A remote sensing-based assessment of gullies along main roads in Emakhazeni, Mpumalanga Province, South Africa

Abstract

Soil erosion is a critical problem in the Southern African environment and often manifests through gully erosion. Such gullies are visible in many regions of South Africa. Various factors increase gully development. These may be broadly classified into natural factors (like rainfall characteristics, geology, soil, and hillslope gradient) and anthropogenic (human) factors (like population size and land use management practices). Road construction, an anthropogenic factor, changes the runoff characteristics of the surface by significantly reducing permeability, potentially reducing it to zero. Roads can thus contribute to accelerated soil erosion, frequently manifesting in gully formation at outlets of road culverts, where runoff concentration is maximised.

This study investigated whether it is possible to assess gully characteristics remotely. Gullies adjacent to national or regional armoured (tarred/paved) roads of the Emakhazeni Local Municipality of South Africa’s Mpumalanga Province were identified on Google EarthTM. Gullies were further digitised on Google EarthTM, sampled, and checked against field data to indicate if the two sets of data demonstrate similarities (i.e., ground-truthing). Gullies were sparsely distributed but tend to be more located on the R540 motorway between eMakhazeni and Dullstroom as well as the R33 motorway between eMakhazeni and Carolina. The approximate volume of material excavated from each gully was calculated. The average overall volume of gullies in the study area is 24,04 m3. Data collection took place between 15 March 2022 and 12 June 2022. Rainfall data for the past 30 years were collected from weather stations owned by Agricultural Research Council (ARC) and the South African Weather Service (SAWS) in the study area. The hillslope gradient was determined from Digital Elevation Models (DEMs) and compared against field data. Furthermore, the aerial extent (surface area) of gullies calculated using LiDAR (overall average of 173,68 m2) and Google EarthTM imagery (overall average of 65,18 m2) were also assessed. Statistical analyses were done with the computer language “R” in R-studio.

This study found that gullies of the type investigated cannot be readily identified on Google EarthTM due to the identification of false positives and that ground truthing is essential. Google EarthTM could, however, be used to measure the aerial extent of gullies. The hillslope gradient of gullies could also not be measured accurately using a DEM possibly because an even finer spatial resolution DEM is required for the size gullies studied. Furthermore, it was found that rainfall gauges ought to be installed at the gullies itself when assessing the gullies to obtain more accurate data. Finally, it was also found that LiDAR cannot be used to measure the excavated volume of gullies. Using proper LiDAR equipment (such as the Leica BLK 360 Mark 1 LiDAR device) and assessing a greater number of gullies using LiDAR might, however, change this finding.