotranspiration (ETo). In this study the<br />potential of the fuzzy inference system (FIS) is investigated as a simple technique for<br />modeling hourly ETo obtained using the FAO-56 Penman-Monteith and ASCE<br />equations. Then, combinations of efficient hourly climatic data namely temperature,<br />wind speed, relative humidity and solar radiation were used as inputs to the fuzzy<br />model. Four fuzzy models were developed based on different combinations of inputs.<br />Common statistics such as Mean square error, average absolute relative error and<br />determination coefficient and two more statistics of Jacovides (t) and R2/t are used as<br />comparison criteria for evaluation of the model performance. Here, Training and<br />testing fuzzy models were done with Fariman meteorological data – an arid region in<br />the northeast of Iran. The fuzzy model whose inputs are solar radiation, air<br />temperature, relative humidity and wind speed, yield the highest correlation and<br />compatibility to reference models of FAO-56 PM and ASCE, based on common<br />statistics. Whereas, the fuzzy model whose inputs are solar radiation, air temperature<br />and relative humidity, are selected as the best model based on combination of<br />common and additional statistics. The fuzzy model with two inputs namely solar<br />radiation and relative humidity has acceptable results, too. The results show that solar<br />radiation is the most effective parameter on hourly reference evapotranspiration and<br />temperature, relative humidity and wind speed were other effective parameters,<br />respectively. These results for training and testing phase are alike. It was found that<br />the developed fuzzy models could be successfully employed in estimating the hourly<br />ETo with a limited weather data.