Inageo

Recently remote sensing technology for agriculture is becoming more available, in the US there are already several offers of apps that enhance a farm’s performance, however in markets such as México the cost is unfeasible for most of the farmers as they are already at a disadvantage with US farmers since the cost of diesel and credit interest rates are twice as high in México.

We propose to solve the cost-efficiency of the technology to make it accessible for a different segment of farmers that can not pay for it, this is done by reducing costs as much as possible and finding a new market that will allow to pay for the operation of the system and R&D to improve it.

If scaled we can globally we can enable every farmer with free access to actionable intel based on public data.

Remote sensing tech to improve agricultural practices is already available but is not very accessible to Mexican farmers, this is mainly due to cost. The average US farmer investment in software is 1,500 USD which is not feasible for most of Mexican farmers.

From more than 4 million the country has, 69% has 5 hectares or less which means that at most their total crop value is around 2000 USD, most of it is for self-consumption so there is not much room to invest in software, in the 2017 national agricultural survey, high input and services costs were listed as the main problem for 75% of the surveyed farmers.

Still 47% of the rural population has internet access, almost all using a mobile device, taking advantage of this we look to distribute information to the farmer. By using public data, building everything with open software and an efficient design, we can solve the software part and make it cheap enough to give it away.

By reusing the software platform we can make low cost tools for agricultural monitoring which are useful to insurance funds, creditors and government entities that provide aid and with that income scale the farmer’s platform.

Cloud software platform that integrates multiple data sources and processes them to make human readable indexes, distributes them through a mobile app to farmers, does it very cheaply.

For now data is obtained from Sentinel-2, Landsat 8, the UC in Santa Barbara and a commercial provider ClimaCell. Raw imagery is processed using ESA’s Sen2-Agri platform to get surface reflectance products and cloud masks, TIF products are optimized for cloud access and stored in AWS, cloud masks are transformed into geometries for storage and querying in a spatial database. Indexes such as yield potential (NDVI) or rainfall accumulation (CHIRPS) are computed and aggregated for an individual plot on request and cached for future requests.

A Python web server provides a REST API to request this data, another Python app serves as a front-end that allows the farmer to specify his plot of land and query the API to get cloud-free observations and visualize indexes, we also include other useful information as weather forecasts, market prices and regional expected yields. We are looking to use the accumulated data to train ML algorithms that produce better indexes such as better yield estimations.

The target are the farmers who have a internet access but can not pay for remote sensing software, there are about 2.8 million small farmers that live in their plots and around 47% of the rural population in México has internet access so we could estimate at least 1.3 million possible beneficiaries in México. For now we started working in our home state Tamaulipas, we have access to the farmers organizations that group local producers where we can promote the user app and get feedback from the users to improve it, we expect the solution to really address their needs after some iteration.

On one hand we provide knowledge to farmers in form of maps that contain useful information for them, on the other hand the cheap monitoring tools for financing stakeholders (insurers, creditors, government aid) can enable them to improve their proceses (claim validation, yield/risk estimation) which translates in a more efficient investment of their budget allowing them to make their products more accesible, we look to connect both ends through the platform to make access to financing easier as information is more available, this is useful as most of farms are informal and do not leave records.

Remote sensing apps for farmers are built around business models that expect the farmer to pay for the value of the technology (climate.com, farmersedge.ca, ceresimaging.net), the main innovation consists of transferring that cost somewhere else, which means reutilizing the developed software to create a new business model for financing stakeholders that is designed to be just profitable enough to sustain development of the farmer side app. From the point of view of the financing institutions they get the opportunity to reduce their current investment in Land Parcel Identification Systems (https://www.sinergise.com/) with low cost alternatives. The solution is unique compared to others mainly because it aims at low income users, which are most of them.

Core technology is remote sensing, satellites provide the raw observations which are processed to obtain useful intelligence, a simple computation of the band values allows to generate an index of potential yield, using more advanced techniques such as Machine Learning we can do better modeling of these indexes. By packing this information in a mobile app we can make it accessible for the farmer.

Commercial farm management apps already do this although they target a higher-end market and can supply themselves with commercial satellite observations, in contrast we use public data which limits the temporal and spatial resolution of the products but is free. Public observations from missions as Sentinel-2 and Landsat-8 are also widely used by institutions to model agricultural variables. Basic indicators like the Normalized Difference Vegetation Index are easily computed and already proven to be useful to model expected crop yield, more complex processes to model yield using Machine Learning have been proven to work in the US academically (https://www.aaai.org/ocs/index...) we are looking to generate the required data to replicate these algorithms locally. We actually replicated this paper in México, but while MODIS remote data is the same globally, there is not nearly as much historic and reliable field data like the USDA's NASS provides.

First we create a basic remote sensing app for farmers that can be accessed for free and make it available in Tamaulipas, they start using it and giving us feedback on how to make the app more useful for them and we iterate the product with this knowledge.

Having direct contact with the farmers we ask them to help us create new features for the app, such as letting us install a weather station in some of their plots and gathering field data as yield samples, with this data we create better localized yield estimation models. Continuously we ask them to let us sample data from them and we will apply current research to develop helpful remote indexes for the app.

The ability of making accurate yield estimations and other automated remote verifications is useful for financing institutions, estimating yield allows them to model insurance risk and credit interest rates, automating claim verifications allows them to act faster and reduce operational costs. Reusing our software we provide this information as a service for 1 MXN per hectare per month with no minimum amount of hectares/month, this is less than 1% of the government subsidy for insurance.

Small insurance funds already pay consultants to help them reduce the amount of field work (operational costs), we let them try our service, if the product is good they will convert into paying customers, this will provide funds to reinvest, switching between expanding to other States and investing in R&D to improve the app. 

Institutions having better means of making field measurements will reduce their operational costs which the interest rates depend on, making their products more accessible for more farmers.

We keep reinvesting everything in R&D to make better data products and scaling to more regions, finally we connect both ends of the system, the farmer’s app and the institution monitoring tools creating a sort of Credit Karma for farms, using remote sensing data to estimate credit scores and insurance risks, while using the app to interact with the farmer.

Currently we are still developing so we are serving zero people, we expect to serve at least 5,000 producers from Tamaulipas, this will be achieved by accessing them through local organizations and showing what the technology can do.

Our rough estimates for the total potential beneficiaries in México is 1.3 million individual farmers, since the number of farmers with 5 hectares or less is around 2.8 million and the percentage of individuals in rural areas with access to internet is 47%.

In five years we want to serve at least 600,000 producers which account for about half of the estimated potential beneficiaries in México.

First year

First impact goal is to make remote sensing technology available to local farmers in Tamaulipas, we do this by removing the cost barrier and making it free and talking to them to see what works best.

Second goal make this information more useful for the farmer, we achieve this by using the help of the farmers themselves to build a dataset that will allow us to make better information products for them.

Third goal sell low cost Land Parcel Identification systems to small financing stakeholders, will achieve this by having a very accessible price.

Five years

Have 600,000 producers using our app, half of estimated total potential beneficiaries in MX, we seek to achieve this by reinvesting our profits into scaling to more regions and improving the app.

1. There is a financial threat of not being able to keep funding the system. Also we do not have any money to invest in R&D, it's just three of us voluntarily working on it part-time since this is not yet profitable.

2. Another threat could be that farmers do not understand the information as we are presenting it to them, or it is not really useful, or they simply do not want to use the app.

3. A technical threat would be the unavailability of the remote sensing data sources such as Sentinel-2 although this is unlikely.

4. In the case of the agro financing market, a barrier could be existing LPIS software available in use by the financing stakeholders, also consultants already providing their services.

5. There is also a corruption barrier, corrupt people make money off inefficiencies in the financing system and will work against something that threatens their income.

1. For now we have cloud credits from an AWS grant that allow us to keep developing and running the system, also for our state these costs are really low and we could pocket them. With time we seek to make our business model profitable to keep funding us.

2. Working closely with them, listening, having a good UI/UX designer, personally promoting and training the farmers on how to use app.Also they have already expressed their interest in getting something similar.

3. Not much of a barrier for now but a threat we need to consider.

4. By targeting small funds which account for the largest share of institutional agro financing in MX and which we have seen the kind of data products they get from consultants and we know we can give them a better product for lower cost by taking advantage of recent technology developments and automation.

5. Mainly hoping that in challenging times, as the need for food grows, we learn to value more our collective well being, rather than our self. Also maybe throwing a blockchain somewhere along the way.

We are three founders who work part-time.

Dr. Wilver Salinas is the Operational Manager of the Instituto de Ingeniería y Ciencias of the Universidad Autónoma de Tamaulipas, he was formed as an agricultural engineer and then moved to GIS, writing his first article more than 20 years ago which at the time convinced the University to purchase Landsat imagery to solve some of the State surveying needs. Since then the IIyC was born as a business unit to do geospatial projects needed by the State of Tamaulipas.

Dr. Cutberto Paredes also works at IIyC with Dr.Salinas as the main technical advisor on GIS matters.

Alan Salinas has 5 years of experience as a software engineer and his role is to code everything that needs to be coded, has worked before with the IIyC in geospatial projects.

Instituto de Ingeniería y Ciencias from the Universidad Autónoma de Tamaulipas, they are helping with technical GIS advice.

Unión Agrícola Regional del Norte de Tamaulipas, this organizations groups all producers from the north of the State and we have reached out to them to present our idea and they are willing to use the app.

There are beneficiaries (users) and clients. The users are farmers who get a free remote sensing app, there are also users of a Land Parcel Identification System which are financial institutions, and they are the clients.

Product is sold in form of Software as a Service in units of hectare/month for the LPIS, initially the cost will be 1 MXN per unit, no minimum units needed for purchase. Distribution of information is done through a web dashboard. They need it because the largest share of institutional financing in México are small funds (23%) and they have less operational resources so low cost LPIS with a flexible purchasing scheme (per hectare/month) becomes attractive.