Utilising Machine Learning for Pan Evaporation Prediction - A Case Study in Western Australia
Evaporation has a significant impact on the management of water resources, irrigation system designs, and hydrological modelling due to its complex and nonlinear nature. This is because evaporation is a result of the interactions of various climatic factors. In Australia, research suggests that evap...
Saved in:
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Conference Paper |
| Published: |
Newswood Limited
2024
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Evaporation has a significant impact on the management of water resources, irrigation system designs, and hydrological modelling due to its complex and nonlinear nature. This is because evaporation is a result of the interactions of various climatic factors. In Australia, research suggests that evaporation causes about 40% of the water in open water lakes to be lost each year. Given the potential consequences of climate change, this water loss could become a major issue. This paper presents efficiency of Transformer Neural Network (TNN) approach in predicting monthly pan evaporation (Ep) through a case study in Perth, the capital of Western Australia. Daily meteorological data from a weather station in Perth was deployed for testing and training the model by utilising weather parameters, including maximum temperature, minimum temperature, solar radiation, relative humidity, and wind speed for the period 2009�2022. The Pearson correlation coefficient was used to determine the optimal ML model input parameters. Several models have been developed by combining different input combinations and other model parameters. To evaluate the ML model's performance, it was compared to Stephens and Stewart, a widely used empirical technique. The model's performance was subsequently assessed using standard statistical measures. The results of the performance evaluation criteria suggest that the Transformer model proposed in this study can effectively predict the monthly evaporation rate, benefiting from its self-attention mechanism. The proposed model performed admirably (R2=0.986, RMSE=0.031, MAE=0.025, and NSE=0.987). Additionally, it was demonstrated that the transformer model was more accurate than the empirical method for the same input sets, leading to a notable improvement in the estimation of monthly evaporation rates. � 2023 Newswood Limited. All rights reserved. |
|---|