Gully erosion mapping based on remote sensing data

Abstract

Gully erosion is the most destructive type of soil erosion, induced by soil structure detachment. As a result, modest to massive incisions are made in the field. The process can degrade the quantity and quality of soil and potentially cause structural damage. Field studies are used to map the position of gullies, but they are inefficient in terms of time and cost, especially on a regional scale. Therefore, another approach is applied to visualize the probability of gully erosion development using geoenvironmental factors. Remote sensing data can be used to examine the condition of the land, leading to an accurate representation of the earth's surface. This research's primary goal is to predict the location of gully erosion using remote sensing data in the upper section of the Sapi Watershed, Banjarnegara, Indonesia. This location primarily consists of mountainous areas used for massive cultivation. Parameters comprising land use and vegetation area (ie. anomaly NDVI) derived from SENTINEL 2A, and topographic data from DEMNAS. The mapping process considers the actual location of the gully and other geographical characteristics using Random Forest. A total of 85 gully location records were collected and verified using Google Earth and field surveys. Nongully data were obtained using median DoD filters to distinguish between river and mountain top. 70% of the data are used for modelling and the rest for validation of model results. RF-generated prediction maps could provide an important instrument for planning and land conservation in the early phases of gully formation.