TITLE:
Mapping and Quantifying White Mold in Soybean across South Dakota Using Landsat Images
AUTHORS:
Confiance Mfuka, Xiaoyang Zhang, Emmanuel Byamukama
KEYWORDS:
White Mold, Soybean, Random Forest, Landsat, South Dakota
JOURNAL NAME:
Journal of Geographic Information System,
Vol.11 No.3,
June
18,
2019
ABSTRACT: White Mold of soybeans (Glycine Max), also known as Sclerotinia stem rot (Sclerotinia sclerotiorum), is among the most important fungal diseases that affect soybean yield and represents a recurring annual threat to soybean production in South Dakota. Accurate quantification of white mold in soybean would help understand white mold impact on production; however, this remains a challenge due to a lack of appropriate data at a county and state scales. This study used Landsat images in combination with field-based observations to detect and quantify white mold in the northeastern part of South Dakota. The Random Forest (RF) algorithm was used to classify the soybean and the occurrence of white mold from Landsat images. Results show an estimate of 132 km2, 88 km2, and 190 km2 of white mold extent, representing 31%, 22% and 29% of the total soybean area for Marshall, Codington and Day counties, respectively, in 2017. Compared with ground observations, it was found that soybean and white mold in soybean fields were respectively classified with an overall accuracy of 95% and 99%. These results highlight the utility of freely available remotely sensed satellite images such as Landsat 8 images in estimating diseased crop extents, and suggest that further exploration of consistent high spatial resolution images such as Sentinel, and Rapid-Eye during the growing season will provide more details in the quantification of the diseased soybean.