AgriField-Manipur: A Hybrid Constraint-Based Dataset for Agricultural Decision Support in Data-Scarce Valley Regions (Version 3)

AgriField-Manipur (Version 3) is a simulated field-level agricultural dataset designed to support machine learning and decision-support research in data-scarce valley agricultural regions, with a focus on the Manipur valley in North-East India.

The dataset provides daily time-series observations for simulated agricultural field plots, generated using a hybrid framework that combines rule-based agronomic constraints with stochastic variability. Although synthetic, all variables are bounded by agro-climatic, soil, and crop-physiological references, supporting agronomic plausibility and internal consistency.

The dataset comprises 94,900 records, where each record represents a single day of observation for a simulated field plot. It includes environmental variables (temperature, humidity, rainfall), soil indicators (soil moisture, soil type, pH), relative macronutrient indices (N, P, K), crop identity, and downstream decision-support outputs. These outputs include a continuous crop suitability score, a binary suitability classification, and rule-based advisory indicators for irrigation, fertilizer application, and pest or disease risk.

The dataset covers five valley districts of Manipur (Imphal East, Imphal West, Thoubal, Bishnupur, and Kakching), spans the period from 1 December 2024 to 30 November 2025, and supports 52 crops relevant to Manipur valley agriculture. Geographic coordinates are provided at an approximate district level to preserve privacy while enabling spatial analysis.

Version 3 represents a documentation-aligned release. The dataset itself remains unchanged from Version 2, with updates limited to documentation and metadata to ensure consistency with the associated research work.

This dataset is synthetic but agronomically grounded and should not be interpreted as direct field measurements. Advisory outputs are generated using expert-defined rules and are provided for research purposes only.