A Wireless Underground Sensor Network Field Pilot for Agriculture and Ecology: Soil Moisture Mapping Using Signal...

Area of Science:

• Agricultural Engineering
• Environmental Monitoring
• Internet of Things (IoT)

Background:

• Wireless Underground Sensor Networks (WUSNs) are crucial for collecting in situ sensor data in agriculture and ecology.
• Reliable year-round operation of WUSNs is essential for continuous environmental monitoring.
• Mapping soil conditions using buried sensor data is a key application of WUSNs.

Purpose of the Study:

• To demonstrate the reliable year-round operation of WUSNs for determining and mapping soil conditions.
• To evaluate a deep learning algorithm for predicting soil volumetric water content (VWC).
• To assess the feasibility of using WUSN data and climatic parameters to replace expensive soil VWC sensors.

Main Methods:

• Designed and deployed a 23-node WUSN at an agricultural field site (530 m radius).
• Collected nine months of data on soil VWC, soil temperature (ST), and soil electrical conductivity.
• Utilized a deep learning model incorporating received signal strength (RSSI), distance (D), ST, air temperature (AT), relative humidity (RH), and precipitation to predict soil VWC.

Main Results:

• The WUSN operated continuously since September 2019, enabling real-time soil monitoring.
• A deep learning model using RSSI, D, AT, ST, and RH achieved an R² of 0.82 and RMSE of ±0.012 for soil VWC prediction.
• The model demonstrated high accuracy in predicting soil VWC across different seasons.

Conclusions:

• WUSNs can operate reliably year-round in field conditions for soil monitoring.
• Deep learning models combined with readily available soil and climatic data can accurately predict soil VWC.
• This approach presents a viable, cost-effective alternative to traditional soil VWC sensors in WUSNs.