Remote Sensing Based Rapid Assessment of Flood Crop Damage



Journal Title

Journal ISSN

Volume Title



Accurate crop-specific damage assessment immediately after flood events is crucial for grain pricing, food policy, and agricultural trade. However, most policy level statements on flood crop loss usually come as a gross estimation such as inundated cropland acreages without the consideration of crop types and the degree of damage. Therefore, it is crucial to bring the right information to the right people at the right time, and this is where remote sensing became important. The main goal of this research is to bring the crop-specific damage information immediately after flood events using satellite earth observation data. A newly developed Disaster Vegetation Damage Index (DVDI) is utilized to assess flood impacts on crop conditions. By incorporating the DVDI index along with the information of crop types and flood inundation extents, this research assessed crop damage for three case-study events. Crop damage are assessed in a qualitative scale and reported at county level. The results of damage assessment are validated through NDVI profile and yield loss in percentage. A linear relation is found between DVDI values and crop yield loss. Results also indicate the association between DVDI class and crop yield loss. Cropland Data Layers (CDL) is available for US croplands many months after the growing season. Therefore, the in-season crop-specific rapid flood damage estimate is not possible without the information of crop types. Thus, this research fills the data gap by providing the methodological frameworks for the identification and prediction of in-season major crop types. Trusted pixels are extracted from historical CDL data using crop rotation patterns. These trusted pixels are used to train supervised classification models to map major crop types using in-season Landsat images. The overall accuracies of single date multi-band image classification are 84%, 89%, and 92% for May, June, and July, respectively. The result also shows higher accuracy (94–95%) can be achieved through multi-date image classification compared to single date image classification. Remote sensing-based crop mapping can be affected by cloud contamination. Thus, this study also presents an alternative approach to predict crop types using a machine learning approach to predict crop types using historical cropping patterns. Results show crop type can be predicted before the crop growing season with 85% accuracy. Flood inundation extent is crucial for the rapid assessment of flood crop damage. This study utilized some options for flood inundation mapping using popular remote sensing data such as Landsat, Sentinel 1, Sentinel 2, and Soil Moisture Active Passive (SMAP). Although the primary challenge is the availability of remote sensing data, this study found that useful remote sensing data either from optical or microwave systems can be available for flood mapping in most of the cases. This study also explores flood mapping options using SMAP-soil-moisture data and Federal Emergency Management Agency (FEMA)’s national flood hazard information. Flood extents are mapped by comparing pre- and post-event water information derived from remote sensing data. Since this damage assessment relies on crop mapping, flood mapping, and crop condition-profile assessment, errors from each step can contribute to the final evaluation. Despite having some limitations and shortcomings, the outcome of this research can significantly contribute to the process of the rapid assessment of flood crop damage.