A New Application of Random Forest Algorithm to Estimate Coverage of Moss-Dominated Biological Soil Crusts in Semi-Arid Mu Us Sandy Land, China

Abstract

Biological soil crusts (BSCs) play an essential role in desert ecosystems. Knowledge of the distribution and disappearance of BSCs is vital for the management of ecosystems and for desertification researches. However, the major remote sensing approaches used to extract BSCs are multispectral indices, which lack accuracy, and hyperspectral indices, which have lower data availability and require a higher computational effort. This study employs random forest (RF) models to optimize the extraction of BSCs using band combinations similar to the two multispectral BSC indices (Crust Index-CI; Biological Soil Crust Index-BSCI), but covering all possible band combinations. Simulated multispectral datasets resampled from in-situ hyperspectral data were used to extract BSC information. Multispectral datasets (Landsat-8 and Sentinel-2 datasets) were then used to detect BSC coverage in Mu Us Sandy Land, located in northern China, where BSCs dominated by moss are widely distributed. The results show that (i) the spectral curves of moss-dominated BSCs are different from those of other typical land surfaces, (ii) the BSC coverage can be predicted using the simulated multispectral data (mean square error (MSE) < 0.01), (iii) Sentinel-2 satellite datasets with CI-based band combinations provided a reliable RF model for detecting moss-dominated BSCs (10-fold validation, R2 = 0.947; ground validation, R2 = 0.906). In conclusion, application of the RF algorithm to the Sentinel-2 dataset can precisely and effectively map BSCs dominated by moss. This new application can be used as a theoretical basis for detecting BSCs in other arid and semi-arid lands within desert ecosystems.

Spectral Response Assessment of Moss-Dominated Biological Soil Crust Coverage Under Dry and Wet Conditions

“…The average potential annual evaporation in this period is 2300 mm, which is six times the annual average precipitation. The soil type is loose eolian sandy soil, which is infertile and vulnerable to wind erosion [17]. After thousands of years of human activities (such as reclamation of farmlands, deforestation, and overgrazing), natural vegetation in the sandy land has almost disappeared.…”

“…The portable spectrometer was used for the measurements at wavelength increments of 2 nm within a 325-1075 nm range, with a 15 ° field of view (FOV), and at a height of approximately 1 m above the ground. Spectral measurements were carried out under bright and sunny conditions from 10:00 to 15:00 China standard time (UTC+8) [17].…”

“…The BSC coverage was estimated as described by Chen et al [17]. Photographs were taken of the BSCs in the hoop (D3000 digital camera, Nikon) ( Figure 2b), after which Adobe Photoshop CC (19.1.6) and CAN_EYE (v 6.491) software were used to clip the hoop area and determine the BSC coverage (Figure 2c,d).…”

“…Weber and Hill [55] emphasized this good performance of CI to distinguish BSC from other ground objectives. Chen et al [17] also noted that a combination of CI bands is better for BSC identification when using random forest algorithms to detect BSCs.…”

“…Landsat-8 Operational Land Imager (OLI) and Sentinel-2 Multispectral Instrument (MSI) spectral bands used in this study[17].…”