Comparison The ANN Models of Penman-Monteith Potential Evapotranspiration with Combination Two Input of Climatological Data in Surabaya
Danayanti Azmi Dewi Nusantara(*); Feriza Nadiar

Department of Civil Engineering, Faculty of Engineering, Universitas Negeri Surabaya, Ketintang, Surabaya, Indonesia.
*danayantinusantara[at]unesa.ac.id

Abstract

In Surabaya, as a part of the equatorial region, determine the rate of daily potential evapotranspiration (PET) turns into a requirement. Along with, during the scarcity season, the rate of PET significantly increases for certifying the water availability. The amount of PET advance into decisive for water supply cases such as irrigation, water supply, hydropower, and other. The PET model founded from several inputs of climatological data that are relative-humidity, wind speed, average daily temperature, and the duration of sun exposure. As a caution, modeling a PET through prolonged and complicated steps. To simplify the development of modeling PET, this research using Artificial Neural Network (ANN) based on data-driven modeling. The PET-ANN model intends to match the PET estimated with Penman-Monteith (PM). This research purpose of learning what the best combination of two climatological data as an input to the PET-ANN model. The perform MSE and R on the validating process of PET-ANN present how the different results come. The results show the best combination of two climatological data from the entire data set as an input. The conclusion is that the combination of relative humidity and wind speed as an input to the PET-ANN presents the best result than other combinations of climatological data. Besides, it approves that the relative humidity and wind speed as an undoubted input to the PET model even using ANN or not.

Keywords: Potential Evapotranspiration, Artificial Neural Network, Climatological Data

Topic: Civil Engineering