Yield Forecasting and Damage Assessment of Paddy to Introduce Crop Insurance in Flash Flood Affected Regional Areas Using Satellite Remote Sensing and Econometric Approach

概要

The unexpected flash flood occurrence and rainfall uncertainty are the most challenging global issues to achieve sustainable development where wetland communities are the key sufferers. In this regard, an initiative was taken through yield forecasting and damage assessment of paddy to introduce crop insurance in flash flood-affected wetland areas using satellite remote sensing and econometric approaches. Bangladesh, which is one of the most climatically vulnerable, as measured by flash floods, floods, droughts, cyclones was selected for this study. Since the overall goal of this study was to establish sustainable livelihoods in flash flood-prone areas, we have classified our research into three specific objectives.

In the first part of our research, an effort has been made to restrict flash flood vulnerability mapping and need assessment based on vulnerability classification at Haor regions in Bangladesh using satellite remote sensing, geographic information system (GIS), and econometric models. At first, flash flood vulnerability mapping was performed using a bivariate statistical regression-based frequency ratio (FR) model. The outputs of vulnerability mapping were classified into very high (7.57%), high (22.56%), moderate (56.96%), low (12.34%), and very low (0.56%), covering areas of 1303.0, 3882.0, 9801.0, 2124.0 and 96.0 km2, respectively. Second, 40 sampled respondents from each vulnerable groups were interviewed to assess their vulnerability and coping strategies against flash floods using field survey data through the composite livelihood vulnerability index (CLVI). CLVI findings showed that the vulnerability differences among the five groups of respondents were relevant to vulnerability mapping and classification done in the remote sensing and GIS platform. Besides, the logit estimates explained that the age of the household head, household size, farming experience, educational status, occupation of the household head, farm size, proximity to the marketplace, and number of earning members affect farmers' attitudes toward coping strategies at different levels significantly.

However, agricultural forecasting getting importance due to its adverse climate change and to ensure food security throughout the seasons. In this context, we attempted to develop rice yield prediction models utilizing satellite remote sensing-based vegetation indices at the optimum harvesting period before flash flooding in the second part of our research. Several empirical yield prediction models for rice production were developed using five relevant vegetation indices: the normalized difference vegetation index (NDVI), normalized difference water index (NDWI), rice growth vegetation index (RGVI), moisture stress index (MSI), and leaf area index (LAI). Both parametric (simple and complex) and nonparametric (artificial neural network, ANN) regression studies were used to validate the obtained models. In simple regression studies, the best estimated results for the RGVI (R2 = 0.44), NDVI (R2=0.63), NDVI (R2 = 0.55), and NDVI (R2 = 0.67) were observed for 2017, 2018, and 2019, as well as the average of seasons from 2017 to 2019. The best-performing vegetation indices for constructing Boro rice yield prediction models using multiple regression were the composite NDVI- RGVI (R2 = 0.65), NDVI-NDWI (R2 = 0.56), and NDVI-MSI (R2 =0.69) indices. However, utilizing a simple regression technique (R2 = 0.84) and multiple regression analysis (R2 = 0.91) of the average NDVI-MSI composite index, NDVI showed superior accuracy in the ANN-based machine-learning outcomes for the average Boro rice season (2017-2019). As a result, Boro rice maturity yield prediction models can be effective for farm risk management, insurance premium assessments, and related stakeholder decision-making to limit the effects of intense flash flood events.

In addition, for any flash flood prone vulnerable countries, establishing appropriate risk management strategies is the topmost concern. Crop insurance, as an innovative risk coping strategy, can thus help poor farmers in emerging nations deal with weather-related output risks. In this regard, we attempted to develop a new and unique damage-based crop insurance system for flash flood-affected wetland areas by integrating GIS, satellite remote sensing, and econometric models simultaneously. A NDWI followed by the decision rule approach was employed to generate an accurate damage map of Boro rice from Landsat-8 Operational Land Imager (OLI) data, and the inundated area was delineated from the very high-resolution land use land cover (LULC) map of the Survey of Bangladesh (SoB). For determining minimum coverage levels and premium amount, expected break-even yield for indemnity and the future value of annuity modeling were applied. Findings of coverage levels suggested that until the 70% coverage level, farmers were not eligible to reach the expected break- even yield for indemnity payouts. Moreover, findings of the expected loss and production function analysis showed that the higher the coverage levels were, the lower the insurance premium was, and the lower the damage class was; subsequently, the insurance premium rate was lower. The lowest insurance premium rate was observed for the high coverage and moderately damaged class at $23.82/ha, and the highest rate was seen for marginal coverage and highly damaged areas ($39.49/ha). Finally, a binary logistic regression technique was applied to determine the important determinants influencing farmers' willingness to adopt (WTA) crop insurance, and a multiple regression approach was utilized to assess the amount of insurance premiums they were willing to pay (WTP). Evidence from the regression findings suggested that farmers’ socioeconomic and environmental awareness features were relevant for adopting a damage-based crop insurance system. Identifying the risk of a flash flood in providing farmers with accurate information, developing advanced risk management strategies, and providing agricultural credit and service provision, policymakers and research institutions may benefit from mapping and evaluating their livelihood vulnerability on a single platform.