A novel‐optimal monitoring model of rocky desertification based on feature space models with typical surface parameters derived from LANDSAT_8 OLI

Abstract

Previous studies monitoring the spatial distribution of rocky desertification have used single indices, a comprehensive index, or image classification. However, these approaches could not distinguish between the degrees of rocky desertification as they did not consider various influencing factors and their interactions. To avoid the above shortcomings, this study used the feature space model and seven typical land surface parameters to establish two categories of rocky desertification model: (1) a point-to-point model; and (2) a point-to-line model. A novel model for the optimal monitoring of rocky desertification was then proposed, which could take the comprehensive impacts of the human-nature system on the process of rocky desertification.The results showed that: (1) the feature space models provided a novel approach to large-scale monitoring of rocky desertification; (2) the point-to-line model incorporating the rock bare index (RBI)-dryness index feature space showed the optimal applicability for monitoring of rocky desertification, with a precision of 93.4%; and (3) the RBI performed the best in indicating the process of rocky desertification, with an average precision of 88.5%. The results of this study can act as a reference within the investigation of the spatiotemporal evolution of rocky desertification for karst mountain areas.

Machine learning algorithm for estimating karst rocky desertification in a peak-cluster depression basin in southwest Guangxi, China

“…The reduction in farmland area improved farmland management and increased regional gross industrial product, which together with the continuously rising gross domestic product of the tertiary industry caused a positive rocky desertification development 38 . In addition, the Program of Conversion from Cropland to Forest and Grassland had been applied to restore the vegetation ecosystem since 2005, which has contributed to the gradual development of the ecological and environmental conditions of karst rocky desertification in a benign direction 39 .…”

Extraction of Rocky Desertification Information in the Karst Area Based on the Red-NIR-SWIR Spectral Feature Space

“…Since karstic desertification occurs only in karst areas, and there were some non-karst areas in the study area, this study used 1:50,000 lithological vector data (http://www. karstdata.cn/, accessed on 22 June 2022) from Huaxi District to mask non-karst areas before constructing the rocky desertification index model [12]. As the reflectance of rocky areas is similar to that of built-up areas and bare land, it is difficult to distinguish them using the spectral features of the images, which can easily lead to misclassification of rocky desertification [38], so built-up areas and bare land need to be removed before the construction of the rocky desertification supervisory information extraction model; as water bodies are also unlikely to undergo rocky desertification, they also need to be removed.…”

“…Remote sensing technology has gradually become the main means of rocky desertification monitoring because of its ability to obtain large-scale rocky desertification information in a rapid and timely manner and to achieve large-scale rocky desertification monitoring and assessment [8][9][10][11]. Conventional remote sensing image classification methods such as supervised classification and unsupervised classification are highly subjective and have difficulty distinguishing the grade of rocky desertification, cannot directly indicate the development status of rocky desertification, and are not suitable for extracting spatial distribution, area and grade information of rocky desertification [12,13]. The mixed image element decomposition method (SMA) is a relatively common method in the study of karst desertification, but due to the discontinuous distribution of topography in karst areas and the influence of human activities, weathering and erosion, end element variation is common and it is very difficult to obtain feature end elements, so the SMA method cannot easily obtain information on rocky desertification [14,15].…”

“…Luo Jie et al [21] constructed the rocky desertification difference index (RSDDI) based on NDVI and Albedo surface parameter feature space, and extracted information on the spatial distribution and extent of rocky desertification in Dougu Township, and the results showed that the RSDDI could extract and classify rocky desertification information more accurately. Guo, Bing et al [12] constructed two types of feature space models (point-to-point and point-to-line) based on seven surface parameters to monitor the rocky desertification in Dafang County, and the results proved that the feature space models constructed based on different surface parameters could extract rocky desertification in karst areas, providing a new research method for the effective large-scale monitoring of rocky desertification.…”

A stratified environmental reference system for better understanding of the relationship between remote sensing observations and ground monitoring of karst rocky desertification

“…The normalized difference vegetation index (NDVI), a key product from MODIS or other remote sensing data, has a high positive correlation with vegetation coverage, suggesting an effective basis for KRD mapping (Li & Wu, 2015; Zhang et al, 2017). One case study showed that as KRD intensity decreases, NDVI increases (Chen et al, 2014; Guo et al, 2021). Yet, NDVI within the same rocky desertification level tends to fluctuate over a large range of values (Chen et al, 2014; Zhu et al, 2016).…”