Modeling Evapotranspiration in IoT based WSN for Irrigation Scheduling: An Optimized DL Approach

The scarcity of freshwater resources throughout the world has raised the demand for the optimal utilization of these resources. Internet-of-things (IoT) based wireless sensor networking is an interesting and vital technology that has seen substantial growth in recent years. Agricultural applications are one of the fields where it is extensively used and effectively implemented to manage irrigation requirements. In smart agriculture, reference evapotranspiration ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$ET_{0}$</tex> ) has a great significance in determining the precise crop water requirement. Hence, accurate estimation of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$ET_{0}$</tex> is critical to avoid under or over-irrigation without compromising the agricultural productivity. This paper presents Genetic algorithm (GA) based optimized Long short term memory (LSTM) model (LSTM-GA) to estimate reference evapotranspiration ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$ET_{0}$</tex> ) using climate data acquired by IoT based wireless sensor networks (WSN). Specifically, we examine the temporal property of climate data by proposing a systematic way of determining the time window size for the LSTM model using Genetic algorithm. The climate variables include daily maximum temperature ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T_{max}$</tex> ), minimum temperature ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T_{min}$</tex> ), solar radiation ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R_{s}$</tex> ), sunshine hours (SSH), wind speed ( <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$U_{2}$</tex> ), relative humidity (Rh), and vapor pressure (Vp) of Ludhiana station. The proposed hybrid model was validated against the benchmark technique FAO-PM for estimating <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$ET_{0}$</tex> . The performance comparison revealed that LSTM-GA provided reliable results and outperformed the stand-alone LSTM model