Daily Root Zone Soil Moisture and Meteorological Dataset for Time Series Prediction (2020–2025)

This dataset contains daily meteorological variables and root zone soil moisture data collected from 1 January 2020 to 31 December 2025. The dataset was constructed to support time series prediction and comparative modeling using machine learning and deep learning approaches. Meteorological data were obtained from BMKG (Badan Meteorologi, Klimatologi, dan Geofisika), Trunojoyo Meteorological Station. The observed variables include air temperature (Tavg, °C), relative humidity (RH_avg, %), rainfall (RR, mm), and sunshine duration (SS, hours/day). Root zone soil moisture data were retrieved from NASA POWER, derived from the MERRA-2 reanalysis system. The parameter used is GWETROOT (Root Zone Soil Wetness), which represents a unitless soil wetness index ranging from 0 to 1 within the 0–100 cm soil depth layer. This index reflects the relative degree of soil moisture saturation in the root zone rather than absolute volumetric water content (m³/m³). The soil moisture data correspond to the same geographic coordinates as the meteorological station (latitude −7.0398, longitude 113.914). All variables were aligned on a daily temporal scale. Special codes for missing or unavailable values (e.g., −999) were identified and handled during preprocessing. Zero rainfall or zero sunshine duration represent valid observations rather than missing data. The dataset is structured for supervised learning, where meteorological variables serve as input features and the GWETROOT soil wetness index serves as the prediction target. The dataset is suitable for time series forecasting, environmental modeling, irrigation recommendation systems, and agricultural decision support applications.