By Min Xu and Kurt Alles
Gradable is helping farms move to and be recognized for sustainable cropping systems. Sustainable agriculture means stewarding the resources farms rely on, including enhancing nitrogen use efficiency, reducing greenhouse gas emissions, improving water quality, investing in soil health, and maximizing farmer profitability. Over decades, producers have led the charge on developing new regenerative practices, namely cover cropping, reducing or eliminating tillage, rotating crops, incorporating livestock, and promoting biodiversity.
Advancements in and widespread adoption of agriculture technology on farms across the country have eliminated much of the burden of field- and, in some cases, bushel-level data collection. Machine data collected by combines and other farm implements now flows seamlessly from the field to the cloud, allowing farmers to monitor their farm and business health in more detail than ever before.
Consumer demand for sustainably grown agriculture products presents a new opportunity for growers to be rewarded for their stewardship and conservation. Gradable enables farmers to anonymously share their data with grain buyers, translating operational complexity into identity preserved sustainability metrics that buyers value. This creates environmental transparency in the agriculture supply chain while maintaining security over farmer data.
Identity preservation in the agriculture supply chain means environmental characteristics can flow from field to consumer. As with organic or non-GMO, identity preservation requires a degree of verification. Gradable is streamlining verification and reducing the burden on producers to access new markets. Using computer vision and machine learning algorithms, Gradable processes satellite imagery and millions of acres of crop production data to analyze results and verify field activity.
Below, read more about Gradable’s methodologies for verifying cover cropping and tillage, both essential practices in regenerative systems.
Gradable uses European wide-swath, 10-meter resolution Sentinel-2 twin satellites imagery, weather data, and practice data, gathered from FBN®’s 21,000 farmer members comprising over 59 million acres, to develop models for verifying tillage practices and cover crop usage at the field level.
FBN data on field-level cover crop use and tillage practices from 2017 to 2019 crop years were used as the ground truth in the model. Off-season (late October to early May) satellite data are sourced from Sentinel-2, and cloudy pixels are removed before image analysis. Precipitation data is used to account for soil moisture effects, and temperature data is used to track the cumulative heat units for cover crop development. Cash crop harvest and planting dates are estimated at the county level if unavailable.
Cover Crop Mapping
The model utilizes a time series of multiple spectral bands and vegetation indices from Sentinel-2 satellite imagery and weather data during the off season to create a classifier with 96% accuracy to examine spatial cover crop occurrences (Table 1). The intensity and range of winter greenness and timing of the greenness peak are the key features to train the cover crop classifiers. The training model uses data from Illinois, Indiana, Iowa, South Dakota, Minnesota, Ohio, Nebraska, Kansas, and Wisconsin from 2017 to 2019. Figure 1 shows examples of Sentinel-2 satellite imagery and Enhanced Vegetative Index (EVI) for fields with and without cover crops.
The challenges and misclassifications associated with cover crop mapping are related to cover crop establishment on cover cropped fields and significant weed presence in non-cover cropped fields. The type of cover crop being used also affects mapping accuracy, especially for cover crops that are not winter hardy, where availability of cloud- and snow-free satellite images can be a limitation.
Table 1. Matrix of cover crops mapping at the field scale using 10-fold cross validation.
Figure 1. Sentinel-2 satellite imagery and Enhanced Vegetative Index (EVI) show the differences between fields with and without cover crops.
Tillage Practice Mapping
To verify tillage practice, Gradable categorizes three different tillage practices (conventional tillage, reduced tillage, and no-till) using remote sensing. Normalized difference tillage index (NDTI) is used to distinguish field crop residue from soil based on cellulose and lignin absorption features found between 2100 nm and 2300 nm in the shortwave infrared region of the electromagnetic spectrum (Hively, et al., 2018). The time series of NDTI along with multiple spectral bands is analyzed for consistency and patterns to detect tillage practices on the farm. Figure 2 shows an example of NDTI change under no-till and conventional tillage practices.
Using FBN network data on field tillage practices, Gradable’s model classifies conventional tillage, reduced tillage, and no-till with an average accuracy of 77% and classifies no-till at 84% accuracy across 2017–2019. Data coverage includes Illinois, Iowa, Nebraska, Kansas, South Dakota, North Dakota, Minnesota, Ohio, and Wisconsin.
Unlike cover crop classification, the accuracy of tillage practice classification is significantly impacted by soil and residue water content, especially under wet conditions. Moreover, the presence of green vegetation can also physically obscure crop residue. Crop residue type also affects tillage estimation, as the amount of residue left varies with crop type.
Table 2. Matrix of tillage practice mapping at the field scale using 10-fold cross validation.
Figure 2. Satellite imagery and NDTI from Sentinel-2 demonstrating the differences between different tillage practices (Top: no-till after corn; Bottom: conventional tillage after corn).
About Farmer’s Business Network, Inc. and Gradable: FBN® is an independent agriculture technology platform and farmer-to-farmer network with a mission to power the prosperity of family farmers around the world, while working towards a sustainable future. FBN’s network consists of over 21,000 farmer members comprising over 59 million acres. FBN launched Gradable to provide technology and services to growers and buyers to facilitate the scoring, sourcing, and pricing of Low-Carbon Grain, building the infrastructure to make environmental transparency in the agriculture supply chain a reality.