Drought prediction data for IoT

Title: Synthetic Daily Drought Prediction Dataset (2-Year)

Description:
This dataset contains synthetic daily measurements of soil and rainfall conditions for 2 years (730 days) designed for drought prediction research and IoT applications. It includes continuous and binary features that simulate realistic environmental conditions, along with a binary drought label.

Features:

date – Daily date (YYYY-MM-DD)

soilMoisture – Soil moisture percentage (continuous, 5–80%)

soilDigitalWet – Soil wetness indicator (binary: 1 = wet, 0 = dry)

rainPercent – Daily rainfall percentage/probability (continuous, 0–100%)

rainDetected – Rain presence (binary: 1 = yes, 0 = no)

drought – Drought label (binary: 1 = drought, 0 = no drought)

How it was created:

Soil moisture and rainfall were simulated using seasonal sinusoidal patterns and random noise to reflect realistic environmental variability.

Binary features were derived from thresholds on soil moisture and rainfall.

Drought labels were generated using rule-based conditions combining low soil moisture, absence of rain, and dry soil.

Additional random noise was added to introduce realistic uncertainty, making the dataset suitable for training machine learning and deep learning models.

Usage:

Can be used for training sequence-based models like GRU/LSTM or classical ML models for drought prediction.

Ideal for IoT simulations, educational purposes, and experimentation with real-time drought detection systems.