Machine learning techniques for reference evapotranspiration and rice irrigation requirements prediction: a case study of Kerian irrigation scheme, Malaysia
Reference evapotranspiration (ETo) is a crucial component in agro-meteorological processes, where accurate estimation is imperative for planning and managing irrigation practices. ETo and rice irrigation requirements were first estimated using FAO Penman–Monteith (FAO-PM56) and the water balance mod...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Springer Science and Business Media Deutschland GmbH
2025
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/122675/1/122675.pdf http://psasir.upm.edu.my/id/eprint/122675/ https://link.springer.com/article/10.1007/s10333-025-01040-9?error=cookies_not_supported&code=55c22652-0c73-41da-a88f-c6dc209c1abd |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Reference evapotranspiration (ETo) is a crucial component in agro-meteorological processes, where accurate estimation is imperative for planning and managing irrigation practices. ETo and rice irrigation requirements were first estimated using FAO Penman–Monteith (FAO-PM56) and the water balance model, respectively, and the obtained results were used as reference values in the machine learning algorithms. Two machine learning algorithms, named Support Vector Regression (SVR) and Random Forest (RF), were applied to predict ETo and rice irrigation requirements using only climatic data (rainfall, temperature, relative humidity, and wind speed). The novelty of this paper is the application of machine learning techniques as an alternative to traditional methods and software solutions for estimating ETo and irrigation demand. Emphasis is placed on the advantages of data-driven models, such as eliminating the need for internal framework factors, facilitating realistic calculations, and providing more straightforward solutions for handling multiple variables. A comparative study is conducted to evaluate the performance of SVR and RF. Comparison results indicated that the SVR model outperforms the RF model in predicting ETo and irrigation requirements, with the minimum error observed in both the training and testing phases. All coefficient determination (R2) values are generally higher than 0.90 except for RF during the testing phase in ETo prediction, where the R2 value is 0.82. Alternatively, the excellent performance of SVR over RF for ETo and irrigation requirements prediction in the training phase was also demonstrated by the mean absolute error (MAE) and root mean squared error (RMSE). The results have demonstrated that machine learning techniques, particularly the SVR model, can be confidently utilized as alternative methods for predicting ETo and irrigation requirements, even with limited input data, with high accuracy. |
|---|