Near Real Time Crop Loss Estimation using Remote Sensing Observations

Natural calamities triggered by erratic weather conditions like cyclone, earthquakes, hail storms, and flood incurs substantial loss to the infrastructure and crops of the region. Countries across the globe are prone to such natural calamities. In India, specifically coastal parts are vulnerable to tropical cyclones. In 2018 east coast districts of Tamil Nadu and Andhra Pradesh, India were affected by the three cyclones namely Titli (11 Oct. 2018), Gaja (16 Nov. 2018) and Pethai (17 Dec. 2018) causing severe damage to seasonal crops such as Rice, Coconut and Areca Nut plantations. Traditional survey-based methods of crop loss assessment are time-consuming and labor-intensive.This study addresses the problem of near-real-time qualitative crop loss assessment due to tropical Gaja cyclone using the temporal data from Sentinel 1 and 2 satellites. The crop damage assessment study has been undertaken for Gaja cyclone in the affected district of Thanjavur, Tamil Nadu, India. The major crops cultivated in the district are Kharif Rice (locally called as Samba and Late Samba) and Coconut plantations. The study addresses qualitative loss assessment in terms of crop area affected. As a first step, we used time series data of Sentinel1 (VV and VH backscatter) available between Aug.-Nov. 2018 to map the Kharif rice area. Also, cloud-free Sentinel 2 scenes available during Mar.-May. 2018 have been used to map the Coconut area. Field visits were conducted to collect the geo-tagged plot boundaries for the rice crop and coconut plantations. The data collected through field visits was used both for model training and crop loss assessment. Google maps satellite layer was used as a base map for identification of other non-crop classes (i.e., forest, water, settlement, etc.). The overall accuracy of crop area classification was 87.23% for rice and 92.22% for coconut.Further, to estimate the crop loss, crop layers along with the NDVI were considered. Two crop loss scenarios, namely minimum damage and maximum damage, were identified for both the crops. The mean NDVI composite before the event, i.e., 1-15 Nov. 2018 was considered as the base. In case of maximum loss scenario, short term NDVI composite available immediately after the event, i.e., 17-25 Nov. 2018 was selected. After the cyclone, long term NDVI composite of the mean (i.e., 17 Nov.13 Dec. 2018) was used to assess the minimum loss scenario. Using field observations, the crop loss was categorized as severe loss, medium loss, low loss, and no loss. Results showed that the coconut plantations in Pattukkottai, Peravurani, and Papanasam blocks of Tanjavur are affected by the cyclone. The significant rice crop loss has been observed in Thanjavur, Orattanadu, Pattukkottai blocks. We have found the remote sensing based crop loss observations are matching with the government reports based on field observations. The remote sensing observations with human participatory sensing (i.e., field observations) has the potential for near-real-time crop loss assessment.