A dataset of seven-year weekly rice pest surveillance, climate, and lunar-phase observations from Midsayap, Philippines.

This dataset contains weekly rice-pest surveillance and environmental observations collected at the Department of Agriculture–Philippine Rice Research Institute, Midsayap Experimental Station (DA-PhilRice-MES), North Cotabato, Philippines. It includes 366 weekly observations and 18 variables describing light-trap counts of Rice black bug (RBB), Scotinophara coarctata (Fabricius) and White stemborer (WSB), Scirpophaga innotata (Walker), meteorological conditions, surface-soil moisture, lunar phase, and calendar information. Environmental variables obtained from NASA POWER include mean, minimum, and maximum air temperature; relative humidity; corrected precipitation; mean, minimum, and maximum wind speed; wind direction; and upper-layer soil wetness. Lunar phase is represented using eight categories ranging from new moon to waning crescent. The repository also provides four derived binary-label datasets corresponding to the percentile thresholds used in the related pest-forecasting study: RBB P75.1, RBB P94.8, WSB P94.3, and WSB P98.9. These represent cumulative classifications of background versus elevated-or-severe abundance and background-or-elevated versus severe abundance. A positive label is assigned when the observed pest count is strictly greater than the stored numerical threshold. Each threshold directory contains: Row-level pest counts and binary labels Exact numerical thresholds and comparison rules Reproducible training and testing membership Class-distribution summaries JSON metadata describing the derivation procedure Combined files containing all binary labels and threshold summaries are also provided. The transparency files preserve observation identifiers and split membership but do not contain newly trained machine-learning models. The dataset may support ecological time-series analysis, pest-seasonality assessment, climate–pest association studies, outbreak-threshold comparisons, rare-event and imbalanced classification, alternative validation strategies, integrated pest-management research, and reproducible teaching exercises. Users should interpret environmental variables as associated covariates and not as demonstrated causal drivers of pest abundance. The dataset originates from one fixed monitoring location, and severe outbreak observations are comparatively rare. Users should consult the accompanying metadata and related research article before analysis.