Background

The Financial Instruments Sector Team (FISTeam) works on processes and tools which aim to help people overcome climate risk through financial tools like index insurance and index-based disaster risk management. Through partners, collaborators, and educational activities, FISTeam is engaging the players necessary to represent the wide array of expertise and perspectives and build community driven solutions.

The FISTeam partnered with the project Adaptation of Agriculture Value Chains to Climate Change (PrAda) in Madagascar, funded by the German Society for International Cooperation (GIZ) to provide support in Climate Risk Insurance Development.

The PrAda team and local partners among others the Ministry of Agriculture and Livestock (MINAE) and the National Department of Meteorology (DGM) conducted a detailed analysis for selected agricultural value chains and demonstrated how the forecasted climate hazards will impact the revenues of smallholder farmers and threaten the livelihoods of rural communities. Adaptation measures, such as access to climate risk insurance products, were identified as relevant for smallholder farmers.

In the Androy region, the peanut value chain was identified as the one with the clearest and strongest vulnerability to drought, thereby allowing the design of a climate risk insurance product. For the 2020-2021 season, an insurance pilot was set up in the Androy region, more specifically in the district of Bekily and Ambovombe, to serve farmers who produce groundnuts. The farmers for the first pilot are located around two communities: Andalatanosy and Ambatosola. Drought is the main risk for these groundnut farmers. To underwrite and design the pilot, the project started a cooperation with the insurance company ARO.

Approach

The initial step to an index co-design process is to conduct preliminary feasibility assessments such as mapping index design potential, gathering site specific information and working with experts to help better understand the need of an insurance product on the ground. This process includes visits to the locations in question to obtain farmer’s agricultural practices and recollected bad years experienced on the ground. Once the initial phase is completed, a pilot index design is co-developed by consulting a variety of rainfall datasets identified in the initial step along with vulnerability data such as farmers’ bad years and expert reporting.

Throughout the pilot season, a series of seasonal monitoring processes are conducted to follow the progress of rainfall and assess the relative historical ranking of the current rainfall distribution in comparison with the farmers’ reported cropping cycle. The monitoring informs stakeholders on the performance of the index and provides insight on ways to improve the design.



The map below shows the locations visited during the initial visit data collection process by the project team (the red dots), as well as the coordinates used for calculating the insurance index in each commune (the blue dots) and the rain gauges installed (the green dots).

Challenges

The 2020 pilot index was designed to calculate the minimum of 30-day running sum of IMERG daily satellite rainfall estimate between December 1st and February 28th. The payout calculation of the index was done by the DGM after a successful skill transfer training. The index did not trigger a payout.

Three types of satellite datasets were assessed by PrAda in order to choose the satellite product that best captures local droughts: rainfall (IMERG and TAMSAT), soil moisture (ESA CCI), and Enhanced Vegetation Index (MODIS Terra EVI). A dense network of point locations is sampled over entire pilot area. Various estimates are calculated and interpolated at each point location. Cross-comparisons are conducted between various estimates at each location. The cross-comparison between different datasets suggests that 2- and 3-month running-sum rainfall estimates are in a higher agreement with Soil Moisture and EVI estimates. However, IMERG and ESA CCI has a higher agreement than that between TAMSAT and ESA CCI.

The 3-month window between December and February were chosen to cover the cropping cycle of groundnut in the pilot region. The planting and flowering dates were collected from farmers using participatory processes in Bekily.

However, the pilot index validation process faced data corroboration challenges which provided valuable lessons learned for the project as it moves forward. These challenges were heightened by the logistical difficulty of collecting valuable information/feedback from farmers and the inability to conduct capacity building trainings in person to the local met office due to the COVID-19 pandemic. As a partner with an advisory role, the FISTeam worked with the DGM and other local partners to breakdown the validation process of the pilot index using tools and workflow processes tailored for the project.

IMERG: https://gpm.nasa.gov/data/imerg

TAMSAT: https://www.tamsat.org.uk/

ESA CCI: https://esa-soilmoisture-cci.org/

EVI: https://modis.gsfc.nasa.gov/data/dataprod/mod13.php

Farmer Reported Seasonal Calendar

When designing a rainfall-based insurance index, an understanding of crop types and the agricultural cycle(s) is very important in identifying critical growth phases and the risks of an index’s failure to perform. In order to verify the impact of below-average rainfall during these critical periods, FISTeam provided a data comparison tool that allows decision-makers to adjust the specific critical timings of drought to hone in on the risk of crop failure.

An important piece of information is how farmers experience the season. This data can be used to align the insurance index window(s) with key seasonal events, like planting and tasseling.

The following charts summarize the seasonal timing information that farmers from each commune provided.

The data provided shows that farmers in the pilot locations tend to plant in November as the average rainfall onset starts around end of October. The index as indicated above covers this period. However, the monitoring of the critical period of plating time is essential in isolating the risk of planting failure.

Reporting of the pilot season experienced by farmers

During the pilot season of 2020/2021, farmers answered keys questions used in the validation process:

  • When did you plant? 2021-02-18

  • When did the rainfall start? 2021-01-17

  • Was there any damage to your crop because of rainfall shortage? yes

  • How does this year compare to the previous year? slightly better

  • How does the rainfall this year compare to the past five years? much worse

Using farmers’ reporting of the season, the onset on the rain as well as the planting time is relatively late (50 days late) compared to the average recorded during the initial visit. It has also been reported that according to farmers, the 2020/2021 season was relatively below average, yet 2019/2020 was reported as a worse year.

What happened at the beginning of the season?

In order to assess the ranking of the early phase, late phase, and the seasonal average of the 2020/2021 season, we compare the cumulative rainfall of this year to history. Below are three graphics comparing the historical rainfall using all available data of IMERG, ARC2, and CHIRPS which represent three different rainfall satellite products available at a daily temporal resolution. Both ARC2 and CHIRPS are rainfall estimate satellite products which are systematically calibrated using rain gauge data.

In the figures above, early_precip refers to the November-December cumulative rainfall, late_precip to the January-February cumulative rainfall, and seasonal_precip to the entire season (Nov-Feb).

All three sources of satellite data show that the cumulative November and December rainfall for the 2020/2021 season was average to below-average. When assessing end of season conditions, the satellite sources show average to above average conditions.

Expert verification

In order to verify the satellite and farmers data with additional information that can provide further validation to the analysis, we can consider expert information about the season to corroborate the index. One common source is the FEWSNet Integrated Phase Classification (IPC). The IPC Score, issued in October of each year going back to 2016 in Madagascar, is a way for us to evaluate how closely food insecurity assessments relate to significant droughts.

While IPC provides a district-level scale of comparison which doesn’t allow us to hone in on the localized commune conditions, it does nevertheless provide us with an overview of the level of food insecurity of the region. This level of assessment can flag major historical droughts that affect larger regions. With an average IPC score between 2 (stressed) and 3 (crisis), out of a maximum of 5 (famine), the 2020/2021 season in southern Madagascar appears to be as bad as any year since 2016.

Interative validation

An interactive tool has been developed as part of the PrAda project to support the capacity building of DGM to design and monitor weather-based indexes. The tool was used as a training platform and was also made available to DGM and local partners to access and compare appropriate satellite data sources as well as farmer information for Index Insurance design and validation co-identified with local stakeholders. Users can use the tool to

  • Quantitatively compare the level of historical matching between all available satellite and on-the-ground sources of information.
  • Define comparison parameters by choosing the seasonal phases (windows) to include in the calculations, and the calculation method (rolling or simple averaging of data)
  • Visualize the historical bad years of each data source.
link to to tool

Conclusion

Piloting an index insurance product is often accompanied with the bottleneck challenges to be solved through Index Insurance design.

With the availability of a wide variety of remote sensing technologies that are becoming more precise and sophisticated, index insurance products face the challenge of informed application of the modern technology to produce better performing products.Performing products which best capture the reality on the ground require an iterative process of co-design and validation. While the choice of a satellite data source is important for providing accurate estimates of seasonal rainfall, the workflow analysis performed by IRI in collaboration with the DGM show that a close examination of key phases of the groundnut cropping cycle such as the planting and flowering periods, and the use of farmers’ feedback can highlight areas of potential design improvements. In this case, the index validation process provided valuable information about the weight of early seasonal rainfall in farmers’ yield outcomes. While the available crowdsourced information is showcasing an important added value in design improvement, additional data from farmers that is systematically collected and integrated in index design is needed.

When the index co-design follows a collaborative approach that promotes the participation of all stakeholders at all stages of the project, these challenges can be solved iteratively by making sure:

  • Farmers’ information is integrated in the design and validation

  • A convergence of evidence process is followed by a thourough comparison of different data sources, complemented by local experts’ intuition.

The PrAda pilot serves as an example of the ways a co-owned due diligence approach can provide insight on concrete next steps of index insurance refinement and improvement.