Background Gully erosion, as an extreme form of land degradation, causes a potential danger for both local ecological security and food security. Gully measurement with high-efficiency and high-precision has important significance for gully erosion monitoring in the black soil area of Northeast China. This study aims to explore the applicability of satellite remote imagery Gaofen-7 for monitoring gully erosion in the black soil area.

Methods A typical watershed with area of 4.2 km² in Binxian county, Heilongjiang province, was selected as the study area to analyze the accuracy of extracting ephemeral gully (EG) and permanent gully (PG) parameters by visually interpretation using GF-7 stereoscopic satellite images (multispectral data of 0.8 m resolution and panchromatic data of 0.65 m resolution) with unmanned aerial vehicle (UAV) measurement data as a reference.

Results 1) The 64 ephemeral gullies were interpreted from UAV imagery, with a density of 15.23 gullies/km² and an average length of 83.43 m. The 80 ephemeral gullies were extracted from GF-7 with only 42 were the same as those from UAV images. The error in the number and density of ephemeral gullies interpreted using GF-7 images was larger than 50%, with many cases of misinterpretation. Using GF-7 images, the water flow paths, plot boundaries, and field roads may be misidentified as ephemeral gullies. It is inappropriate to be used for monitoring EG erosion at the small watershed scale. The 17 permanent gullies were interpreted from UAV images, with a density of 4.04 gullies/km² and a gully density of 2.02 km/km². The measurement in linear and areal density of the permanent gullies with GF-7 was reliable with the errors were 1.00% and 5.62%, respectively. 2) Comparing the length of the 42 EGs interpreted from both GF-7 and UAV images, the average error of gully length using GF-7 images was 4.54 m, with the relative error below 10% accounted for 40.48%. However, there was no statistically significant difference (P > 0.05) between the gully length measurements of the GF-7 and UAV and a good linear relationship was fitted (R²=0.93). 3) There was high reliability of gully length, width and area of permanent gullies extracted using the GF-7 images with the mean errors were 2.89%, 14.92% and 16.02%, respectively. The average error of the gully depth of the permanent gullies was 1.24 m, with a relative error of 24.46%, and the average error of gully gradient was around 7%, and the average relative errors of the permanent gullies' volumes were 69.95%, indicating the three-dimensional morphological parameters of permanent gullies from GF-7 were not reliable.

Conclusions There is high accuracy in the measurement of density and two-dimensional parameters of permanent gullies at the scale of small watershed using GF-7 images, while there is a large error in extraction of three-dimensional parameters of permanent gully. This study may provide an important reference for the investigation and monitoring of gullies in Chinese black soil area.