Agriculture flood mapping with Soil Moisture Active Passive (SMAP) data: A case of 2016 Louisiana flood

Md Shahinoor Rahman,Liping Di,Junmei Tang,Ranjay Shrestha,Eugene G Yu,Li Lin,Lei Hu,Chen Zhang,Zhengwei Yang

doi:10.1109/agro-geoinformatics.2017.8047062

Abstract

Agriculture is one of the most affected sectors by the flood. Spaceborn remote sensing is widely used for flood mapping and monitoring in recent decades. Some applications such as flood crop loss assessment require data with fine temporal resolution to monitor short-lived flood. MODIS is providing remote sensing data with 1–2 days temporal resolution which has frequently been used for flood mapping for a large area. However, incapability to penetrate through the cloud hindered the application of optical remote sensing in flood mapping in many cases. Thus, radar remote sensing especially synthetic aperture radar (SAR) already shows the capability for the flood mapping in cloud condition. However, monitoring of short-lived flood is not possible using freely available SAR data because of the long revisit capacity of these SAR systems. Therefore, microwave remote sensing with fine temporal resolution might be helpful for flood inundation mapping. Soil Moisture Active Passive (SMAP) is a microwave remote sensing initiative which is providing 3-hourly soil moisture data. Therefore, this study tries to map agriculture flood based on SMAP soil moisture data and soil physical properties. Soil moisture above the filed capacity might be the indication of soil inundation. Moreover, It has been observed that there is an increment in soil moisture during the flood. Therefore, this approach considered three conditions to map the flooded pixel: a minimum of 0.05 increment in soil moisture, a soil moisture threshold 0.40 (moisture above the field capacity) and the 72 consecutive hours. To avoid the random increment in soil moisture a 3-day moving window is applied to the time series data. The flood map extracted from SMAP data is validated with FEMA declared inundated crop land. The overall accuracy is 60% and about 32% of commission error. The over estimation of the flood by SMAP data due to the coarse spatial resolution (9km) of SMAP data.

Full Text