**2. Materials and Methods**

#### *2.1. Study Area*

Yuxi City is in the center of Yunnan Province on the southwest border of China, between latitude 23◦19 ~24◦53 north and longitude 101◦16 ~103◦09 east (Figure 1a). It is a prefecture-level city under the jurisdiction of Yunnan Province, bordering the provincial capital Kunming City in the north, Honghe Prefecture in the southeast, Pu'er City in the southwest and Chuxiong Prefecture in the northwest (Figure 1b). The terrain is high in the northwest and low in the southeast, with staggered distribution of mountains, canyons, plateaus, and basins. The west is mainly a deep-cut alpine valley landform. The central and eastern regions belong to the mountains of central Yunnan Province, which is dominated by a middle mountain; the terrain of most areas is undulating in the shape of waves, and there are many intermountain basins of different sizes scattered among the mountains. The eastern region is mainly dominated by plateau lake-basin landforms, with three plateau faulted lakes, Fuxian Lake, Xingyun Lake and Qilu Lake. Around three lakes, Chengjiang, Jiangchuan and Tonghai, lacustrine basins are formed, with flat and open terrain in the basins.

**Figure 1.** Overview of study area. (**a**) the location of Yuxi City in China; (**b**) the elevation of Yuxi City; (**c**) the landscape grid of mountainous areas and flatland areas.

Yuxi City is located in the low-latitude Yunnan Plateau, with long hours of sunshine and abundant heat. It belongs to the subtropical plateau monsoon climate. Due to the great difference in elevation of the terrain, the three-dimensional climate is clear. Under the comprehensive influence of the Indian Ocean and the Gulf of Tonkin, the climate in most areas is mild, with distinct dry and wet seasons, no severe cold in winter and no severe heat in summer. The annual average temperature is 15.4–24.2 ◦C, and the annual precipitation is 787.8–1000 mm, which mostly occurs from June to October, with heavy rain mainly from June to August. Due to the complex terrain and the great elevation difference, the rainfall is heavier, and the temperature is lower in the mountainous area than in the flatland area. From the top of the mountain to the bottom of the valley, the temperature difference is significant throughout the year and between day and night.

#### *2.2. Data Source and Processing*

In this study, six Landsat satellite remote sensing images in 1995, 2000, 2005, 2010, 2015 and 2018 were downloaded from the geospatial Data Cloud website (www.gscloud.cn, accessed on 10 June 2019), mainly including Landsat 5 TM images, Landsat 7 ETM+ images and Landsat 8 OLI images. The selection of remote sensing images was based on the premise of little cloud and good quality. Winter and spring were selected as the image months, and the imaging times of the remote sensing images were all in January, February, and March of each year. The cloud content (CC) of all images was less than 10%, and the cloud content of most images was less than 1%. Based on the ArcGIS 10.8 software, the maximum likelihood classification method was used to classify landscape units. Referring to the standard of "Classification of Land Use Status" and combining with the research needs of landscape patterns in Yuxi City, the landscape units were divided into five types: cultivated land, forest and grass, construction land, water area and unused land (Figure 2). Kappa coefficients are all above 80%, and the accuracy meets the research requirements.

**Figure 2.** Interpretation of landscape types in Yuxi City from 1995 to 2018. (**a**) 1995; (**b**) 2000; (**c**) 2005; (**d**) 2010; (**e**) 2015; (**f**) 2018.

Based on the terrain characteristics, area size and elevation of Yuxi City, 500 m × 500 m, 1.5 km × 1.5 km and 3 km × 3 km grids were constructed as preselected evaluation units. With the help of analysis tools such as ArcGIS 10.8 software Create Fishnet, Dissolve, Clip and Merge, a total of 7570 grids with a size of 1.5 km × 1.5 km were finally determined. Therefore, the equal space system sampling method was used to divide the study area into 1.5 km × 1.5 km landscape plots, which were used as the basic unit for the study of landscape pattern evolution. There were 1775 landscape plots in the flatland area and 5795 landscape plots in the mountainous area (Figure 1c). The raster maps of land use in mountainous areas and flatland areas from 1995 to 2018 were cropped to obtain the raster data of landscape units over the years at the grid scale for the subsequent calculation of the landscape index. Based on the data of landscape units in each grid, the landscape index of each grid was calculated by Fragstats 4.2 software (Oregon, USA).

#### *2.3. Index Calculation*

By referring to the relevant literature [38] and combining with the unique characteristics of the mountain–flatland, landscape indices were selected respectively from the class level and landscape level to characterize and analyze the landscape pattern of Yuxi City. Plaque Density (PD), Edge Density (ED), Largest Plaque Index (LPI), and Mean Plaque Area (Area\_MN) were selected at the class level. At the landscape level, Landscape Shape Index (LSI), Largest Patch Index (LPI) and Shannon Diversity Index (SHDI) were selected. The calculation formula and ecological significance of each landscape index are shown in the reference [23,39].

The Fragstats 4.2 software was used to calculate the landscape pattern index at the class level first, and excel software was used for statistics and mapping analysis. The landscape indices of each landscape plots in the mountainous area and the flatland area in six periods was then calculated by using the processed raster data of landscape units over the past years, and the spatial distribution maps of landscape indices were obtained. Finally, this study divides each landscape index into five levels according to the natural breakpoint method, which are represented by I, II, III, IV and V, respectively.

#### **3. Measurement and Comparison of Landscape Pattern Change at Class Level**

#### *3.1. Variation Characteristics of Landscape Pattern at Class Level in Mountainous Areas*

During the study period, at class level, the variation in the landscape index in the mountainous area was significantly different (Figure 3). The PD and ED of cultivated land increased first and then decreased at the turning point of 2000, and the trend of the Area\_MN was in an "S" shape, while the LPI continued to decline. The PD and ED of forest and grass, water area and construction land increased first and then decreased, while the Area\_MN first decreased and then increased. The patches number and density of forest and grass decreased and the Area\_MN of construction land changed only slightly. The ED of cultivated land, forest and grass was higher than that of the water area, construction land and unused land. The LPI and the Area\_MN of forest and grass were the largest, which had a great influence on other landscape types.

#### *3.2. Variation Characteristics of Landscape Pattern at Class Level in Flatland Areas*

During the study period, the PD of cultivated land in the flatland area increased continuously, the Area\_MN decreased significantly, and the ED first increased and then continued to decrease, but it always ranked in first place among all regions, while the LPI continued to decline (Figure 4). As the first dominant type, forest and grass, the Area\_MN first decreased and then increased, the PD and ED first increased and then decreased, and the LPI continued to decrease, indicating that the dominant position of large patches in the landscape gradually decreased. The PD and ED of the water area first increased and then decreased, and the degree of fragmentation gradually increased; the LPI changed little during the study period, but the Area\_MN showed a downward trend. The Area\_MN of construction land reached its lowest level in 2005, and then continued to rise with significant changes, while the LPI also increased significantly. However, the PD and ED first increased and then decreased with 2005 as the turning point, suggesting that the distribution of construction land was relatively scattered before 2005, but that after 2005, the construction land patches gradually concentrated with contiguous distribution.

**Figure 3.** Landscape index at class level in mountainous areas of Yuxi City from 1995 to 2018. (**a**) Patch Density (PD); (**b**) Edge Density (ED); (**c**) Mean plaque Area (Area\_MN); (**d**) Largest Patch Index (LPI).

Combined with the implementation time of regional land management policies, the changes in the landscape indices of cultivated land and forest and grass in mountainous and flatland areas were mainly affected by the policy of returning cultivated land to forest and grass during the study period. The overall fragmentation degree of construction land and forest and grass in the flatland area is clearly higher than that in the mountainous area, and the layout of the construction land is more concentrated. This is mainly because the social and economic development level of the flatland area has improved rapidly since 2005, and the urban land has expanded significantly. However, due to the migration of populations to the flatland area and for other reasons, the farmland in the mountainous area was abandoned, while the forest and grass became more concentrated and contiguous.

**Figure 4.** Landscape index at class level in flatland areas of Yuxi City from 1995 to 2018. (**a**) Patch Density (PD); (**b**) Edge Density (ED); (**c**) Mean plaque Area (Area\_MN); (**d**) Largest Patch Index (LPI).

#### **4. Measurement and Comparison of Landscape Pattern Change at Landscape Level**

*4.1. Variation Characteristics of Landscape Pattern at Landscape Level in Mountainous Areas* 4.1.1. Landscape Diversity

During the study period, the landscape diversity in the mountainous area showed obvious phased characteristics, and the change during the period 1995–2005 was significantly higher than that during the years 2005–2018. The proportion of the SHDI was the highest in class II and III, the proportion of the low value area decreased, while the proportion of the high value area increased (Figure 5a). The area where the SHDI increased was greater than the area where the SHDI decreased in each period. From 1995 to 2018, the area where the index increased accounted for 26.74% of the total, while the index reduction areas accounted for 5.48%, indicating that the landscape diversity showed an increasing trend year by year. The high value area of the landscape diversity index mainly concentrated in the relatively low flat areas of the east and the north, where the main landscape units are cultivated land, forest and grass, while the low value area is mainly located in the central and northwest Ailao mountain area, where the main landscape unit is forest and grass (Figure 6).

**Figure 5.** Graded area proportion of landscape index. (**a**) graded area proportion of SHDI in mountainous areas; (**b**) graded area proportion of SHDI in flatland areas; (**c**) graded area proportion of LSI in mountainous areas; (**d**) graded area proportion of LSI in flatland areas; (**e**) graded area proportion of LPI in mountainous areas; (**f**) graded area proportion of LPI in flatland areas.

#### 4.1.2. Landscape Shape

Over the years, in the mountainous area, the LSI of grade I and II accounts for a relatively high proportion, and the combined proportion of the two is as high as over 70%, while the proportion of grade IV and V is about 14%, indicating that the landscape shape is relatively simple (Figure 5c). In terms of time change, the proportion of grade I showed a downward trend, while the proportion of other levels showed an upward trend. The area where the LSI increased accounted for 23.85% of the total, while the index reduction area accounted for 3.14%, indicating that from 1995 to 2018, the LSI continued to increase, and the landscape shape tended from simple to complex. However, from 2005 to 2010 and 2010 to 2015, the index increased area was smaller than the index decreased area, indicating

that the landscape in these two periods was clearly contiguous, and the shape tended to be simple (Figure 7). During the study period, the areas with increased LSI were mainly distributed in the northern and southwestern regions, while the areas with decreased LSI were mainly concentrated in the eastern regions.

**Figure 6.** Spatial distribution of SHDI in mountainous areas of Yuxi City from 1995 to 2018. (**a**) 1995; (**b**) 2000; (**c**) 2005; (**d**) 2010; (**e**) 2015; (**f**) 2018.

**Figure 7.** Spatial distribution of LSI in mountainous areas of Yuxi City from 1995 to 2018. (**a**) 1995; (**b**) 2000; (**c**) 2005; (**d**) 2010; (**e**) 2015; (**f**) 2018.

#### 4.1.3. Largest Patch

From 1995 to 2018, the LPI levels in mountainous areas were mainly II, III and IV, but the proportion of I, II and III level increased, while the proportion of IV and V level decreased (Figure 5e). In each period, the area of the LPI that increased was much less than the decreasing area; the area of the increased LPI accounted for 7%, and the area of the index decreased accounted for 24%. The high value area of the LPI and the change in the LPI mainly occurred in the northern, central and western regions, with large topographic relief and high altitude. The low value area of the LPI with little change was mainly distributed in the eastern region, mainly in the relatively low and flat terrain area (Figure 8). From 1995 to 2018, the LPI decreased significantly in the central and western regions, especially in the northern part of the country, mainly because the forest and grass cover in the northern Yimen County was cut by a large amount of arable land, which resulted in serious fragmentation and rapid decline of the LPI.

**Figure 8.** Spatial distribution of LPI in mountainous areas of Yuxi City from 1995 to 2018. (**a**) 1995; (**b**) 2000; (**c**) 2005; (**d**) 2010; (**e**) 2015; (**f**) 2018.
