Making crop innovations more accessible for farmers with cloud data strategies

Microsoft has been conducting research using data-driven agriculture and artificial intelligence (AI) to solve some of the hard problems of agriculture.

In its work to accelerate innovation across the agriculture value chain, Microsoft has recently announced Azure Data Manager for Agriculture, which extends the Microsoft Intelligent Data Platform with industry-specific data connectors and capabilities to connect farm data from disparate sources, enabling organizations to leverage high-quality datasets and accelerate the development of digital agriculture solutions.

It’s not just about growing food. It’s about growing good, nutritious food - and growing it in a sustainable way without harming the plant. One of the most promising ways to address that problem is through data-driven agriculture. If farmers could capture data from different parts of the farm and then use artificial intelligence on top of that data to add value, they can improve efficiencies and production across every acre of land.

A key challenge for farmers is accurate weather forecasting. By accessing historical weather sensor data from a particular farm, an AI model can be trained to produce more accurate micro-climate information on an acre-by-acre basis.

There are other benefits to combining various data sources to unlock new insights and create AI models and solutions. As an example - using airborne sensors and drones for weed detection in agriculture and using common data models to harmonize data in supply chains.

Common data models can open up synergies across industries.

The Microsoft Azure cloud-based solution also allows for privacy-preserving data sharing. With low friction and fluidity in the data level, it is possible to reduce friction and increase velocity, thus lowering time to value.

To enable precision agriculture, a farmer needs to know what the farm looks like, not just what’s above the soil but also what’s below it. How to create that view? By bringing data from sensors, satellite imagery and drone imagery together using AI and computer vision techniques, then merging them to create views of the farm that a farmer otherwise couldn’t get. Farmers can then make more granular decisions about managing their fields. AI can also be used to predict solar and wind energy production, and how technology can help with climate adaptation.

One data source for farmers is satellite imagery. By combining and fusing high-resolution satellite imagery from multiple satellites, Microsoft’s Project SpaceEye uses AI to create daily cloud imagery. Built on Azure by Microsoft Research, SpaceEye is an AI-based system that generates daily cloud-free optical and multispectral imagery of Earth. SpaceEye uses the Synthetic Aperture Radar (SAR) instrument from the Sentinel-1 mission to produce baseline data, since radar isn't affected by cloud cover, and then combines the radar data with historical optical data to generate an AI image that predicts what a scene may look like under the clouds. Microsoft states that this can unlock significant use cases in agriculture, land-use monitoring and disaster response among others.

Every crop, soil and region is different, which makes it difficult to apply one model everywhere. The top models are specific to crops and need to be trained by experts. However, the cost of obtaining data is high, and the target customer is not clear. The pipeline's entire supply chain, from farm to retailer, can benefit from AI technology and when enough value is created to reduce costs, it can be made accessible to individual farmers.

One of the use cases of interest to retailers is product recalls, as AI can make the process faster and more efficient. However, other aspects to consider include who pays for the data and who benefits from it.

Alternate sources of data can be used in many interesting agricultural applications. Data sources such as tractor information, drone fleet, farm activity data and John Deere operations center information are a few examples. Historical data of farm activities is required to compute the carbon sequestration depending on the processes and activities done in the next year or two. However, there are challenges with collecting and integrating farm activity data, which can be messy because it relies on operators with different levels of expertise. AI tools can be used to massage the data and move it into a common model.

Carbon sequestration: The process of capturing and storing atmospheric carbon dioxide. It is one method of reducing the amount of carbon dioxide in the atmosphere with the goal of reducing global climate change.

When scaling up the use of technology and AI in farming, there is still a need to maintain a “human-in-the-loop” (HITL) element to ensure that context is not lost, and more information can be learned from experts in the field. HITL refers to systems that allow humans to give direct feedback to a model for predictions below a certain level of confidence. The goal is not to replace farmers with technology but to augment them with more technology and AI-enabled insights.