Ecological Restoration Strategies for the Topography of Loess Plateau Based on Adaptive Ecological Sensitivity Evaluation: A Case Study in Lanzhou, China

Abstract

Existing studies related to ecological sensitivity evaluation are comparatively mature; however, few approaches are concerned with distinctive topographical conditions that enable response to specific environmental restoration requirements. In this paper, an adaptive ecological sensitivity evaluation framework is established according to a representative plateau topography (i.e., the Loess Plateau region) via a case study of Gaolan County, China. Through the process of factor selection, index system construction, formula adaption, factor weight determination by analytic hierarchy process (AHP) as well as the support of RS and GIS technology, the ecological sensitivity of the research region has been evaluated and analyzed in combination with land use types and typical gully problems. The results show that high sensitivity areas account for 11.16 percent of the total area in the research region and a tendency for fragmentation in spatial distribution. Forest lands with steep slopes in gullies’ upstream have the highest ecological sensitivity, the proportion of forest and garden land dominated by trees in the highest sensitivity area is around 80 percent. Evidence-based ecological restoration strategies are proposed in terms of the topography of Loess Plateau. This study shows possibilities to adapt existing sensitivity evaluation model for medium-scaled specific topography problems and provides useful clues as a basis for knowledge acquisition that can feed into spatial design, planning and restoration processes.

1. Introduction

Human beings are closely related to the natural ecological environment. With the development of the global economy, problems such as biodiversity reduction, aggravation of ecological disasters, and fragile ecological environment have become increasingly prominent [1]. Therefore, ecological environment restoration is one of the most important issues nowadays, especially for fragile ecological landforms such as mountains, wetlands and waterfront areas with complex terrains. Carrying out ecological restoration is an urgent problem to be solved at present. The Loess Plateau of China features a unique mountainous terrain. It is mostly found in the provinces of Gansu, Shaanxi, Ningxia, Shanxi, Henan and Qinghai, along the middle and upper reaches of the Yellow River [2], which is an extremely fragile ecological area [3]. The Loess Plateau’s ecological environment problem of soil erosion [4] is a typical example of this weak ecological environment.

Ecological sensitivity is defined as the adaptability of ecological factors to external pressure or external interference under the premise of no loss or reduction in eco-environmental quality [5]. Ecological sensitivity analysis is primarily used to measure the maximum capacity of an ecosystem to withstand external, natural or man-made disturbances, and its ability to recover after being disturbed by them, which provides a reliable scientific basis for ecological restoration [6]. By dividing the target region into different levels of sensitivity areas, evaluating the current situation of each sensitivity area, analyzing its causes and problems, and proposing corresponding strategies, ecological sensitivity evaluation can provide the premise and basis for analyzing and predicting the occurrence of ecosystem imbalance and environmental problems in the region [7]. The ecological restoration strategies proposed based on ecological sensitivity evaluation can more precisely address the actual ecological problems in a specific region.

A large volume of studies on ecological sensitivity evaluation has been carried out worldwide. In 1967, McHarg first proposed the ecological adaptability analysis model [8]. Then, Steiver optimized and proposed the ecological sensitivity evaluation model on this foundation [9]. Horne studied the ecological sensitivity of Australian rainforests to selective logging [10]. Kumark selected soil, temperature, net agricultural income and other factors from 271 regions in India in the past 20 years to build a mathematical model, obtained the sensitivity of agriculture to climate change, and then conducted a spatio-temporal evaluation [11]. Ouyang identified soil erosion, desertification, salinization, acid rain and other ecological problems in China, then analyzed the whole of China’s ecological sensitivity and spatial distribution [12]. Yan, Yuan, Yang and other researchers used the method of ecological analysis to analyze different problems of provincial, municipal, and natural reserves [13,14,15]. Li combined ecological sensitivity evaluation and mountain restoration strategies [7]. Roman Biek used ecological sensitivity analysis to explore amphibian decline [16]. Going by the above studies, we can see that ecological sensitivity evaluation has been applied in a variety of sectors lately, including climate and hydrology [17,18]. Ecological sensitivity evaluation is divided into two types based on the object: comprehensive ecological sensitivity evaluation and sensitivity evaluation of a single ecological environment problem [19] such as soil erosion or hydrological sensitivity. While the methodologies for medium and large-scale ecological sensitivity evaluations are more complex, small-scale evaluations tend to focus on individual factors.

The Loess Plateau’s ecological studies are mostly focused on large-scale issues [2]. Tong carried out an ecological sensitivity evaluation in the typical small watersheds of The Loess Plateau [19]. Mou summarized the engineering projects in the Yellow River basin and the changes to grassland, farmland, forests and other land types, and proposed the strategy for the overall ecological restoration of the Loess Plateau region [20]. Li and other researchers focused on the engineering measures and strategies [3,4]. In the Loess Plateau region, previous ecological restoration research has concentrated on two aspects. The first, large-scale ecological restoration planning that involves ecological restoration strategies for the entire Loess Plateau region or for a specific province or city [21], and is primarily carried out from the perspectives of soil and water conservation, soil erosion control, and ecological restoration evaluation [22].

The second aspect is the small-scale ecological restoration project, which is a construction project that focuses on a single gully or slope surface [23,24]. In the Loess Plateau, there is a lack of a mesoscale ecological restoration strategy that integrates an area having ecological problems with urban planning and design. Ecological sensitivity evaluation may serve as a foundation for ecological restoration, and the Loess Plateau’s unique topographic terrain necessitates ecological sensitivity to play an analytical role at the mesoscale. As a result, choosing the Loess Plateau’s mesoscale plot for research is critical in filling the research gap of ecological sensitivity evaluation in mesoscale and mountainous landforms.

Consequently, the goal of this research is to build an adaptive evaluation model that can be used to evaluate ecological sensitivity in mesoscale and special landforms and lay the foundation for ecological restoration in the Loess Plateau region by applying the model. This research has two main components. The first is to construct an ecological sensitivity evaluation system suitable for the Loess Plateau region and carry out ecological sensitivity evaluation. The second is to analyze the ecological sensitivity evaluation results and propose ecological restoration strategies suitable for the research region. These approaches provide theoretical basis and optimization strategies for future ecological restoration practices in the Loess Plateau and other special geomorphic regions.

2. Materials and Methods

2.1. Overview and Data

Lanzhou city’s land, which has typical Loess Plateau geomorphic features, is suitable for this research. The research region is in Gaolan County, on the edge of Lanzhou’s main urban area and adjacent to the Yellow River. The area of the research region is 105 square kilometers in size. Except for a portion of the southern region, which is developed as an urban residential area, most of the land remains in the form of natural villages, mainly farmland, forest land and wasteland, with dispersed settlements. As a result of severe debris flows in the research region, the ecological environment here is weak and urgently needs to be restored (Figure 1).

The topographic data utilized in this study are ASTER GDEM’s 30M resolution digital elevation data of 2009, which can be freely downloaded (https://www.gscloud.cn/sources/accessdata/310?pid=302, accessed on 20 January 2022). The high-resolution geographic remote sensing maps are Landsat 8 OLI TIRS satellite digital deliverables (https://www.gscloud.cn/sources/accessdata/411?pid=263, accessed on 25 December 2017). Since a geological hazard assessment has been conducted in the research region, fundamental geology, soil and geological disaster-prone area data can be found in the assessment report, which is used to redraw them as vector data in ArcGIS10.2. The current land use data are produced as vector data in ArcGIS10.2 from the image file of the 2016 Lanzhou City Land Resources Survey. By using ArcGIS10.2 software, all the data in this research are eventually unified under the WGS 1984 coordinate system.

2.2. Methodology

Ecological sensitivity evaluation is usually conducted according to the process of “factor selection—construction of index system—determination of factor weight—comprehensive evaluation and analysis” [6]. The whole research framework system is determined based on this process (Figure 2). First, based on available data and the background situation of the research region, current situation research is conducted to identify the present key problems and determine evaluation factors. The weights of the factors are then determined using appropriate methods, and ecological sensitivity calculations are then performed to obtain comprehensive ecological sensitivity evaluation results and ecological sensitivity areas of various levels, such as high sensitivity area, mid-high sensitivity area, mid-low sensitivity area, and low sensitivity area. Finally, to complete the application of the evaluation results, we analyze each level of area and offer matching ecological restoration strategies.

2.2.1. Construction of Index System

In previous studies, Tong selected five factors, which are slope, undulation, soil erosion, land use and vegetation cover, to analyze comprehensive sensitivity according to the actual condition [19]. Pan, in keeping with the main and potential ecological problems of Karamay, China, which has a similar climate to the Loess Plateau, subjectively selected the factors of natural environmental evolution (including soil erosion, land desertification, soil salinization and biodiversity), and the disturbance factor of human activity (oilfield exploitation) [25]. Li integrated practical situation and expert opinions, selected terrain, water buffer, vegetation, land use and road buffer to accomplish similar special terrain analysis [7]. In conclusion, there is no unified standard for selecting factors for ecological sensitivity evaluation [26], so it is necessary to follow the principles of science, comprehensiveness, operability, typicality and regionality, and make a concrete analysis of the specific situation [27]. Ouyang divided China into several sensitive regions, and the Loess Plateau’s sensitive region included the whole Loess Plateau zone with a semi-arid climate. The terrain is a unique loess tableland, ridge and hill. The texture of loess is relatively loose, so loss of soil and water is a very serious matter [12]. Therefore, this research begins with the current situation, incorporates the opinions of 15 experts in urban planning, landscape design, ecology and other related fields, and creates an index system according to the characteristics of the Loess Plateau and the features of the research region.

The main ecological problems in the Loess Plateau basin include the deterioration of water environment and ecology, the severe situation of soil erosion control, and the pressing need for biodiversity conservation [28,29]. Among them, soil erosion is the most serious problem in the Loess Plateau region [3,21]. The Loess Plateau’s loose soil, diverse terrain and steep slopes are the primary causes of soil erosion. Simultaneously, several studies have demonstrated that unreasonable land use and poor vegetation coverage, together with water resource limits, contribute to soil erosion [30]. As a result, the factors considered in this research include four aspects: topography, vegetation, land use, and water area.

Topography, which includes elevation, slope and aspect, among other factors, is the carrier of ecological environment [31]. The current land in the research region is of a typical Loess Plateau topography, mainly composed of ridges and hills. With an elevation of 1541.2–2025.1 m, the topography is high in the northeast and low in the southwest. Small basins with high height and flat topography dominate the north, whereas finger-like mountains dominate the south (Figure 3a). According to the causes of soil erosion and other problems, the elevation, slope, and aspect are selected as evaluation factors. Furthermore, it is anticipated that there is a geological disaster-prone area in the research region based on the content of the geological disaster report and the geological structure. Therefore, the factor is also included in the evaluation system to assure for ecological security.

As previously stated, the vegetation cover impacts soil erosion in the Loess Plateau region and is crucial to the stability of the ecosystem. The normalized difference vegetation index ($\mathrm{NDVI}$) may indicate the initial state and growing status of vegetation, and its calculation Formula (1) is as follows:

$$\mathrm{NDVI}=\frac{R_n-R_r}{R_n-R_r}$$

(1)

where $R_n$ is the reflectance of near-infrared band and $R_r$ is the reflectance of red band. In this research, satellite digital deliverables of Landsat 8 OLI_TIRS are selected for remote sensing image calculation in ENVI software, and the $\mathrm{NDVI}$ in the research region is finally obtained (Figure 3b).

The research region is typical of sparse vegetation in the Loess Plateau. A majority of the region’s current vegetation has been artificially replanted, and Yanchi Village in the south has been built into an urban residential area, making it lose part of its natural ecosystem function. Therefore, land use type is also taken into consideration in this research. Agricultural land, construction land and other land are divided into three categories in current land use documents. Agricultural land is subdivided into cultivated land, garden land, forest land, pastureland, facility agricultural land and pond. Construction land is divided into urban and rural construction land (including urban land, rural residential land and mining land), transportation and water conservancy land (including railway land, highway land and hydraulic construction land), and other construction land (including land for scenic facilities and special uses). Other lands include water areas, beaches and natural reserves (Figure 3c). In this research, current land use is reclassified according to the existing ecological environment. The land with the highest sensitivity is forest land, garden land, and other land mostly formed of trees and shrubs, whereas pastureland and cultivated land primarily composed of shrubs and herbs are classified as sub-high sensitivity. The third level is made up of bare land, pond (there are none in the research region) and facility agricultural land, while the fourth level is composed of construction land and other land.

In the research region, there is no water body, but the terrain is undulating and crisscrossed by the gullies. Rain and floods create more than 20 gullies, among which Dasha Gully, Shengou Gully and Xianshui Gully are the primary tributaries of the Yellow River, directly flowing into it. Because the region receives a lot of seasonal precipitation, and surface runoff has a lot of scouring and erosion, the gully can prevent and control natural disasters in this situation [32]. Therefore, the gully buffer area is included in the evaluation factor. Furthermore, the soil texture in the study area is collapsible loess, which has a consistent and fine texture, a loose structure, and considerable porosity. The granular structure will be destroyed by external force or gravity, and the soil will sink, making slope runoff more likely. There are still a considerable number of gullies generated by catchment runoff in the mountains, in addition to the more than 20 identified gullies in the current data. In this research, the hydrological analysis module of ArcGIS is used to extract catchment lines in order to conduct a complete analysis of the hydrological condition in the research region (Figure 3d). The establishment of buffer area is an essential way to protect water environment [33]. Hence, this research will also analyze the buffer area in addition to the gullies and catchment lines.

Finally, this study selects seven factors as the main factors of ecological sensitivity evaluation based on the examination of various graphic data and model data, as well as field inquiry and expert opinions. There is no criterion for the level of ecological sensitivity in the evaluation. In this research, the sensitivity of a single factor is divided into four levels, which are high sensitivity, mid-high sensitivity, mid-low sensitivity, and low sensitivity, and assigned values of 7, 5, 3 and 1, respectively, dominated by the impact mode and degree of factors on the ecosystem in the Loess Plateau. The construction of the ecological sensitivity evaluation index system is shown in

Table 1. Ecological sensitivity evaluation index system.

Index (Unit)		Evaluation Level of Ecological Sensitivity (Assigned Value)
		High Sensitivity (7)	Mid-High Sensitivity (5)	Mid-Low Sensitivity (3)	Low Sensitivity (1)
Topography	Dem (m)	≥1900	1800–1900	1700–1800	<1700
	Slope (%)	≥25	15–25	8–15	<8
	Aspect	South	Southeast, southwest, east	West, northwest, northeast	North
	Geological disaster-prone area	Geological disaster-prone area	--	--	Not geological disaster-prone area
Plant	NDVI	≥0.15	0.12–0.15	0.1–0.12	<0.1
Land	Land use	Forest land, garden land	Pastureland, cultivated land	Bare land, pond, facility agricultural land	Construction land, other land
Water	Gully buffer (m)	<10	10–50	50–100	>100

.

2.2.2. Determination of Factor Weight and Overlay Analysis

In terms of methodology, ecological sensitivity evaluation often starts with a single-factor evaluation using AHP, GIS technology, grey relational analysis, and other methods to conduct a single-factor systematic or multi-factor comprehensive evaluation [28].

In this research, AHP is used to determine the factor weight. AHP contains only two layers; the index layer is the selected factor, and the target layer is the comprehensive ecological sensitivity. The factor weight judgment matrix is built with AHP. Fifteen experts in urban planning, landscape architecture, ecology and related fields are selected and asked to make a pairwise comparison of the relative importance of ecological sensitivity indexes within the target layer, based on a comprehensive consideration of conditions such as the proportion of each single ecological factor covering the area of the research region, and the number of patches formed by nonlinear areas that are distinct from the surrounding environment. A 7 × 7 judgment matrix was constructed using the numerals 1–5 and their reciprocals as scales. The matrix column items are normalized and the arithmetic mean of each row is calculated, yielding the eigenvector ω of the matrix. The matrix’s consistency is then tested. Firstly, Formula (2) is used to obtain the matrix’s largest eigenroot

$$A\omega ={\lambda }_{max}\omega$$

(2)

where $A\omega$ is the product of the eigenvector $\omega$ and the matrix $A$, and ${\lambda }_{max}$ is the matrix’s largest eigenroot. After calculation, ${\lambda }_{max}$ = 7.7108, and then use Formula (3) to calculate the consistency index ($CI$)

$$CI=\frac{{\lambda }_{max}-n}{n-1}$$

(3)

where $n$ represents the order of the matrix. After calculation, the matrix $CI$ = 0.1185. Finally, Formula (4) is used to calculate consistency ratio ($CR$)

$$CR=\frac{CI}{RI}$$

(4)

thereinto, the consistency index RI of the 7 × 7 matrix is 1.36 [34] by checking the table, and the matrix $CR$ = 0.0871 by calculation. When $CR$ < 0.10, the consistency of the judgment matrix is considered acceptable and meets the requirements. Thus, we get the weight of each factor, as shown in

Table 2. Hierarchical judgment matrix and weight for ecological sensitivity evaluation.

Evaluation Factor	Elevation	Slope	Aspect	Vegetation Coverage	Land Use	Gully Buffer Area	Geological Disaster	Weight
Elevation	1	1/2	3	1/2	1/2	1/3	1/3	0.0759
Slope	3	1	5	4	3	2	3	0.2922
Aspect	1/4	1/5	1	1/3	1/4	1/5	1/5	0.0332
Vegetation coverage	2	1/4	2	1	1/3	1/3	1/5	0.0675
Land use	3	1/3	3	2	1	1/4	1/3	0.1030
Gully buffer area	5	1/2	5	2	3	1	3	0.2333
Geological disaster	4	1/2	4	4	4	1/3	1	0.1949

.

After determining the above single-factor selection, assigned value of sensitivity and weight, the single-factor ecological sensitivity evaluation of each factor is completed in ArcGIS10.2, and the spatial analysis tool in the software is used for overlay analysis according to Formula (5), and finally the comprehensive ecological sensitivity evaluation result of the research region is obtained.

$$W={\displaystyle \sum }_{i=1}^n\left(K_i\times C_i\right)$$

(5)

where $i$ is the number of evaluation factors; $n$ is the total number of evaluation factors. In this study, $n$ = 7. $K_i$ is the weight of the ith evaluation factor. $C_i$ is the ecological sensitivity evaluation score of the ith evaluation factor.

3. Results

3.1. Single-Factor Sensitivity Evaluation Results

Figure 4 depicts the evaluation results for elevation, slope, aspect, and geological disaster-prone area. The area with high elevation sensitivity is only 1.89 percent, which is found in Qianjiayao village in the north, Jiuhe village in the southwest, and Zhujiajing village in the south. The mid-high sensitivity is concentrated in the ridge with higher terrain in the north and the mountain near Zhujiajing village in the southeast, accounting for 38.34 percent of the total area of the research region. The mid-low sensitivity area and low sensitivity area, accounting for approximately 60 percent of the total area, are in the northwestern and southern areas where the terrain is relatively flat. The distribution of slope sensitivity and elevation sensitivity is similar to that of elevation sensitivity. High sensitivity makes up 3.22 percent of the overall area of the research region, which is mainly found at high altitudes with obvious slope changes as well as at the intersections of villages, roads or other construction land and mountains.

The distribution of gullies in the research region is initially retrieved by using the hydrological analysis module of ArcGIS, then optimized and adjusted by combining existing paper data. The gully sensitivity is determined according to the distance of each gully catchment line. The proportion of high sensitivity area to mid-low sensitivity area in the total distribution is minimal, but low sensitivity area distribution is more prevalent. The distribution of high sensitivity, mid-high sensitivity and mid-low sensitivity of vegetation coverage in the research region is relatively average. The high sensitivity areas cover 7.37 percent of the research region’s total area, whereas low sensitivity areas are concentrated along Dasha Gully in the south and in the area that has been used for urban residential development and construction. The high sensitivity areas of land use account for 8.64 percent of the total area of the research region, mainly distributed in the current garden and ecological public forest of each village. The urban green belt built during the construction of urban residential land near Caojiawan Village and Yanchi Village in the south is a high sensitivity area. The evaluation results of each single factor are shown in

Table 3. Evaluation results of single-factor ecological sensitivity.

Evaluation Factor	High Sensitivity		Mid-High Sensitivity		Mid-Low Sensitivity		Low Sensitivity
	Area (ha)	Proportion (%)	Area (ha)	Proportion (%)	Area (ha)	Proportion (%)	Area (ha)	Proportion (%)
Elevation	198.45	1.89	4025.70	38.34	5068.35	48.27	1207.50	11.50
Slope	338.10	3.22	2680.65	25.53	4123.35	39.27	3357.90	31.98
Aspect	982.80	9.36	3538.50	33.70	4357.50	41.50	1621.20	15.44
Vegetation coverage	773.85	7.37	5597.55	53.31	2936.85	27.97	1191.75	11.35
Land use	907.20	8.64	6654.90	63.38	2442.30	23.26	495.60	4.72
Gully buffer area	290.85	2.77	1174.95	11.19	1376.55	13.11	7657.65	72.93
Geological disaster	53.55	0.51	--	--	--	--	10,446.45	99.49

(Figure 5).

3.2. Comprehensive Ecological Sensitivity Evaluation Results

The comprehensive ecological sensitivity index of the research region ranges between 1 and 5.3044. The natural breaks classification method is adopted to reclassify the calculated results and, finally, the comprehensive ecological sensitivity evaluation results of the research region are obtained, as shown in Figure 6. The comprehensive ecological sensitivity evaluation results also suggest a significant fragmentation condition due to the research region’s mountainous topography and average vegetation coverage.

The high sensitivity areas, which are concentrated in Jiuhe Village and Jianshuigou Village in the southwest; Central Village in the north; and Sanping Village and Zhujiajing Village in the east, account for 11.16 percent of the total area in the research region. The mid-high sensitivity areas make up 29.37 percent of the total area in the research region, which are mostly found in the radiation zones of high sensitivity areas with well-dispersed distribution. The mid-low sensitivity areas account for 36.57 percent of the entire area of the research region, occupying the largest proportion of the area, and the distribution is dispersed, covering all villages in the research region, particularly Tougou Village and Sanping Village, which were the most extensively distributed. The low sensitivity areas account for 22.90 percent of the total area of the research region, which are found in the north and east, with a concentration in the urban residential area in the south.

3.3. Analysis of Ecological Sensitivity Evaluation Results under Different Land Uses

Each level of ecological sensitivity area is retrieved from the results of the comprehensive ecological sensitivity evaluation, and the results are displayed in Figure 7. The analysis of the ecological sensitivity evaluation results is conducted under three categories of land use, which are construction land formed by the combination of the third and fourth-level land; forest land and garden land represented by trees as the highest level land; and pastureland and cultivated land represented by shrubs and herbs as the second-level land, according to the classification principle in the single-factor analysis of land use types. ArcGIS overlay analysis can be used to obtain the distribution of different levels of ecological sensitivity area under diverse land uses, as illustrated in Figure 7.

High sensitivity areas have a fragile ecological environment, which are in urgent need of key protection. High sensitivity areas can be divided into two categories. One is located in mountain areas of high altitude, with complex topographic slope changes and landslide hazards. The other is located in the gully and its primary buffer area. The mid-high sensitivity areas, which serve as a transition zone between the high sensitivity area and the construction land, play a vital role in sustaining the ecological stability of high sensitivity areas. The combined size of the two areas is more than 40 percent of the total area of the research region, and the proportion of forest and garden land dominated by trees in each area is around 80 percent of each area (Figure 8).

The mid-low sensitivity areas are mainly the secondary and tertiary buffer areas of gullies and the natural green land of gentle slope hills. Low sensitivity areas encompass roughly 24 square kilometers, and are mostly found in built-up areas and farmland (Figure 9). The low sensitivity areas are mainly concentrated in the relatively flat areas and urban residential areas in the lower reaches of Dasha Gully in the south, interspersed with villages that are scheduled to be converted into urban residential areas. However, according to the results of field investigation, serious ecological problems exist in low sensitivity areas of the Loess Plateau, such as single vegetation level and low farming efficiency.

3.4. Ecological Sensitivity Analysis in Gully Buffer Area

The buffer areas of the existing 20 gullies in the research region are overlaid with the results of the comprehensive ecological sensitivity evaluation to analyze the sensitivity situation within the 100 m buffer areas of the gullies, and the results are shown in Figure 10. The upstream portions of each gully, such as Dasha Gully, Shengou Gully, Yangchenjia Gully and Xianshui Gully, account for 11.16 percent of the total area of the gully buffer areas, which are largely spread in the upstream areas of each gully. Some gullies, such as Huoshao Gully and Xiaosha Gully, have complicated terrains and are wholly located in high sensitivity areas. Mid-high sensitivity areas make up 29.37 percent of the entire gully buffer areas, mostly found within the 10–50 m buffer areas and the downstream areas of some gullies. Mid-low sensitivity areas, which are mostly located within the 50–100 m buffer areas and outside the 100 m buffer areas of some gullies, account for 36.57 percent of the entire gully buffer areas. Low sensitivity areas, which are concentrated in the downstream area of Dasha Gully and scattered outside the 100 m buffer area of each gully, account for 22.90 percent of the entire gully buffer areas.

According to the above analysis, comprehensive results of ecological sensitivity evaluation in the research region show an overall tendency of fragmentation in spatial distribution, with the highest ecological sensitivity in the upper reaches of the gullies and the forest lands with steep slopes. However, due to the poor ecological environment in the Loess Plateau region, low sensitivity areas also have problems such as single vegetation level and low farming efficiency.

4. Discussion

4.1. Genetic Analysis of High Sensitivity Regions

Based on the single-factor analysis of high sensitivity regions in the research site, it can be seen that the main factors leading to high sensitivity are slope aspect, vegetation and land use. Although the weight assigned to slope direction is only 0.0332 in our calculation due to its large area, a large number of high sensitivity aspect regions are distributed in the highly sensitive area (Figure 11).

According to the comprehensive sensitivity analysis results, the high sensitivity regions are mainly distributed among Jiuhe Village, Jianshuigou Village, Sanping Village and Zhujiajing Village. We further analyzed the single-factor situation of each village, as shown in Figure 12. The actual situation of Sanping Village and Zhujiajing Village in the east is mostly hilly landform, and the area of high sensitivity regions in the site is small. However, there is land at a high altitude in the eastern edge of the site, so the main reasons for high sensitivity are elevation, slope and disaster area. Jiuhe Village and Jianshuigou Village now have more woodland inside, and their high sensitivity is mainly due to slope, landuse and NDVI. The high sensitivity regions scattered in other parts of the research site are mainly caused by aspect and gullies. Under the condition of intense precipitation, the whole situation of vegetation, slope and land use will change in the gully development zone, resulting in serious soil erosion. Because the terrain of Loess Plateau has aspect variation to a certain extent, the aspect within the research site is not aggregated but distributed over a wide area. The scattered and abundant distribution of gullies and terrain is also the main reason why the research region and Loess Plateau as a whole show highly sensitive distribution and dispersion, and each area needs different strategies for restoration.

4.2. Ecological Restoration Strategies for Different Levels of Ecological Sensitivity Area

Ecological restoration is the process of assisting in the restoration of ecosystems that have been degraded, damaged or destroyed [35]. Forest land protection and enhancement, as well as gully restoration, are possible in high sensitivity areas. In terms of forest land protection and enhancement, micro-interventions mixed with natural growth patterns can be performed. To improve vegetation coverage and achieve ecological conservation, the original ecological public welfare forests and gardens could be safeguarded, and natural reserves could be greened and restored [36,37]. For mid-high sensitivity areas and mid-low sensitivity areas, it is possible to give full play to the function of soil and water conservation of native vegetation on the slope, and adopt external soil spray seeding technique and soilless sod [38] to implement ecological restoration of the slope. For construction land in low sensitivity areas, the potential of growing diverse plants should be expanded, so multi-level vegetation of trees, shrubs and grasses with good landscape effect may be considered [39]. On the one hand, it can improve the stability of the plant community, optimize ecosystem structure and enhance biodiversity. On the other hand, it can enrich the effect of landscape beautification. The recycling of water resources can be fully considered for farmland in low sensitivity areas to alleviate soil erosion of the Loess Plateau caused by water resource constraints. Using native tree species to re-green existing depressed mountains is also an option. For example, in the research region, trees such as Populus tomentosa, Robinia pseudoacacia and Pinus tabuliformis can be selected, and shrubs and grass plants such as Amorpha fruticosa, Tamarix ramosissima and Caragana tibetica KOM. can better adapt to the local environment and reduce conservation costs [40].

4.3. Ecological Restoration Strategies for the Gully

Dasha Gully is the largest gully in the site. It is used as an example to provide specific ecological restoration strategies for gullies based on the results of comprehensive ecological sensitivity evaluation. The total length of Dasha Gully is about 13 km. Its upstream reach is dominated by high ecological sensitivity areas, the middle reach by mid-high sensitivity and mid-low sensitivity areas, and the lower reach by mid-low sensitivity and low sensitivity areas. Therefore, the gully can be divided into three reaches: upstream reach, middle reach and lower reach as per ecological sensitivity (Figure 13).

The upstream reach of Dasha Gully is located in an ecological conservation area or a high-altitude area with steep terrain. The mountains along the gully and the source of the gully can be regreened through ecological measures such as plateau preservation, slope protection and gully fixation. The middle reach of Dasha Gully is mostly hilly with gentle slopes, so the slopes can be repaired to prevent soil erosion, and wet-tolerant and drought-tolerant vegetation can be selected in the buffer area of the gully for greening construction, so as to realize ecological flood control and water retention transmission [37,40,41]. At present, the lower reach of Dasha Gully has been developed to a great extent. It can be designed as a park in line with the current situation to achieve both ecological restoration and landscape utilization.

5. Conclusions

Taking the site in Lanzhou as an example of the Loess Plateau, this research develops an ecological sensitivity evaluation model for special topography and a suitable ecological sensitivity evaluation system. Based on the ecological sensitivity evaluation of the research region, we reach the following conclusions:

(1)

We obtain four levels of area, including high sensitivity area, mid-high sensitivity area, mid-low sensitivity area and low sensitivity area, through ecological sensitivity evaluation. High sensitivity regions account for 11.16 percent of the total area in the research region.

(2)

Comprehensive results of the ecological sensitivity evaluation in Lanzhou show a tendency for fragmentation in spatial distribution. The proportion of forest and garden land dominated by trees in each area is around 80 percent of the total area of high sensitivity areas.

(3)

This research combines land use with a specific analysis of the problems, with special ecological elements such as gullies on a small and medium scale, and puts forward targeted ecological restoration strategies.

For researchers, this study fills in the knowledge gap of ecological sensitivity evaluation methods in the Loess Plateau region, and provides ideas for related research in the future. This research also offers ecological restoration strategies for designers to apply in special mountainous topography areas such as the Loess Plateau.

There are two major limitations to this research. First, only relevant examples are used as the research object, and factor selection and weight determination are performed using the expert scoring method, making it somewhat subjective. In other types of mountainous topography analysis, the model and parameters might be adjusted to generate more accurate results based on local characteristics. Second, this research focuses primarily on the current state of ecological sensitivity analysis and ignores the temporal dimension. On the premise of sufficient data, the historical evolution might be analyzed. If the temporal and spatial evolution form of the gully can be determined, the development of gully would be better simulated to implement ecological restoration. The above issues can be further explored and supplemented in future studies.