Abstract
Under the stress of the global change, rocky desertification has become a serious environmental problem in the Karst Mountain area. At present, few remote sensing monitoring research works on rocky desertification based on feature space model have been conducted and reported. In this study, the Albedo-LST feature space remote sensing monitoring index based on point-point model has been proposed, and subsequently the spatio–temporal evolution pattern and driving mechanism of rocky desertification in Dafang district from 1986 to 2019 were analyzed. The results show that: (1) The point-point Albedo-LST feature space model of rocky desertification has good applicability with the overall accuracy of 90.79%; (2) From 1986 to 2019, the rocky desertification in Dafang district first showed an increasing trend (1986–2005) and then a decreasing trend (2005–2019); (3) The comprehensive evolution frequency of rocky desertification during 2001–2005 was the largest with 7.51% a−1, which was related to the implementation of the Grain for Green Project; (4) The single factor with the largest contribution rates to rocky desertification are land use type, landform, and temperature. The interactive factors with the largest explanatory power are temperature ∩ land use type and landform ∩ land use type. The research results can provide decision support for the prevention and control of rocky desertification in Southwest China.
1 Introduction
Rocky desertification is one of the typical types of land degradation, mainly due to the influence of natural factors and human activities, resulting in surface vegetation degradation, soil erosion, bedrock exposure, fragile ecological environment, and the decline of carrying capacity of surface ecological processes. Rocky desertification areas cover 22 million km2 globally, accounting for 12% of the total land area, which are one of the most important monitoring and researching critical zones of Earth. In addition, it is the third largest ecological problem after soil erosion in the Loess Plateau and desertification in the north of China, which is a hot issue in the study of karst ecological vulnerability at home and abroad [1–3]. It not only destroys the ecological environment, but also affects agriculture and forestry production and seriously hinders the sustainable development of local social economy, which is a major cause of local poverty (Sustainable development goals (SDGs) 1: No poverty). In addition, rocky desertification caused by human activities and climate change pose a major challenge to the sustainable development and affect the lives and livelihoods of millions of people (SDGs 13: Climate action). Dynamic monitoring of rocky desertification is essential to improve the regional livelihoods, reduce vulnerability, and mitigate economic risks (SDGs 15: Life and land).
In recent years, experts and scholars at home and abroad have carried out a lot of research in identification and dynamic monitoring of rocky desertification [4,5]. James et al. [6] proposed a global desertification monitoring method based on a DDP model. Sauro et al. [7] mainly analyzed the historical change in the rocky desertification and explored the interaction of climate change and human activities on the evolution process of rocky desertification. Combining the GIS technology and a landscape type model, Yang et al. [8] quantitatively analyzed the dynamic change in the landscape pattern of rocky desertification, revealed the significance of ecological management of rocky desertification, and provided reference for subsequent planning in the given area of the study. Wang et al. [9] utilized RS and GIS technology to extract the information of rocky desertification and then obtained a spatial differentiation rule of rocky desertification. Zhang et al. [10] obtained the abundances of vegetation and exposed rock in Southwest China using linear spectral unmixing method. Wen and Su [11] applied the bedrock exposure rate, slope, and vegetation index to extract the rocky desertification information and then explored the dynamic evolution patterns of the rocky desertification and predicted their future development trend. Chen et al. [12] established the relations between the rocky desertification and soil erosion by investigating changes in soil magnetic susceptibility profiles on a karst slope in Southwest China. Based on a fuzzy analytic hierarchy process combined with climate, topography, geomorphology, and socio-economic factors, Gu et al. [13] investigated the spatial and temporal evolution characteristics of rocky desertification and predicted the trend of change. Luo et al. [14] utilized the normalized differences vegetation index, net primary productivity, surface albedo, and slope to calculate and evaluate the rocky desertification condition in the Southwest China with ArcGIS 10.5 and found that Guizhou had the most widely distributed area of rocky desertification, followed by Yunnan and Guangxi. Zhang et al. [15] proposed a novel karst rocky desertification index to monitor and evaluate the level of rocky desertification in Pingguo County, Guangxi Province. However, most of the above studies applied the image classification method and comprehensive index method to monitor the rocky decertification. The image classification method could only identify the distribution range of rocky desertification, but could not obtain its spatial differentiation pattern. The comprehensive index method could consider the effects of natural factors and human activity, but ignored the interaction and nonlinear influences among different factors. In recent years, the feature space monitoring model based on typical surface parameters has been widely used in land degradation (desertification, salinization, etc.) monitoring, and has achieved good results. However, few studies on remote sensing monitoring of rocky desertification based on feature space models have yielded reports [16,17]. The model deployed in this study can better reflect the interaction and nonlinear relationships between influencing factors on the evolution process of rocky desertification [18–20]. In addition, the influences of natural factors and human activity on the process of rocky desertification are complex, which is difficult to be distinguished. Many scholars applied the linear regression model and principal component analysis to investigate the effects of natural factors and human activity, which could not quantitatively discriminate the above factors. Meanwhile, the dominant factors of the rocky desertification would change with the increasing intensity of human activity. However, most previous studies were conducted in certain periods. In recent years, the geodetector has been widely used in the analysis of the driving mechanism of ecological environment problems, such as soil erosion, desertification, and land use changes, which has provided a new approach to analyze the dominant factor and its changes in the rocky desertification in different periods.
Based on the above analysis, this study constructed the remote sensing monitoring index of rocky desertification based on Albedo-LST feature space, and then introduced the geodetector to analyze the spatial–temporal evolution law and the dominant factors of the rocky desertification in Dafang district.
2 Methods
2.1 Study area
Dafang district, located in the Northwest Guizhou Province, belonging to Bijie City (105°15′47″–106°08′04″E, 26°50′02″–27°36′04″N, Figure 1), is one of the most seriously affected zones by rocky desertification with the area of 3505.21 km2. The land that is composed of mountainous and hilly terrain is higher in the middle and northeast parts. Land use types are mainly composed of cultivated land, forest, and grassland, with the forest accounting for the largest proportion, at 44.61%. The study area was dominated by a northern subtropical humid monsoon climate with abundant rainfall and mild temperament. The annual average precipitation is 1,155 mm, while the precipitation is mostly concentrated in summer, accounting for 78.8% of annual precipitation.
Location of the study area and its topography.
2.2 Data source and preprocessing
The Landsat images (Path 128/Row 41) used in this study in 1986 (July, 13), 2001 (August, 7), 2005 (July, 29), and 2019 (August, 23) were downloaded from the Geospatial data cloud. The spatial resolution is 30 m and the cloud cover is less than 10%. The images have been processed by atmospheric correction with the digital number (DN) value converting into band reflectance by ENVI 5.3. Digital elevation model (DEM) data with a spatial resolution of 90 m is derived from the Resource Environment and Data Center of Chinese Academy of Sciences (https://www.resdc.cn/). The land use type datasets (1985, 2000, 2005, and 2018) with a scale of 1:100,000 are obtained from the Aerospace Information Research Institute, Chinese Academy of Sciences, and the average precision of these above datasets is 87.8%, which can satisfy the need of this study. Population and other statistical data of Dafang district are from Dafang District People’s Government portal (http://www.gzdafang.gov.cn/) and statistical yearbook data. The meteorological station data, including daily precipitation and daily average temperature, is downloaded from China Meteorological Data Network (https://www.sogou.com).
2.3 Method
2.3.1 Principle of feature space model
Different types of feature spaces based on typical surface parameters have been widely used in the monitoring of land degradation processes and soil moisture. In this article, the feature space model had been introduced to monitor the rocky desertification process. Taking Albedo-LST feature space as an example (Figure 2), Albedo increases with the intensification of the rocky desertification, because the reflectivity of underlying surface increases with the increase in the exposed rocks. In addition, due to the difference in heat capacity of different substances, the temperature of the bare rocks tends to increase faster and higher than that of the vegetation cover, water, and bare soil. In addition, vegetation can also regulate regional microclimate (humidifying and cooling) through transpiration. Therefore, in Albedo-LST feature space, with the increase in Albedo and surface temperature, the degree of rocky desertification showed an increasing trend. The closer to the AB line, the slighter the rocky desertification. On the contrary, the farther away from the CD line, the more serious the rocky desertification [21].
Principle of feature space.
2.3.2 Calculation and standardization of typical surface parameters
In this study, the above two typical surface parameters, namely, Albedo and LST, were selected to construct the remote sensing monitoring index of the rocky desertification.
Albedo:
LST:
where B blue, B red, B nir, B swir1, and B swir2 refer to the band reflectance of blue, red, swir1, and swir2, respectively. a and b are the coefficients, corresponding to atmospheric transmittance and surface emissivity. T refers to the radiation temperature. T a refers to the average atmospheric temperature.
In the formula, M i represents the i normalized index. MI i represents the i index. MI i,min represents the minimum value of the i index. MI i,max represents the maximum value of the i index.
2.3.3 Dynamic degree model
Similar to the dynamic degree model of land use [22], the frequency of single rocky desertification evolution and the frequency of comprehensive rocky desertification evolution are applied to analyze the temporal evolution patterns of rocky desertification. The formulae are as follows:
Single rocky desertification evolution frequency:
Comprehensive rocky desertification evolution frequency:
In the formula, S m and S n are the area of early and late rocky desertification. The unit is km2, T is time, f is a single rocky desertification evolution frequency, S mi and S ni are the area of the first type of rocky desertification intensity grade in the study area at the beginning and the end of the study. The unit is km2, and T is time. When T represents the year, the unit is % a−1 (a−1 represents the annual measurement unit). F is the evolution frequency of comprehensive rocky desertification, and the unit is % a−1 [23].
2.3.4 Geodetector
The factor detector is utilized to quantitatively detect whether a geographical factor affects the spatial distribution difference. The interactive detector calculates and compares the q value of each single factor and the q value after the superposition of the two factors to determine whether there is an interaction between the two factors [24]. The models are as follows:
In the formula,
3 Results
3.1 Remote sensing monitoring index of rocky desertification based on feature space
3.1.1 Construction of feature space model
Using the 2D scatter plot tool of ENVI 5.3, two typical surface parameter factors were combined to construct a 2D feature space. By analyzing the distributions of different degrees of rocky desertification in the feature space, it was found that the point clusters that correspond to different degrees of rocky desertification pixels showed point-to-point spatial characteristics, that is, in the feature space, the farther the distance from the point O, the higher the degree of rocky desertification (Figure 3).
Distribution law of different degrees of rocky desertification in feature space.
According to the point-to-point distance formula, the remote sensing monitoring index of rocky desertification was as follows:
where KRMI refers to the karst rocky monitoring index; m and n refers to the coordinate of a specific point O.
With the above remote sensing monitoring index of rocky desertification based on feature space model, the spatial distributions of rocky desertification were calculated using the grid calculator of ArcGIS 10.2. In order to reduce the interference of built-up land, paddy field, and water, these above specific land use types that were derived from the land use datasets in 1985, 2000, 2005, and 2018 were extracted to enhance the inversion accuracy. Combing the natural breaks method and traditional classification standard of rocky desertification (vegetation coverage, bedrock exposure rate, geological lithology, and soil depth) with the observation of Google Earth and field observed data, the rocky desertification index could be divided into five grades, namely, no rocky desertification (KRMI ≤ 0.24), slight rocky desertification (0.24 < KRMI ≤ 0.41), moderate rocky desertification(0.41 < KRMI ≤ 0.63), intensive rocky desertification (0.63 < KRMI ≤ 0.81), and severe rocky desertification (KRMI > 0.81). The detailed processes are as follows: (1) First, to obtain the annual average rocky desertification index during 1985–2018; (2) Second, to obtain the average value of rocky desertification index with different types or levels of vegetation coverage, bedrock exposure rate, geological lithology, and soil depth based on the annual average rocky desertification index; (3) Third, to calculate the thresholds of rocky desertification index with natural breaks methods; (4) Fourth, to combine the above thresholds and statistical average values to obtain the final thresholds. The inversion results based on the point-point Albedo-LST feature space index of rocky desertification are shown in Figure 4.
Spatial distribution of rocky desertification based on point-point Albedo-LST feature space model.
3.1.2 Validation of inversion accuracy
In order to verify the inversion accuracy of the monitoring index based on the point-point Albedo-LST feature space model, 239 validation samples (30 × 30) were evenly set in the study area via use of Google Earth. Then, utilizing the 86 field observed data of 2018 that could match the validation samples in position, the classification results of rocky desertification for validation samples were determined by visual interpretation and traditional classification standard of rocky desertification. The distribution of validation samples is shown in Figure 5. Finally, the confusion matrix of field observed value and inversion value for rocky desertification were obtained, which are displayed in Table 1, and the overall inversion accuracy of rocky desertification in 2018 was 90.79%, which indicated that the monitoring index of the rocky desertification based on the point-point Albedo-LST feature space model had better applicability in the study area.
Distribution of validation samples of rocky desertification in 2018.
Confusion matrix of field observed value and inversion value for rocky desertification in 2018
| Levels of rocky desertification | Inversion values in 2018 | User accuracy (%) | ||||||
|---|---|---|---|---|---|---|---|---|
| NRD | SLRD | MRD | IRD | SRD | Sum | |||
| Observed values | NRD | 53 | 4 | 0 | 0 | 0 | 57 | 92.98 |
| SLRD | 3 | 68 | 1 | 3 | 1 | 76 | 89.47 | |
| MRD | 0 | 2 | 27 | 1 | 0 | 30 | 90.00 | |
| IRD | 2 | 1 | 3 | 48 | 1 | 52 | 92.31 | |
| SRD | 0 | 2 | 0 | 1 | 21 | 24 | 87.50 | |
| Sum | 58 | 77 | 31 | 53 | 23 | 239 | — | |
| Producer accuracy (%) | 91.38 | 88.31 | 87.10 | 90.57 | 91.30 | — | — | |
| Overall accuracy (%): 90.79 | Kappa coefficient: 0.88 | |||||||
Note: NRD refers to no rocky desertification, SLRD refers to slight rocky desertification, MRD refers to moderate rocky desertification, IRD refers to intensive rocky desertification, and SRD refers to severe rocky desertification.
3.2 Area changes in different levels of rocky desertification
Table 2 shows that the total area of rocky desertification in Dafang district increased from 2409.37 km2 in 1986, accounting for 69.02% of overall area, to 76.51% in 2019 with an increased area of 263.14 km2. The rocky desertification was mainly distributed in the southeast and northwest parts. Overall, the area of severe rocky desertification increased mostly from 7.15 to 16.93% during 1986–2019, with an increase of 341.85 km2. The area of no rocky desertification showed a decreasing trend, from 1081.6 km2 in 1986, accounting for 30.98%, to 818.45 km2 in 2019, with a decrease of 7.54%. The area of no rocky desertification was mainly distributed in the northeast.
Area comparison of different levels of rocky desertification
| Levels of rocky desertification | 1986 | 2001 | 2005 | 2019 | ||||
|---|---|---|---|---|---|---|---|---|
| Area (km2) | Percentage | Area (km2) | Percentage (%) | Area (km2) | Percentage | Area (km2) | Percentage | |
| SRD | 22.75 | 0.65 | 316.84 | 9.07 | 48.65 | 1.39 | 298.54 | 8.54 |
| IRD | 249.69 | 7.15 | 657.40 | 18.82 | 401.10 | 11.48 | 591.54 | 16.93 |
| MRD | 752.90 | 21.57 | 877.30 | 25.12 | 959.29 | 27.46 | 757.16 | 21.68 |
| SLRD | 1384.03 | 39.65 | 816.88 | 23.39 | 1160.29 | 33.23 | 1025.27 | 29.36 |
| NRD | 1081.60 | 30.98 | 822.56 | 23.56 | 921.62 | 26.40 | 818.45 | 23.44 |
Note: NRD refers to no rocky desertification, SLRD refers to slight rocky desertification, MRD refers to moderate rocky desertification, IRD refers to intensive rocky desertification, and SRD refers to severe rocky desertification.
From 1986 to 2001, the area of severe rocky desertification showed a larger increase, from 0.65% in 1986 to 9.07% in 2001. The area of slight rocky desertification showed a decreasing trend, from 36.65 to 23.39%, with a total decrease of 567.15 km2, which accounted for 16.25% of the total area. It could be seen that the rocky desertification showed a trend of deterioration during this period. During 2001–2005, more than 25% of the total area of rocky desertification showed a decreasing trend, with the largest decrease in the severe rocky desertification area, which decreased from 316.84 km2 in 2001 to 48.65 km2 in 2005, and the proportion decreased from 9.07 to 1.39%. The area of slight and no rocky desertification showed an increasing trend, from 816.88 and 822.56 km2 to 1160.29 and 921.62 km2, respectively. This was mainly because the implementation of the ecological project of returning farmland to forest and grassland effectively curbed the aggravation of rocky desertification in the large area. From 2005 to 2019, the area of rocky desertification in Dafang district exhibited an increasing trend. The area of severe and intensive rocky desertification areas increased, while the area of moderate, slight, and no rocky desertification areas decreased. The area of severe rocky desertification increased from 48.65 to 298.54 km2, accounting for 7.16% of the increase, and the area increased by 249.89 km2.
In conclusion, from 1986 to 2019, the rocky desertification in Dafang district first showed an increasing trend (1986–2005) and then a decreasing trend (2005–2019).
3.3 Evolution frequency of rocky desertification
Combined with the evolution rate of rocky desertification and Equations 6 and 7, the single rocky desertification evolution frequency and comprehensive rocky desertification evolution frequency of rocky desertification at different times were further calculated (Table 3 and Figure 6). It could be seen from Table 3 that the frequency of rocky desertification evolution in the first stage (1986–2001) was relatively large, and there was a two-stage differentiation compared with the second stage (2001–2005). The frequency of rocky desertification evolution in the second stage (2001–2005) was the largest, and the severe rocky desertification area had the largest frequency of evolution. This indicated that the comprehensive treatment project of rocky desertification in Bijie City had achieved remarkable results, which reduced the evolution speed of rocky desertification. The frequency of rocky desertification evolution in the third stage (2001–2019) was the lowest. Slight and intensive rocky desertification changed from negative change in the first stage to the positive change in the second stage, and finally again negative change, which might be due to the small influence of slight and moderate rocky desertification on the local area, which had not become key prevention and control areas of rocky desertification. The intensive and severe rocky desertification areas showed a positive change before 2001, indicating that the intensity of rocky desertification in the region was relatively lower before 2000. Human activity and natural factors at that time did not have very serious consequences for rocky desertification, while there was a negative correlation after 2001. There was a positive correlation between rocky desertification in 2005–2019. The increase in rocky desertification area was mainly due to the transformation of areas of no rocky desertification, slight and moderate rocky desertification areas, while intensive and severe rocky desertification areas changed places repeatedly. In summary, the evolution frequency of intensive and severe rocky desertification fluctuates greatly, and other intensity frequency was smaller.
Evolution frequency of single rocky desertification at different times (% a−1)
| NRD | SLRD | MRD | IRD | SRD | |
|---|---|---|---|---|---|
| 1986–2001 | −2.10 | −4.63 | 0.95 | 4.13 | 6.19 |
| 2001–2005 | 2.69 | 7.40 | 2.14 | −15.97 | −137.80 |
| 2005–2019 | −0.90 | −0.94 | −1.91 | 2.30 | 5.98 |
| 1986–2019 | −0.97 | −1.06 | 0.02 | 1.75 | 2.80 |
Note: NRD refers to no rocky desertification, SLRD refers to slight rocky desertification, MRD refers to moderate rocky desertification, IRD refers to intensive rocky desertification, and SRD refers to severe rocky desertification.
Evolutionary frequency of comprehensive rocky desertification in different periods.
As shown in Figure 6, the comprehensive rocky desertification evolution frequencies in different periods were 3.16, 7.51, 1.80, and 1.08 % a−1 in 1986–2001, 2001–2005, 2005–2019, and 1986–2019, respectively. With the change in study periods, the comprehensive rocky desertification evolution frequency decreased, and the area transferred of rocky desertification was the largest with a stronger fluctuation from 2001 to 2005. During this period, the comprehensive control project of rocky desertification was implemented in most areas, indicating that reasonable conversion of farmland to forest and grassland, afforestation, changing the traditional land use mode, and developing new economic growth points were effective approaches for rocky desertification control.
3.4 Spatial distributions of rocky desertification
3.4.1 Spatial distributions of rocky desertification in different slope zones
The spatial distribution characteristics of rocky desertification were significantly affected by slope. As shown in Table 4, zones of severe desertification, intensive rocky desertification, and moderate rocky desertification were mainly distributed in the slope of <5°, 5–8°, and 8–15°, with the areas of 68.95, 131.94, and 308.93 km2, respectively. There was a decreasing trend with the increase in slope, while zone of no rocky desertification was mainly distributed in the steep slopes, accounting for 72.84% (Figure 7). This was because steep slopes were not suitable for economic development and human habitation, where the interference intensity of human activities was lower.
Area transfer matrix of different levels of rocky desertification
| Slope (°) | <5° | 5–8° | 8–15° | 15–25° | >25° | |
|---|---|---|---|---|---|---|
| Area (km2) | NRD | 62.83 | 92.72 | 257.16 | 250.98 | 127.63 |
| SLRD | 116.11 | 161.77 | 389.00 | 278.12 | 96.20 | |
| MRD | 123.76 | 155.09 | 308.93 | 165.06 | 45.62 | |
| IRD | 116.73 | 131.94 | 224.88 | 95.18 | 22.67 | |
| SRD | 68.95 | 68.87 | 102.17 | 36.85 | 8.49 | |
Note: NRD refers to no rocky desertification, SLRD refers to slight rocky desertification, MRD refers to moderate rocky desertification, IRD refers to intensive rocky desertification, and SRD refers to severe rocky desertification.
Area percentages of rocky desertification in different slopes.
3.4.2 Spatial distributions of rocky desertification in different land use types
There were significant differences in the spatial distribution characteristics of rocky desertification among different land use types. Table 5 and Figure 8 show that zones of severe rocky desertification, intensive rocky desertification, and moderate rocky desertification were mainly distributed in the cultivated land, of which moderate rocky desertification area accounts for the largest proportion of 26.67%. Zones of slight rocky desertification and no rocky desertification were mainly distributed in forest land, with slight rocky desertification area accounting for the largest proportion at 36.22%. It indicated that afforestation was an effective way to control rocky desertification. Consequently, the implementations of returning farmland to forest projects were conducive to control rocky desertification.
Area transfer matrix of rocky desertification in different land use types
| Land use types | Cultivated land | Forest | Grassland | Other land | |
|---|---|---|---|---|---|
| Area (km2) | NRD | 173.01 | 521.56 | 96.11 | 0.32 |
| SRD | 321.26 | 566.66 | 151.29 | 1.70 | |
| MRD | 383.16 | 282.26 | 128.62 | 4.14 | |
| IRD | 367.88 | 140.60 | 75.83 | 6.67 | |
| SRD | 191.18 | 53.31 | 31.90 | 7.72 | |
Note: NRD refers to no rocky desertification, SLRD refers to slight rocky desertification, MRD refers to moderate rocky desertification, IRD refers to intensive rocky desertification, and SRD refers to severe rocky desertification.
Area percentage of rocky desertification in different land use types.
3.4.3 Spatial distributions of rocky desertification in different landforms
Zones of severe desertification, intensive rocky desertification and moderate rocky desertification in different landforms are mainly distributed in the middle and low mountains and hills, accounting for 40.08 and 73.80% (Table 6 and Figure 9), respectively. The area of rocky desertification in hills is 1172.8 km2, accounting for 92.16%, while zones of no rocky desertification were mainly distributed in the middle and low mountains and high mountains, with a total area of 710.59 km2. In addition, the area of slight rocky desertification in the middle and low mountain areas was 809.09 km2, accounting for the largest proportion of 33.65%. In addition, zone of no rocky desertification had the largest proportion of 42.92% in the high mountain areas.
Area comparison of levels of rocky desertification in different landforms
| Landform | Platform | Hills | Middle-low mountains | High mountains | |
|---|---|---|---|---|---|
| Area (km2) | NRD | 3.59 | 67.49 | 631.76 | 78.83 |
| SLRD | 6.75 | 158.05 | 809.09 | 56.32 | |
| MRD | 5.64 | 250.12 | 505.26 | 29.13 | |
| IRD | 2.90 | 253.33 | 314.20 | 15.28 | |
| SRD | 0.69 | 131.64 | 144.25 | 4.12 | |
Note: NRD refers to no rocky desertification, SLRD refers to slight rocky desertification, MRD refers to moderate rocky desertification, IRD refers to intensive rocky desertification, and SRD refers to severe rocky desertification.
Area comparison of levels of rocky desertification in different landforms.
3.5 Dominant factors of rocky desertification during different study periods
In order to investigate the dominant factors affecting the rocky desertification, five typical natural factors (slope, landform, land use type, temperature, and precipitation) and two socio-economic factors (population density and GDP) were selected in this study. The factor detector and interaction detector in the geodetector were used to explore and analyze the spatial differentiation law and driving mechanism of the rocky desertification. Rocky desertification intensity grade was a dependent variable Y, influence factors, including slope, land use type, landform, temperature, precipitation, population density, and GDP were X 1, X 2, X 3, X 4, X 5, X 6, and X 7, respectively.
3.5.1 Single factor
The factor detector was utilized to calculate the explanatory power of influence factors on the rocky desertification. The explanatory power of factor X represents the extent to which this factor explains the spatial differentiation of rocky desertification.
The explanatory power of each factor on the rocky desertification in different periods is shown in Table 7. The dominant factors of rocky desertification evolution in the past 33 years were mainly land use type, landform, and temperature. The contribution rate of land use type, landform, and temperature in four periods (Figure 10) were greater than 30, 20, and 15%, respectively. The contribution rate of temperature was greater than that of slope, which was mainly because high temperature resulted in drought and vegetation degradation, which would aggravate the process of rocky desertification [25].
Q value of single factor weight (1986–2019)
| Time factor | Slope | Land use type | Land form | Temperature | Precipitation | Population density | GDP |
|---|---|---|---|---|---|---|---|
| 1986 | 0.01 | 0.05 | 0.03 | 0.03 | 0.01 | 0.01 | 0.01 |
| 2001 | 0.04 | 0.24 | 0.09 | 0.07 | 0.02 | 0.03 | 0.04 |
| 2005 | 0.01 | 0.16 | 0.07 | 0.06 | 0.01 | 0.05 | 0.05 |
| 2019 | 0.07 | 0.15 | 0.11 | 0.08 | 0.02 | 0.03 | 0.04 |
Single factor contribution rate at different times.
3.5.2 Interactive factor
Interactive detection was to mainly determine whether the impact of driving factors on vegetation change was independent, which could be mainly divided into five categories, namely, double-factor nonlinear weakening, double-factor enhancement, mutual independence, single-factor nonlinear weakening, and double-factor nonlinear enhancement. The expressions are:
Figure 11 showed that there were significant differences in the explanatory power of interaction between different historical periods and different factors. The greatest explanatory power of interaction factors in 1986 was landform ∩ land use type (q = 0.073), while that of population density ∩ slope was the smallest with q = 0.02. In 2001, the explanatory power of land use types ∩ temperature and landform ∩ land use type were larger, with the q values of 0.286 and 0.296, respectively. The q value of population density ∩ precipitation was 0.052. The interactive factor with the largest explanatory power in 2005 was the same as that in 2001 and 1986, but the q value was larger than that in 1986 and smaller than that in 2001. In 2019, the dominant interactive factor was the same as the previous period, and the q values of land use types ∩ temperature and landform ∩ land use type were 0.21 and 0.22, repectively.
Detection weight of interaction factors of rocky desertification in different periods: (a) 1986, (b) 2001, (c) 2005, and (d) 2019. ++ represents double factor enhancement, # represents double factor nonlinear enhancement, * represents independence.
In 1986, the interaction was mainly double-factor nonlinear enhancement, whereas in 2001 and 2005, the interaction was mainly double-factor enhancement, and in 2019, the interaction between the two factors was mainly double-factor enhancement. This showed that the explanatory power of the interactive factor on rocky desertification was larger than that of the single factor. In 2019, the explanatory power of the interactive factor was larger than that of the maximum explanatory power of the single factor, which showed that the combined effect of the two factors on the rocky desertification was weakening.
The contribution rate of interaction between population density, GDP, and natural factors in socio-economic factors was larger than that of single socio-economic factors In addition, the contribution rate of the interaction between the two factors was larger than that of the single factor, mainly showing double factor enhancement and double factor nonlinear enhancement, which indicated that the evolution of rocky desertification intensity was not only related to a single factor, but also to the interaction between the factors [27].
4 Discussion
4.1 Advantages of the new proposed model
In recent years, the image classification method, single indices method, and comprehensive index method have been widely applied in the monitoring of the rocky desertification on large scale, which obtained better results [28,29]. The image classification method can better determine the scope and boundary of the rocky desertification, while it cannot obtain and reflect the internal spatial variation information. Due to the fact that the evolution process of rocky desertification has been influenced by both the natural factors and human activity, the single indices method can reflect the condition of rocky desertification to some extent, which also ignored other characteristics of the rocky desertification. The comprehensive index method can fully consider most of the influencing factors to evaluate the condition of rocky desertification: however, it ignored the interactions among factors. Moreover, the linear weight always expands the degree of influence of one factor on the process of the rocky desertification.
The feature space model can consider the interactions among influencing factors, which has been utilized to monitor the land degradation information, such as desertification and salinization. However, few studies on rocky desertification based on feature space model are reported [30]. In this study, on the basis of fully considering the ecological environment characteristics of karst rocky desertification in southwest mountainous areas, the surface albedo and surface temperature are selected as sensitive parameters to construct the rocky desertification feature space, and then the rocky desertification remote sensing monitoring index based on point-point Albedo-LST feature space model has been proposed with the overall precision of 90.79%. The proposed novel monitoring index has better applicability than previous studies [31,32]. The processes of rocky desertification are often affected by multiple natural and human factors such as vegetative cover, geology, geomorphology, climate change, and over-exploitation. Albedo increases with the intensification of rocky desertification, due to the fact that the reflectivity of the underlying surface increases with the increase in the exposed rocks [33]. In addition, due to the difference in the heat capacity of different substances, the temperature of bare rocks tends to increase faster and higher than that of the vegetation cover, water, and bare soil. This model can not only fully take into account the comprehensive influence of multiple factors, but also take into account the nonlinear characteristics of the interaction between multiple factors. This method provides a new approach for rocky desertification information extraction in southwest mountainous areas.
4.2 The main causes of spatial–temporal evolution pattern of rocky desertification
During 2000–2005, the rocky desertification in the vast area had the highest frequency of evolution and the most intensive fluctuation. This was mainly because the effective implementations of the ecological project of returning farmland to forest in 1999, the ecological project of returning farmland to grassland in 2003, and the argument that “green water and green mountains superior to mountains of silver and gold” proposed in 2005 were conductive to the control of karst rocky desertification in Southwestern China [34]. At the same time, the Bijie experimental area was a key area of the national dynamic monitoring of rocky desertification initiative. It was also the only experimental area in China with the theme of “poverty alleviation and development, ecological construction, and population control.” The experimental area included seven districts, namely, Jinsha, Zhijin, Weining, Nayong, Dafang districts, Qixing in Hezhang, and Qianxi. It was representative and typical in the rocky desertification mountainous areas in Southwestern China. Slope was one of the main factors affecting rocky desertification. Rocky desertification was mainly concentrated upon flat land and smaller slope zones. This was mainly because these regions were suitable for human habitation and frequent human activities had greater impacts on rocky desertification in these regions [35]. Lower population density, less intensive human activities, and higher vegetative coverage could effectively prevent the occurrence and development of rocky desertification. During the study periods, the dominant factors of rocky desertification were land use type, landform, and temperature. The local farming methods were one of the main reasons for rocky desertification, and rocky desertification was the most serious in the cultivated land area. Surface water resources are key factors in the evolution process of rocky desertification, which would directly affect the vegetation grown and recovery [5,10]. However, in the karst mountain area zones, the water sources are scare and mainly derived from precipitation. During the past decades, the precipitation showed an overall decreasing trend, while the temperature showed an increasing trend [9,15]. The increased dryness has led to varying degrees of drought, which greatly constraint the vegetation recovery in zones with severe rocky desertification. Meanwhile, the precipitation is mainly concentrated in July, August, and September. The intensive and heavy precipitation can carry away a large amount of surface soil, which would lead to serious soil erosion and rocky desertification [20,27].
5 Conclusion
In this study, the Albedo-LST feature space remote sensing monitoring model based on a point-point model was constructed fully considering the ecological environment of southwest karst mountainous area, and then the spatial–temporal evolution patterns and the dominant driving factors of rocky desertification in Dafang district were analyzed. The main conclusions are as follows:
The rocky desertification remote sensing monitoring index based on the point-point Albedo-LST feature space model has good applicability with the overall accuracy of 90.79%.
From 1986 to 2019, the rocky desertification in Dafang district first showed an increasing trend (1986–2005) and then a decreasing trend (2005–2019).
The comprehensive evolution frequency of rocky desertification in 2001–2005 was the largest with 7.51% a−1, which was related to the implementation of the Grain for Green Project.
The single factors with the largest contribution rates to rocky desertification are land use type, landform, and temperature. The interactive factors with the largest explanatory power are temperature ∩ land use type and landform ∩ land use type.
-
Funding information: This work was supported by National Natural Science Foundation of China (Grant no. 42101306); Natural Science Foundation of Shandong Province (Grant no. ZR2021MD047); Open Research Fund of the Key Laboratory of Digital Earth Science, Chinese Academy of Sciences (Grant no. 2019LDE006); A grant from State Key Laboratory of Resources and Environmental Information System; Open Fund of Key Laboratory of Meteorology and Ecological Environment of Hebei Province (Grant no. Z202001H); Open fund of Key Laboratory of National Geographic Census and Monitoring, MNR (Gant no. 2020NGCM02); The Open Fund of Key Laboratory of Urban Land Resources Monitoring and Simulation, Ministry of Natural Resources (Grant no. KF-2020-05-001); Open fund of Key Laboratory of Land use, Ministry of Natural Resources (Grant no. 20201511835); Undergraduate teaching research and reform project of Shandong University of Technology.
-
Author contributions: Ye Wen and Bing Guo: conceptualization, methodology, software, data curation, and writing-original draft preparation; Wenqian Zang and Jibao Lai: investigation and revision. Ran Li: supervision, writing-reviewing and editing, and revision.
-
Conflict of interest: All authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.
-
Data availability statement: The data that support the findings of this study are available from the corresponding author, Guo B, upon reasonable request.
References
[1] Yao YH, Sou ND, Zhang JY, Hu YF, Kou ZX. Spatiotemporal characteristics of karst rocky desertification and the impact of human activities from 2010 to 2015 in Guanling County, Guizhou Province. Prog Geogr. 2019;38(11):1759–69.10.18306/dlkxjz.2019.11.011Search in Google Scholar
[2] Wang Q, Zhao XQ, Pu JW, Li SN, Miao PP. Temporal and spatial pattern evolution of rocky desertification in Karst region of southeast Yunnan-A case study of Guangnan county. Carsol Sin. 2021;1–14.Search in Google Scholar
[3] Li Y. Analysis of element characteristics of cave drip water in the Karst rocky desertification areas. Guizhou Norm Univ. 2017;2:61.Search in Google Scholar
[4] Liu CL, Wang XH, Shu TZ. Atlas analysis of spatial-temporal evolution of Karst Rocky desertification in Luodian County based on GIS and RS technology- taking Luodian County of Guizhou Province as an example. J Anhui Agri Sci. 2016;44(2):150–3.Search in Google Scholar
[5] Chen HM, Zhang JW, Tang LY, Su M, Tian D, Zhang L, et al. Enhanced Pb immobilization via the combination of biochar and phosphate solubilizing bacteria. Env Int. 2019;127:395–401.10.1016/j.envint.2019.03.068Search in Google Scholar PubMed
[6] James FR, Smith DMS, Eric FL, Turner BL, Michael M, Simon PJB, et al. Global desertification: building a science for dryland development. Science. 2007;316(5826):847–51.10.1126/science.1131634Search in Google Scholar PubMed
[7] Sauro U. Human impact on the karst of the Venetian Fore-Alps, Italy. Env Geol. 1993;21(3):115–21.10.1007/BF00775294Search in Google Scholar
[8] Yang XY, Zhou ZF, Li B. Dynamic analyzing in Karst rocky desertification based on the landscape model spatial pattern. Procedia Environ Sci. 2011;10:2083–90.10.1016/j.proenv.2011.09.325Search in Google Scholar
[9] Wang PS, An YL. Spatial-temporal analysis of rocky desertification in Guizhou province during 2000–2010. J Guizhou Normal Univ (Nat Sci). 2014;32(3):10–5 + 46.Search in Google Scholar
[10] Zhang X, Shang K, Chen Y, Shuai T, Sun YL. Estimating ecological indicators of karst rocky desertification by linear spectral unmixing method. 2014;31:86–94.10.1016/j.jag.2014.03.009Search in Google Scholar
[11] Wen LQ, Su ZF. Evolution characteristics of rocky desertification during 2004–2016 in Guizhou Province. China Acta Ecol Sin. 2020;40(17):5928–39.10.5846/stxb201906251343Search in Google Scholar
[12] Chen ST, Guo B, Zhang R, Zang WQ, Wei CX, Wu HW, et al. Quantitatively determine the dominant driving factors of the spatial–temporal changes of vegetation NPP in the Hengduan Mountain area during 2000–2015. J Mt Sci Engl. 2021;18(2):427–45.10.1007/s11629-020-6404-9Search in Google Scholar
[13] Gu ZF, Bai G, Liu ZK. Prediction and evaluation of karst rocky desertification evolution trend in southwest Guizhou based on fuzzy analytic hierarchy process. J Zhengzhou Univ Aeron. 2021;39(2):95–102.Search in Google Scholar
[14] Luo XL, Wang SJ, Bai XY, Tan Q, Ran C, Chen H, et al. Analysis on the spatio-temporal evolution of rocky desertification in Southwestern Karst area. Acta Ecol Sin. 2021;41(2):680–93.Search in Google Scholar
[15] Zhang J, Liu L, Liu XN, Luo WQ, Wu L, Zhu LH. Spectral analysis of seasonal rock and vegetation changes for detecting karst rocky desertification in southwest China. Int J Appl Earth Obs. 2021;100:102337.10.1016/j.jag.2021.102337Search in Google Scholar
[16] Guo B, Zang WQ, Yang F, Han BM, Chen ST, Liu Y, et al. Spatial and temporal change patterns of net primary productivity and its response to climate change in the Qinghai–Tibet Plateau of China from 2000 to 2015. J Arid Land. 2020;12(1):1–17.10.1007/s40333-019-0070-1Search in Google Scholar
[17] Chen HM, Tang LY, Wang ZJ, Su M, Tian D, Zhang L, et al. Evaluating the protection of bacteria from extreme Cd(ii) stress by P-enriched biochar. Env Pollut. 2020;263(PA):114483.10.1016/j.envpol.2020.114483Search in Google Scholar PubMed
[18] Guo B, Zang WQ, Luo W. Spatial-temporal shifts of ecological vulnerability of Karst Mountain ecosystem-impacts of global change and anthropogenic interference. Sci Total Env. 2020;741:140256.10.1016/j.scitotenv.2020.140256Search in Google Scholar PubMed
[19] Guo B, Zang WQ, Yang X, Huang XZ, Zhang R, Wu HW, et al. Improved evaluation method of the soil wind erosion intensity based on the cloud–AHP model under the stress of global climate change. Sci Total Env. 2020;746:141271.10.1016/j.scitotenv.2020.141271Search in Google Scholar PubMed
[20] Feng J, Ding JL, Wei WY. A study of soil salinization in Weigan and Kuqa Rivers Oasis based on Albedo-MSAVI feature space. Rural Water Conserv Hydropower China. 2018;2:147–52.Search in Google Scholar
[21] Lu J, Zhang XJ, Ye BS, Wu H, Wang T. Remote sensing monitoring of salinization in Hetao irrigation area based on SI-MSAVI feature space. Remote Sens Land Resoucesr. 2020;32(1):169–75.Search in Google Scholar
[22] Wu LL, Lu Y, Zhou X, Chen QH. GIS analysis on spatial-temporal pattern evolution of land rocky desertification in northwestern Guangxi. Earth Environ. 2009;37(3):280–6.Search in Google Scholar
[23] Shi YC, Shu YG. Analysis on Karst rocky desertificaion temporal and spatial variation characteristics and driving factors – A case study of Qinglong county of Guizhou Province. For Resour Manag. 2017;1:135–43+ 152.Search in Google Scholar
[24] Wang ZX, Jiang YJ, Zhang YZ, Duan SH, Liu JC, Zeng Z, et al. Spatial distribution and driving factors of Krast desertification based on GIS and geodetectors. Acta Geograph Sin. 2019;74(5):1025–39.Search in Google Scholar
[25] Shi JN, Lu HY, Tang DS, Zhang DH. Correlation analysis between rocky desertification and slope degree in Karst area of Shaodong county. J Cent South Univ Forestry Technol. 2012;32(10):84–8.Search in Google Scholar
[26] Shen D, Guo JJ, Wang ZH, Chen JH. Sensitivity assessment of geological hazards based on hotspot analysis and geographic detectors. Environ Ecol. 2021;3(4):83–9.Search in Google Scholar
[27] Chang YB, Zhu R, Xiao SC, Li YP. Sandy land change from 1980 to 2015 in Alxa League, China and its driving factors. J Desert Reserch. 2020;40(6):82–90.Search in Google Scholar
[28] Wang JF, Xu CD. Geodetector: principle and prospective. Acta Geograph Sin. 2017;72(1):116–34.Search in Google Scholar
[29] Xia XQ, Tian QJ, Du FL. Retrieval of rock-desertification degree from multi-spectral remote sensing images. J Remote Sens. 2006;10(4):469–74.Search in Google Scholar
[30] Zhang PP, Hu YM, Li XZ, Xiao DN, Yin J, Li YB. Analysis of rocky desertification landscape pattern change in Karst plateau area based on GIS. Trans CSAE. 2009;25(12):306–11.Search in Google Scholar
[31] Zuo TA, Diao CT, Su WC, Sun XF, Guan DJ. Spatial and temporal evolution process and characteristics of rocky desertification in Bijie experimental area. Acta Ecol Sin. 2014;34(23):7067–77.10.5846/stxb201401140108Search in Google Scholar
[32] Guo B, Wei CX, Yu Y, Liu YF, Li JL, Meng C, et al. The dominant influencing factors of desertification changes in the source region of yellow river: climate change or human activity? Sci Total Env. 2022;813:152512.10.1016/j.scitotenv.2021.152512Search in Google Scholar PubMed
[33] Yang QY, Jiang ZC, Ma ZL, Luo WQ, Yin H, Yu QW, et al. Spatial variability of Karst rock desertification based on geostatistics and remote sensing. Trans CSAE. 2012;28(4):243–7.Search in Google Scholar
[34] Luo XL, Wang SJ, Bai XY, Tan Q, Ran C, Chen H, et al. Analysis on the spatio-temporal evolution process of rocky desertification in Southwest Karst area. Acta Ecol Sin. 2021;41(2):680–93.Search in Google Scholar
[35] Wu HW, Guo B, Fan JF, Yang F, Han BM, Wei CX, et al. A novel remote sensing ecological vulnerability index on large scale: A case study of the China-Pakistan Economic Corridor region. Ecol Indic. 2021;129:107955.10.1016/j.ecolind.2021.107955Search in Google Scholar
© 2022 Ye Wen et al., published by De Gruyter
This work is licensed under the Creative Commons Attribution 4.0 International License.
