Creating world's first digital identity for farmers using satellite imagery and AI

The current problem is the lack of historical and in-season data to assess credit risk on smallholder farmers, this due to a lack of infrastructure consisting of agricultural land classification, crop classification, long-term productivity assessment (20-30+ years) on the individual farm-land and field boundaries. Currently, field boundaries are manually created by field agents walking the corners of a physical agricultural field and geo-tagging those boundaries, this is time-consuming, expensive and often inaccurate. In order to provide a reliable and affordable solution the only way is to automate this through the use of deep learning object detection and high-resolution Satellite data.

An additional challenge in the smallholder markets is firstly that 84% of the world’s 570 million farms are smallholdings; that is, farms less than two hectares in size. Many smallholder farmers are some of the poorest people in the world. Tragically, and somewhat paradoxically, they are also those who often go hungry. Lastly, currently 29% of the world's agricultural food production is produced in smallholder market but this is forecasted to change drastically as smallholder farmers gains access to better agronomic advise, crop-input prices and micro-financing.

Smallholder farmers face significant challenges worldwide, including improper infrastructure, lack of funding, gender biases and economic difficulties. The broader agricultural value chain system has been developed in a way that the farmer has been marginalised and left with no real options of escaping economic hardship. This project will revolutionise how we can help smallholder farmers own their own data, independently from the supply chain, and use this data to negotiate better financing, crop-input supply and insurance (among other) terms, this will give farmers more control, autonomy, power and control and be able to leverage this towards the broader system operators.

The wider societal impacts of our solution includes alignment to UN’s Sustainable Development Goals including: 2. Zero Hunger through contributing towards “achieving food security and improved nutrition and promoting sustainable agriculture”, 1. No Poverty through helping low-income farmers (smallholder markets) build preventive actions against “climate-related extreme events”, 15. Life on land: through contributing to “halt and reverse land degradation and halt biodiversity loss”, 13. Climate change: enabling farmers (end-users) to reduce urful angle for negotiating terms with agricultural lending institutions, credit rating, insurance premiums and crop input suppliers, enabling farmers to take back the control of their own data.

Additionally, in order to provide additional context on the agricultural market in general:

• 40% of all agricultural fields are over fertilised

• Farmers are losing 10-15% on adequate input application (crop protection, seeds and fertiliser)

• In the most advanced agricultural nation, US, still only 25% use precision ag-services

• Agricultural fields in Tanzania and Kenya are on avg. smaller than 0.5 ha’s rendering publicly and freely available SatEO insufficient, i.e. this is a largely untapped market

The solution we're building will enable smallholder farmers to easier access financing, agronomic advisory and build their credit profile. Furthermore, with nearly 80% of households in Tanzania engaging in agriculture and at least one third gaining more than half of their income from agricultural activities, while the agriculture sector in Kenya employs more than 40 percent of the total population and 70 percent of the rural population, access to finance for small-scale producers is a major catalyst to broad based economic growth. For a long time and especially in traditional forms of financing, one of the key limiting factors for access to loans for smallholder farmers has been lack of collateral. Looking at land ownership registration for instance, data collected by Tanzania’s bureau of statistics in 2018 shows that out of 8.7 million farms surveyed only 18% were registered.

In addition to this - Small-scale farming systems already grow 50% of our food calories on 30% of the agricultural land. When access to inputs and conditions are equal, smaller farms tend to be more productive per hectare than much larger farms.

What we're developing in this project is an "enabler" or a baseline of fundamental data that the ecosystem can build further upon as a community, this is an important component of the vision as digital field boundaries, historical productivity data in a fully independent and unique format (block-chain) represents significant value not only for the Cardano ecosystem as a positive influence on sustainability, digitisation and "oracle" in agricultural farming market as well as this project will directly affect the lives of smallholder farmers in both Kenya and Tanzania, opening up a dataset to entire 68 million crop land to create opportunities, this has not been done before due to several bottlenecks such as:

Availability of high resolution Satellite data: 

• DigiFarm's deeply resolved Sentinel-2 at 1 meter represents a unique technology ideally suited for the region, seeing as the high resolution is needed to be able to derive analytics for the boundaries which are typically very small, i.e. 0.5 hectares and below, the alternative to using our data would be to have to buy commercial grade SatEO and the closest resolution wise would be having to buy Airbus Pleaidas data at 50cm which costs €8 per sq.km, considering both countries include a total land area of 680,000 sq.km and we'd require 3-dates the total cost would be €16,320,000 just for the input data to deliver and provide these data layers in the region.

  
  

• Additionally, seeing as DigiFarm leverages open source and publicly available data from Sentinel-2 and then run our deep-resolution on it from 10m to 1m this is ideal to be able to keep promoting the open source and independent use of data to support the ecosystem.

  
  

• Lastly, if we could reach 50,000 farmers which is the goal at the end of this project (and to extend this to 500k the following 18 months) and save 10% on crop input and lower insurance/financing cost by 10%, this is a critical and high potential project.

So far we have made the following progress in the project:

• Created a technical accuracy report validating the accuracy of the model output, i.e. accuracy of automatically delineated field boundaries in Kenya based on the deeply-resolved Sentinel-2 at 1 meter per pixel resolution, verifying also the accuracy of above 90% (IoU of 0.90 and above), based on where training data is comprised of 90% and 10% is validation data (i.e. from manually digitised boundaries).
  
  Here is the full technical report (PDF) which includes extensive examples and data + metrics on the model output of the broader agricultural field boundaries delineated from the Kenya area of interest targeted.
  
  

• Created a custom web-based application for Kenya to be displayed with the deeply-resolved Sentinel-2 imagery to show the accuracy of the model output.

  
  
  
  

  

  
  
  Here is the demo link which was created custom for this project which highlights one of the AOIs, with approx. 50 fields randomly selected which highlights visually the accuracy of the output model along with the meta-data and visual basemaps including deeply-resolved Sentinel-2 at 1m per pixel (3 dates) as well as Mapbox (outdated Maxar data). se of chemical fertilizer/crop protection and Co2 emissions and 8. Decent Work and Economic Growth: enabling farmers (cereal-producers) to optimize their operations, i.e. save costs and increase economic productivity.

Benefits included improved access to micro-financing for crop-input, i.e. ~10-12% lower cost

Digitisation of field boundaries and measuring of this opens up a new market for carbon credit accessibility, new revenue streams for farmers, i.e. ability to increase revenue per ha of ~40%

Ability to have an independent source of field-level data, historically up to 35 years creates a powerful angle for negotiating terms with agricultural lending institutions, credit rating, insurance premiums and crop input suppliers, enabling farmers to take back the control of their own data.