Image for post
Image for post
NDVI Anomaly Detection script applied to the area South of Leipzig, Germany. The goal of the script is to calculate the “anomaly” of an index, avoiding clouds. See the multi-temporal custom script description on the Sentinel Hub GitHub repository for more information.

Sentinel Hub supporting IBISA in the Crop Micro-Insurance Field

Sentinel Hub API used by IBISA to develop a service of crop protection for smallholder farmers in developing countries

Jul 2, 2019 · 6 min read

A guest blog post by Jean-Baptiste Pleynet

Image for post
Image for post

Overcoming the Challenge

The challenge of the project is to offer a very low-cost crop risk sharing (close to micro-insurance). For example, the risk premium hypotheses in the business plan of IBISA are around 10$ to 20$ per year, of which the IBISA service takes only 20% fee, as compared to the usually observed micro-insurance ratio ranging from 50% to 80%. By consequence IBISA costs has to be limited to 2$ to 3$ per year to manage each policy, including claim assessment. Only by doing so, IBISA can bring a sustainable service to more than 500 millions smallholder farmers worldwide that today don’t contract micro-insurance because it doesn’t bring enough value for money to them.

To satisfy this challenge of low-cost operations, IBISA combines the use of blockchain and Earth Observation (EO) data. Assessing damages using EO avoids the need for ground truth to do field assessments, even if this introduces other problems, well known in parametric insurance.

Image for post
Image for post

Thus, IBISA project needs EO data, in an easy-to-use presentation.

Finding the Right Tools

It has been possible since long to access the Copernicus data service and download processed EO satellite images. One can then store locally the data and use open source tools such as QGIS to analyse them. Although these tools are relatively stable and easy to use, the learning curve is steep and one needs a lot of local storage and computing power.

We found everything needed in the Sentinel Hub, a cloud API for satellite imagery providing the access and manipulation of EO data. It eliminates the major hassle of downloading, archiving and processing petabytes of data and simply makes the full global archive available via web services. Application developers can therefore focus on value-adding services and end-user applications rather than dealing with the complexity of remote sensing data. Both free web applications, Sentinel Playground and EO Browser enable the user to interactively test scripts on an image from different data sources.

Because of the quality and reactivity of support from the Sinergise team, IBISA started to use Sentinel Playground to explore how to practically use EO data for the needs of IBISA.

IBISA’s Need for EO Data

IBISA has two main needs for EO data:

  • elaborating a risk model based on time series to spread the risk equitably among all farmers in a given region and
  • asses month after month the loss observed for the members.

Risk modelling

The first IBISA usage of the EO data is to do risk modelling. By risk modelling, as with classical insurance, we mean calculating individual contributions, and the minimum amount of treasury required to be kept in the global system to insure the solvency of the system.

Considering that EO data are used to assess loss, we naturally need also those EO data to build the calculation of individual contributions in order to match the potential loss. In this case, we more need historical data, or time series, rather than individual images.

The loss assessment is based on one (or multiple) indexes, observable from space. The most classical one is the Normalized Difference Vegetation Index (NDVI). It is an indicator of the greenness of the biomass at a given geolocation (at a pixel). Even though it is not a physical property of the vegetation cover, its very simple formulation makes it widely used for ecosystems monitoring:

where and are the spectral reflectances measured in the near infrared and red wavebands respectively.

We use the Sentinel Hub Statistical Information API, which allows us to request figures on and index (like NDVI) in a zone over a long period of time.

Then, we can extract NDVI time series, which are the historical value of the NDVI in different locations over the time.

Here for example for 170 different plots in India:

Image for post
Image for post

As we did previously, we can then compute the NDVI Anomaly, and then have NDVI Anomaly time series.

The following graph shows the NDVI Anomaly time series of the same 170 plots in India:

Image for post
Image for post

These time series are the first step of a more complex data analysis in order to calculate contribution or solvency.

Introducing the Watchers Platform

Watchers are human volunteers doing “crowd-watching” regularly to assess loss using free EO data and any kind of data that are freely available on the Internet. In exchange of this work and depending on their reliability, watchers receive a remuneration from part of the fees levied on the contributions of farmers.

Periodically (monthly), the IBISA platform will trigger questions to watchers to assess EO images, where are represented groups of plots of land. Those groups are plots of land belonging to farmers, at a village scale (some kilometre square) and each group will be assessed as one single unit.

Thus, the watchers interface displays the different groups to assess, like the following ones:

Image for post
Image for post
Image for post
Image for post
A group of plots in true color (left) and in NDVI (right).

To assess the state of vegetation as compared to previous years at the same date, the IBISA watcher needs to assess Anomaly images. The anomaly is the current Index (NDVI) compared to the average of the same index, in the same period, in past years (3 last years for example).

Immediately, you can understand that if the NDVI Anomaly is greater than 0, then the vegetation is better than it is usually at the same period. The opposite if this is lower than 0.

In order to realise this, IBISA developed a custom script, awarded at the Sentinel Hub custom script contest, and that watcher helper tool uses the WMS API of Sentinel Hub platform, and the multi-temporal computation mechanism. The NDVI computation image is done at Sentinel Hub side, and a large-scale image is computed.

The IBISA watcher helper tool just has then to draw squares on the screen and ask for answers from the human watcher.

Image for post
Image for post
The Sentinel Hub provides a computation where pixels with positive NDVI anomaly are in a green scale and pixels with negative values are in red scale.

IBISA is highly dependent of Earth Observation data, both for designing the product and for running it. Most satellite data are provided for free by space agencies like ESA, but those raw data need to be processed before being utilised. This need was the beginning of partnership between Sinergise and IBISA, as the use of the Sentinel Hub API allows external systems (like watchers’ helper tool or risk model) to connect and to request EO data in an understandable format, and at the end give life to IBISA.

Sentinel Hub Blog

Stories from the next generation satellite imagery platform

Medium is an open platform where 170 million readers come to find insightful and dynamic thinking. Here, expert and undiscovered voices alike dive into the heart of any topic and bring new ideas to the surface. Learn more

Follow the writers, publications, and topics that matter to you, and you’ll see them on your homepage and in your inbox. Explore

If you have a story to tell, knowledge to share, or a perspective to offer — welcome home. It’s easy and free to post your thinking on any topic. Write on Medium

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store