Development of A Satellite-based Near-Real-Time Precipitation Product for Index-based (Re)Insurance Applications in Central America

This paper presents an ongoing research work undertaken by MiCRO (Microinsurance Catastrophe Risk Organisation), focusing upon the development of a reliable satellite-based rainfall product for the near-real-time (NRT) index-based excess rainfall (XSR) calculation over the pilot countries in Central America. MiCRO (https://www.microrisk.org/) is a reinsurance company that specializes in providing natural catastrophe protection for the vulnerable and low-income segments of the population by designing and operating holistic risk management strategies that use index-based insurance products. MiCRO’s efforts to expand traditional insurability boundaries largely depend upon having access to indices like precipitation estimates that are open-access, quantitatively and operationally reliable, and widely accepted (used) by meteorological and hydrological communities, and by risk takers, mainly insurers and reinsurers.

The ongoing research work comprises two stages: 1) identification of a suitable satellite precipitation product, and 2) exploring the possibility of using satellite-gauge data merging techniques to improve the applicability of the original satellite-only precipitation products. The requirements of satellite precipitation products for MiCRO’s NRT applications are stringent. In addition to accuracy, the methodological and statistical consistency and short processing latency (within 1 day) are equally critical. In the first stage of the work, a review was first conducted of widely-used satellite products that are capable of NRT operation; based upon it, two satellite products (i.e. IMERG NRT Late Run and GSMaP NRT) were selected and further compared against ground rain gauge records and TRMM 3B42RT product (i.e. the current data source used by MiCRO), in terms of accuracy and XSR payout calculation. The comparison results suggest that IMERG NRT Late Run generally has the highest consistency with ground rain gauge data; however, it tends to underestimate extreme rainfall values and consequently XSR payouts during the period of 2014-2017. As compared to IMERG NRT Late Run, both GSMaP NRT and TRMM 3B42RT show a better ability to produce high rainfall values during the period of 2014-2017; however, their consistency with ground rain gauge records is significantly lower. In addition, both IMERG and GSMaP appear to have issues regarding statistical consistency. More specifically, it was observed that both of their statistical features have changed since 2018 over our pilot countries.

In the second stage, the possibility of using satellite-gauge data merging techniques to improve the accuracy and consistency of satellite products was explored. It is worth mentioning that local ground rain gauge records are supposed to be the most suitable data source to improve satellite products. At this moment, the real-time access to ground rain gauge data is however infeasible in our pilot countries, so the CPC Global Daily Unified Gauge-Based Analysis of Precipitation product was used as additional data source to be merged with satellite products. A total of three different data merging techniques were employed for the comparison; these include mean field bias correction (MFB), local bias correction (LB) and kriging with external drift (KED) methods. In addition, some sensitivity analyses were conducted to identify the key factors that affect the quality of the final merged precipitation products. The results suggest that LB and KED products generally outperform MFB and the original satellite products in terms of both standard and extremal statistical properties. Regarding the key factor for merging performance, the consistency between satellite and gauge-based data is found to be the most critical one. In light of this, IMERG NRT Late Run appears to be the most suitable satellite product for merging applications for our pilot countries; and, for operation, it is recommended to conduct data merging at 7-day timescales, so the consistency between satellite and gauge-based data can be effectively improved and at the same time the short-term rainfall dynamics can be sufficiently preserved.