The COVID-19 pandemic affected food supply for a number of reasons. An obvious reason is the interruption of a large part of road transport. A less obvious, but as much significant, is the drop in demand. Indeed, the imposed closure of restaurants and the recourse to dry food in order to going out as little as possible during this period forced wholesalers to reduce their purchases of crops from farmers. The latter, caught off guard, found themselves with a large quantity of groceries on hand, such as leeks, strawberries or lettuce. In addition, with much less labour to harvest, many fields are filled and yet, opposite, famine is still present. The fate of all this food is not a surprise, they will end up in the garbage when they would have made the happiness of some unfortunate people.
Very often, the farmer would have been delighted to be able to donate them instead of throwing them away, but he does not know how to proceed. He also does not want to waste too much time, since farming is his livelihood, he must at all costs find a way to compensate for his financial loss by starting another plantation for example.
The ambition of the Asparagus Mowers Association (AMA) is to notify charities of potential fields with available produce. The application uses satellite images from all over the world and identifies crop plots that have not been harvested abnormally at any given time. For this purpose, a Deep Learning model first identifies the type of crop. We can then cross-reference this with information on its usual harvest period. Last, using an advanced indicator that we built upon satellite-based NDVI time-series - the NDV2I - we automatically translate specific patterns into the identification of abnormally full field at any given time. Thanks to the geolocation of the sites, the charity can go directly to the producer who may or may not agree to join the programme. In the positive case, volunteers from the association will be able to harvest the food and finally redistribute it to people in need.
With AMA, we make life much easier for donors, which is one of the main constraints. But also those of the volunteers who will know exactly which farmers are potential donors. The strength of the application is that it will not only be useful during this pandemic. Indeed, food wastage among agricultural producers is very common. Industries refuse to buy “ugly” produce, of the wrong size (too small or too big). And yet, they are of the same quality as others. Thanks to AMA, the application will be able to direct any charity to producers who are likely to throw them away otherwise. We are convinced that with this win-win system, we will take a step forward in solving world famine.
After a two-months lockdown, finally allowed to leave our home, we decided to spend some days in the countryside. Talking with a local asparagus producer, we realized the disaster he was experiencing: many fields were no longer harvested and its production was destined for the garbage. A lack of available workers and the strong current inclination for sustainable products like rice & pasta rather than fruits & vegetables were condemning his production.
While we really like asparagus, we also thought that this issue was probably far more important, both in terms of locations and impacted crops. This decided us to dig into this topic in order to to better understand the Covid19 consequences for farmers and the implications for global food security. Good guess: the reality was far worse than what we imagine. One figure: if nothing is done, 300M people in the world will suffer from food insecurity by the end of 2020, that's twice as many as in 2019. How can we let so many people in the world starve to death when at the same time, agricultural production of fruit and vegetables is destined to become a historic waste?
SpaceApp seemed to be the perfect opportunity to have our say in the endeavour to tackle this issue. Indeed, what tool could be more appropriate to such a global, information based issue than remote sensing?
This is how we have imagined AMA - Asparagus Mowers Association - an application using satellite imagery to detect fields not harvested during Covid19 and to put farmers & humanitarian associations in contact to distribute food to the most deprived.
Once this goal was defined, we tried to identify appropriate tools to ensure:
It appeared quickly that Sentinel 2 data, thanks to the constellation global coverage, its relatively frequent revisit rate, the available bands and their easy access was a good way to start. Furthermore, this was the most appropriate tool to cross with data from European countries we are used to. Of course, we are convinced, that the solution could be easily extended to use Landsat8 imagery for instance.
The natural tool to access such data, prototype solutions and think in terms of scalability is Google Earth Engine. With such a cloud-based GIS, we did not need any more hardware. As a team with a strong data science component, also not afraid of minimalist documentation, we opted for the Python API. Last, we strongly relied on Jupyter for quick iteration on the code.
Ok, our issues were with the minimalist documentation GEE. We also had to navigate in the heterogeneous ecosystem of open data providing crop information. We had two main achievements. On the one hand, the identification of both the strong COVID19 effect at the national level for some crops and the creation of a indicator to automatically identifies them. And on the other hand, to have enjoyed this wonderful hackathon !
Data:
Other resources: