Evaluating the 22 June 2017 South Dakota Hail Damage Using KHawk UAS: Accuracy Quantification and Cross Validation with Satellite Imagery

Natural disasters, including tornado, hail, wildfire, etc., can cause severe damages to lives, property, and agriculture. With the increasing threat of climate change, real time situation awareness during or right after the disaster is very important. In 2017, USA suffered a loss of at least $306 billion due to natural disasters. It has been estimated that the 2017 natural disasters costed American agriculture over $5 billion, due to destruction caused on fertile regions including beef cattle ranches in Texas and cotton and rice farms in Louisiana. Efforts are being made by various researchers to develop robust damage assessment techniques in preparation for future disasters.

Unmanned aircraft systems (UASs) have become extremely popular over the last decade for post-disaster damage assessment. UASs can be launched with a short notice and flown over dangerous and severely damaged fields for aerial image/video acquisition. The advent of high resolution multispectral or hyperspectral cameras have made it possible to use UASs to quantify crop damage and land variations using indices like normalized difference vegetation index (NDVI). Unmanned sea vehicles (USVs) and UASs have been used to inspect the structural damages caused by Hurricane Katrina, Wilma and Ike. Multiple UASs and wireless sensor networks (WSNs) were used as a part of the AWARE project to validate its use in disaster management. In addition, UASs have been recently used for tornado damage tracking and crop damage quantification using multispectral aerial imagery. However, to ensure the quality of these images and resulting orthomaps, a comprehensive error analysis and cross validation with existing data, such as, satellite imagery are needed.

This paper focuses on the use of KHawk unmanned aircraft systems (UASs) for the evaluation and quantification of the 60 mile scar damage generated by the 22nd June, 2017 hail storm in South Dakota area. Equipped with high resolution RGB and near-infrared (NIR) cameras, the KHawk UASs are deployed over damaged fields for aerial video acquisition. A total of six flights were performed to cover a total area of about 6 square miles on July 23-24, 2017. Collected data is processed to produce high quality RGB and multispectral orthomaps. The emphasis of the analysis will be put on UAS data error quantification (2D ortho-map, multispectral images and NDVI calculations) and cross validation with satellite products. A comprehensive 2D error analysis will be conducted to validate the location accuracy of the UAS map using landmark feature points from KHawk aerial map and satellite images. For the multispectral image analysis, the KHawk surface reflectance calibration procedure will be used to compare UAS maps with satellite images (Landsat 8 taken on July 7 and August 8, Sentinel 2A and 2B, and MODIS from Aqua). An initial qualitative analysis will be performed to validate the NDVI trends followed by a comprehensive pixel-based error analysis. Further processing will be implemented based on the resolution difference to make a fair comparison between UAS and satellite images.

The main contributions of this paper include the development of protocol for calibration and deployment of KHawk UASs for disaster damage assessment, comparison and error quantification of KHawk NDVI results with NASA satellite products, and the collection and sharing of a representative hail damage data set with collaborators such as National Weather Service (NWS), and other federal and state emergency management teams.