Estimation of daily reference evapotranspiration by hybrid singular spectrum analysis-based stochastic gradient...

Area of Science:

• Hydrology
• Meteorology
• Data Science

Background:

• Accurate estimation of reference evapotranspiration (ETo) is crucial for water resource management and agricultural planning.
• Traditional methods for ETo calculation can be data-intensive and computationally demanding.
• Soft computing methods offer promising alternatives for efficient and accurate ETo estimation.

Purpose of the Study:

• To estimate reference evapotranspiration (ETo) in the Adiyaman region of Türkiye using Stochastic Gradient Boosting (SGB).
• To enhance ETo prediction accuracy by decomposing time series data using Singular Spectrum Analysis (SSA).
• To evaluate the performance of the proposed SGB-SSA framework for both short-term and long-term ETo forecasting.

Main Methods:

• Reference evapotranspiration (ETo) was calculated using the FAO-56-Penman-Monteith method.
• Stochastic Gradient Boosting (SGB) was employed to estimate ETo using meteorological data (maximum/minimum temperature, relative humidity, wind speed, solar radiation).
• Singular Spectrum Analysis (SSA) was used to decompose ETo time series into sub-series for improved prediction accuracy, with SGB applied to each sub-series.

Main Results:

• The SGB-SSA framework demonstrated statistically acceptable outcomes for ETo estimation.
• The model achieved high prediction accuracy, validated by root mean square error (RMSE) and Nash-Sutcliffe efficiency (NSE) indicators.
• The framework successfully estimated both short-term and long-term ETo values, considering three lag times.

Conclusions:

• Stochastic Gradient Boosting combined with Singular Spectrum Analysis provides a robust and accurate method for estimating reference evapotranspiration.
• The proposed approach enhances prediction capabilities for hydrological and meteorological applications in data-scarce regions.
• This study contributes a valuable tool for optimizing water resource management and agricultural practices through improved ETo forecasting.