Efficient Vineyard Irrigation Series #2: How to Predict Soil Moisture for the Next Two Weeks
Photo: James Dimas on Unsplash
This is Part 2 of Deep Planet’s three-part series tackling efficient vineyard irrigation scheduling using VineSignal. You can find Part 1 (about optimising soil moisture measurement) here and Part 3 (about dynamically optimising irrigation scheduling) here.
As viticulturists and grape growers know, understanding current and future soil moisture content is a prerequisite for taking actions to reduce crop variability, optimise water usage, and tackle irrigation problems. Adequate soil moisture is critical for ensuring fruit quality. Water stress and over-vigorous growing conditions can result in inferior fruit quality and reduce profitability.
Precision viticulture tools with soil moisture forecasting capability thus provide valuable input into growers’ and viticulturists’ decision-making.
Traditional methods
Current best practice involves growers installing soil moisture sensors in multiple locations (typically 1-6 sensors per 100 hectares) and interpolating to get an average reading of soil moisture across the vineyard. There are three main challenges with this method:
Lack of forecasting. Ground sensors only measure current soil moisture content, not predicted content. Using current data to estimate future soil moisture content can result in non-optimal irrigation decisions, especially given the increasing unpredictability of weather conditions brought about by climate change.
Accuracy. Changing soil composition and depth across even small irrigation blocks means that while sensors give accurate readings, viticulturists risk a significant margin of error in manual estimation between the sensors.
Cost. Sensors come at a high cost, both in terms of capital equipment and the manual labour needed to install and maintain them.
Our solutions
We have worked with viticulturists and growers to develop our VineSignal product. VineSignal is an AI-powered platform that analyses data from satellite imagery, ground sensors, weather stations, irrigation systems and proprietary sources to map actual and predicted soil moisture readings at every point across your vineyard.
VineSignal’s soil moisture prediction functionality enables growers and viticulturists to:
Map predicted soil moisture content across the vineyard and to a depth of 1.5m for the next 1-2 weeks. Our world-class machine learning models adapt for landscape particulars like ground roughness, soil type and vegetation to forecast with 92% accuracy. We will also provide predictions of soil moisture content at each probe point the next 1-2 weeks.
Compare soil moisture change over the previous week and the previous month to see the precise impact of irrigation and weather patterns on the sub-terrain.
Example soil moisture probe two-week prediction (solid lines actual values, dotted lines predicted)
Example prediction of soil moisture interpolation 3D model at all points and depths across 35 hectares
VineSignal’s AI-backed recommendations are intended to supplement customers’ existing soil moisture monitoring methods in a cost-effective and scalable way. Accurate soil moisture forecasts provide valuable input into vineyard management decisions that optimise fruit quality, yield and profitability.
Our team has conducted extensive research on applying machine learning models to satellite imagery, publishing research papers at ICLR and NIPS conferences. See our papers here and here.
Our team has co-developed our Artificial Intelligence tools for efficient vineyard irrigation by working with growers in Australia. Schedule a VineSignal demo here.