search

FedoAIworld / Time-Series-Soil-Moisture-Prediction-using-One-dimensional-Convolutional-Neural-Network

Analyzed time series data of soil moisture using a 1D convolutional neural network to capture the non-linearity and dependency between electrical conductivity and soil moisture across depths.
2 stars 2 forks source link

readme

Soil Moisture Prediction using One-dimensional Convolutional Neural Network

Soil moisture prediction using one-dimensional convolutional neural network. Soil moisture is the amount of water stored in the unsaturated zone, where the soil is made of different layers (soil profile) each with different properties. Estimation of soil moisture is highly useful in irrigation scheduling, crop yield forecasting, drought warnings and runoff predictions.

Reference Data

The reference dataset was collected in the Chicken Creek catchment in Lusatia, Germany. Further description of the catchment is provided on the paper "Machine learning to estimate soil moisture from geophysical measurements of electrical conductivity" written by Davood Moghadas and Annika Badorreck (2019). Measurements of soil moisture, temperature and soil electrical conductivity were collected at different depths (0.2, 0.4, 0.6 and 0.8 m) for a period of 5 months. A 1D CNN was constructed with the dataset considering soil electrical conductivity and temperature as input to the network. As a result, the network was able to capture the interdepencies between soil electrical conductivity and soil moisture content in different depths and showed strong correlation.

Time series plot

Loss plot