Grain Storage Deterioration Risk Dataset Mold and Insect Infestation -via IOT

IoT sensor-fusion dataset simulating hourly grain storage conditions (temperature, humidity, moisture, CO2) across 20 storage lots spanning wheat, rice, and maize, three storage architectures, and three Indian regional climates. Supports mold/mycotoxin risk classification, insect infestation risk classification, and multi-hazard early-warning prediction, grounded in published water-activity mycology and stored-insect thermal-ecology research.

About Dataset

This dataset addresses post-harvest grain storage losses, a major documented component of India's overall food-grain post-harvest losses, by jointly modelling two spoilage hazards usually handled with a single undifferentiated temperature/humidity threshold: fungal growth/mycotoxin risk (driven by water activity, not raw moisture or humidity alone) and stored-product insect infestation risk (driven by a different temperature/humidity optimum and a multi-week population-growth timescale). The dataset simulates an hourly IoT sensor stream (grain-mass temperature, headspace relative humidity, grain moisture content, derived water activity, CO2 concentration) across 20 storage lots spanning three grains (wheat, paddy rice, maize), three storage architectures (bagged warehouse, steel silo, traditional kothi), and three Indian regional climate archetypes (humid eastern, semi-arid northwestern, subtropical northern) over a 90-day simulated period. The underlying models are grounded in the modified-Henderson equilibrium-moisture-content equation (calibrated to a published wheat benchmark), published water-activity/temperature growth windows for Aspergillus flavus/parasiticus and aflatoxin production, and published thermal-ecology data for Sitophilus oryzae and Tribolium castaneum, the two most economically damaging global stored-cereal pests. Three labelled ML tasks are provided: mold/mycotoxin risk classification, insect infestation risk classification, and an hours-to-high-risk early-warning regression, evaluated with a storage-lot-grouped train/test split. The dataset is synthetic-but-literature-grounded (no public IoT-resolution multi-hazard dataset for Indian grain storage exists); it is intended as a research/prototyping asset and as a template pipeline to be recalibrated against grain-specific lab EMC isotherms before food-safety-relevant deployment.

Purpose of Dataset

To Enable Early, Differentiated Warning Of The Two Dominant Post-harvest Grain Storage Hazards -- Fungal/mycotoxin Risk And Insect Infestation -- Which Existing Threshold-based Iot Grain Monitors Typically Conflate Into A Single Alert Despite Having Different Environmental Drivers And Timescales. The Dataset And Companion Model Let A Storage Manager Get Advance Notice (Rather Than A Same-day Alert) To Schedule Aeration, Turning, Or Fumigation Before Either Hazard Becomes Actionable, Directly Targeting India's Substantial Storage-stage Post-harvest Grain Losses. It Is Also Intended As A Reusable Template Pipeline: The Emc And Risk Models Can Be Recalibrated With Grain-specific Lab Isotherm Data For A Deployable, Region-specific Version.