<![CDATA[Reliable microclimatic records are a prerequisite for evidence-based agricultural planning, yet high-resolution ground-truth datasets for tropical coastal environments remain scarce in the Indonesian literature. This study analyzes six months (December 2025 – May 2026) of 10-minute interval observations across 18 meteorological variables from the Parangtritis Automated Weather Station (AWS), Bantul Regency, Yogyakarta, Indonesia. Rainfall was reconstructed via a differential method from the cumulative station counter, yielding a period total of 1748.4 mm over 180 days. Of these, 112 days (66.1%) recorded measurable precipitation, punctuated by seven dry-spell episodes; the longest extended 12 consecutive days (5–16 May 2026). Schmidt-Ferguson classification returned Q = 20%, placing the site in Climate Type B (Wet/Basah). Reference evapotranspiration (ET₀, Hargreaves–Samani) averaged 10.20 mm/day (total: 1796.0 mm), and a PDSI proxy indicated extreme drought conditions by the close of the observation period (PDSI = −7.45), a deficit attributable primarily to persistently high ET₀ demand rather than rainfall deficiency per se. Cross-correlation analysis identified relative humidity as the dominant concurrent temperature predictor (lag 0; r = −0.676). An XGBoost model achieved short-term temperature forecasting accuracy of MAE = 0.56°C and RMSE = 0.71°C (R² = −1.419, reflecting the site's narrow diurnal thermal variance rather than model failure). Prophet projected sharply reduced rainfall for June–August 2026 (0.25–4.77 mm), which the generic staple-crop suitability framework scores as Not Recommended, reflecting unsuitability for rainfed rice, yet these drier conditions are agronomically favourable for shallot in coastal Yogyakarta.]]>
Copyrights © 2026
