Vegetation status parameters are continuously varying and can accurately describe the minute differences in vegetation status in the same land use/cover category. Although these three vegetation status parameters have different definitions, they have relatively high correlations because they are all related to the vegetation status. The PCA approach is used in this study to reduce the correlations. PCA is an effective approach for compressing information in multidimensional data by calculating an orthogonal projection that maximizes the variance in the data (Ifarraguerri and Chang, 2000). In this study, the main objective of PCA is to find a lower-dimensional representation that can account for most of the variance in NPP, LAI, and NDVI. Before the PCA, all three vegetation status parameters should first be dimensionless and normalized with Eq. (1) to avoid the influences caused by an imbalance in units. Then, the comprehensive assessment result of the vegetation status can be calculated with Eq. (2) as follows:
where NIi is the normalized vegetation status value in pixel i, Ii is the original value in pixel i, Imax is the original maximum value in all pixels, Imin is the original minimum value in all pixels, VS is the comprehensive assessment result of the vegetation status, and NINPP, NILAI, and NINDVI are the normalized NPP, LAI, and NDVI values, respectively.
An ecosystem service is a comprehensive multifunctional indicator that includes the supply, regulation, support, and cultural services provided by ecosystems (Millennium Ecosystem Assessment, 2005; Hu et al., 2015). Costanza et al. (1997) defined ecosystem service as the flows of materials, energy, and information from natural capital stocks that are combined with manufactured and human capital services to produce human welfare. Daily (1997) considered ecosystem services as the conditions and processes of natural ecosystems that fulfill human life. In Brauman and Daily (2008) and Douglas (2015), ecosystem services are the benefits human beings derive from natural systems, including the provision, regulation, and cultural services that directly affect people and the supporting services needed to maintain other services. Based on the research of Costanza et al. (1997), Xie et al. (2008) provided a detailed look-up table (Table 1) of the equivalent values of ecosystem services based on land use/cover data in China from a survey of > 700 ecologists. As seen in Table 1, the equivalent values of ecosystem services are evaluated for four primary ecosystem service types: supply services, regulation services, support services, and cultural services. Each primary ecosystem service type includes one or more subtypes. Different services have different weights: the largest score for supply services is 2.98, which is from the raw materials provided from forestland; the largest score for regulation services is 18.77, which is from the water regulation provided by aquatic areas; the largest score of support services is 4.51, which is from the biodiversity protection support provided by forestland; and the largest score for cultural services is 4.69, which is from the recreation and culture opportunities provided by wetlands. Obviously, regulation services have the highest weight, while supply services are not as important as the other services. Table 1 displays the relationships between land use/cover and human activities, natural resources, ecological protection, and climate change. Among these relationships, human activities appear to be more of a burden because the scores of cultivated land and construction land are both very low.
Type Subtype Forestland Grassland Cultivated land Wetland Water area Unused land Construction land Supply services Food 0.33 0.43 1.00 0.36 0.53 0.02 0 Raw materials 2.98 0.36 0.39 0.24 0.35 0.04 0 Regulation services Gas regulation 4.32 1.50 0.72 2.41 0.51 0.06 0 Climate regulation 4.07 1.56 0.97 13.55 2.06 0.13 0 Water regulation 4.09 1.52 0.77 13.44 18.77 0.07 0 Waste treatment 1.72 1.32 1.39 14.40 14.85 0.26 0 Support services Soil formation and retention 4.02 2.24 1.47 1.99 0.41 0.17 0 Biodiversity protection 4.51 1.87 1.02 3.69 3.43 0.40 0 Cultural services Recreation and culture 2.08 0.87 0.17 4.69 4.44 0.24 0 Total 28.12 11.67 7.90 54.77 45.35 1.39 0
Table 1. Equivalent value per unit area of ecosystem services in China (Xie et al., 2008)
To overcome the disadvantage of “one ecosystem service value for one land use/cover category”, the improved equivalent values of ecosystem services will not be constant but will linearly change with changes in vegetation status. The original equivalent values of the ecosystem services in Table 1 are assumed to represent the average cases for each land use/cover type, which means that the higher vegetation status values will have higher equivalent values and the lower vegetation status values will have lower equivalent values even in the same land use/cover type. The improved equivalent ecosystem service value evaluation model should be expressed by Eq. (3) as follows:
where EVimpr,i is the improved equivalent value of an ecosystem service of land use/cover type i, EVoriginal,i is the original equivalent value of an ecosystem service of land use/cover type i (based on Table 1), VSvalue is the vegetation status value, and VSmean,i is the mean vegetation status value of land use/cover type i. Figure 2 illustrates the ﬂowchart of the improved equivalent ecosystem service value evaluation model.
3.1. Principal component analysis (PCA) of vegetation status parameters
3.2. Improvement in the equivalent values of ecosystem services
Figure 3 shows the three normalized vegetation status parameter maps for China in 2018: Fig. 3a is the NPP map, Fig. 3b is the LAI map, and Fig. 3c is the NDVI map. As seen in Fig. 3a, the values in the NPP map in western China and Inner Mongolia areas are mostly lower than 0.1, and the values in the Greater Hinggan, the Lesser Hinggan, the Changbai mountain ranges, and southern China are relatively higher but still lower than those in the LAI map and the NDVI map. Compared to the values in the LAI map and the NDVI map, the values in the NPP map are generally lower because the original NPP (before normalization) data have a large range of values from 0 to over 2000 gCm−2, but most of the values are lower than 1000 gCm−2. As seen in Fig. 3b, the values in the LAI map in western China and Inner Mongolia areas are close to those in the NPP map because low vegetation cover areas and desert areas mainly exist there. However, the values in southern China in the LAI map have a clearer distinction compared to those in the NPP map and the NDVI map, and the values in these last two are very close. In the LAI map, the values in southeastern Tibet, southwestern Yunnan Province, the southeastern coastal regions, Hainan, and Taiwan are clearly higher than those in other places, demonstrating the higher vegetation cover and greater potential of the ecosystem services in these areas. As seen in Fig. 3c, in comparison to the values in the other maps, the values in the NDVI map are generally higher while having less distinction in southeastern Tibet, southwestern Yunnan Province, the southeastern coastal regions, Hainan, and Taiwan. This is because NDVI tends to be saturated when biomass increases to a high value (Myneni et al., 1995; Hao et al., 2008), which is especially obvious in evergreen forest areas (Rahman et al., 2001). Overall, the LAI map performs better in the northeastern, central, and southern regions of China than in the other areas because the map clearly distinguishes the high and middle vegetation cover regions. Nevertheless, the NDVI map performs better in western China because it can clearly distinguish the low vegetation cover and desert regions.
Figure 3. Three normalized vegetation status maps: (a) NPP map, (b) LAI map, and (c) NDVI map for China in 2018.
Table 2 shows the eigenvalues of each principal component. The results demonstrate that the first two principal components contain over 96% of the vegetation status information, which means that these two components could reasonably be substituted for the original three vegetation status parameters. Eqs. (4) and (5) show that the first and second principal components contain most information from the LAI and NDVI maps.
Component Initial eigenvalue Variance (%) Cumulative (%) 1 85.9954 85.9954 2 10.0202 96.0156 3 3.9844 100.0000
Table 2. The eigenvalues of the PCA
Figure 4 is the comprehensive assessment result map of the vegetation status after the PCA, in which higher vegetation status values indicate a greater potential for ecosystem services, and lower vegetation status values indicate a lower potential for ecosystem services. As seen in Fig. 4, the low value regions are mainly distributed in Inner Mongolia, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang, where the vegetation status values are normally under 0.20 because many low vegetation coverage areas and deserts (such as the Taklimakan Desert and the Gurbantunggut Desert) exist due to the high altitudes and dry climates. Slightly higher value regions include the North China Plain, the Sichuan Basin, the Yangtze River Delta (mainly in Shanghai and parts of Jiangsu, Anhui, and Zhejiang provinces), the Pearl River Delta in Guangdong Province, the middle reaches of the Yangtze River, and the Northeast China Plain (including southwestern Heilongjiang, the eastern part of Inner Mongolia, and the western parts of Jilin and Liaoning provinces), where the main grain production areas and urban agglomeration areas of China exist. One of the reasons for the lower vegetation status values in these regions is the tremendous pressure from socioeconomic development and urbanization (Chen et al., 2019b). High value regions are mainly located in high forest coverage areas, such as the Greater Hinggan, the Lesser Hinggan, and the Changbai mountain ranges in the northeast, and Fujian, Jiangxi, Guangxi, Zhejiang, Taiwan, Guangdong, Hunan, Hainan, Yunnan, and Guizhou, which are the areas with the highest forest coverage rates in China. Figure 4 displays the differences in not only the different land use/cover types but also the same land use/cover type, such as the differences in the Northeast Plain and in the Middle-Lower Yangtze Plains, which are both mainly covered by cultivated land and different crop types, and the differences in the Greater Hinggan and Fujian Province, which both have high forest coverage areas but different forest types, growing years, canopy densities, etc.
Based on Figs. 1, 4, we can quantitatively determine the vegetation status differences in the same land use/cover data. Figure 5 shows the vegetation status differences in cultivated land, forestland, grassland, wetland, and unused land. We can see that except for the water areas and construction lands, which are less affected by the vegetation status, the other five land use/cover types are all apparently affected by vegetation status. The upper and lower endpoints are the largest and smallest vegetation status values of each land use/cover type, respectively, and the rhombus points represent the mean values. As seen in Fig. 5, the largest and smallest vegetation status values in the five land use/cover types are 0.85 and 0.21 (cultivated land), 1.00 and 0.26 (forestland), 0.70 and 0.11 (grassland), 0.51 and 0.10 (wetland), and 0.48 and 0.08 (unused land), respectively. The largest difference between the maximum and minimum is 0.74 in forestland, and the smallest difference is 0.40 in unused land. The large vegetation status differences in the same land use/cover type show that evaluating the ecosystem service values with one value in one land use/cover type is inadequate. Figure 6 shows the improved equivalent value map of the ecosystem service values based on both the land use/cover type and the vegetation status calculated by Eq. (3). Figure 6 (left) is the traditional equivalent value map of ecosystem services based only on the land use/cover type, and Fig. 6 (right) is the improved equivalent value map of ecosystem services based on both the land use/cover type and the vegetation parameters. As seen in Fig. 6, the left map has only seven values because the land use/cover types are divided into seven primary types; however, the values in the right map continuously vary between 0 and 57 because the vegetation status values continuously vary. In comparing the right and left maps, there are two main advantages in the improved evaluation model. One advantage is that the equivalent values of the ecosystem services in Fujian, Jiangxi, Guangxi, Zhejiang, Taiwan, Guangdong, Hunan, Hainan, Yunnan, and Guizhou are apparently higher than those in the Greater Hinggan, the Lesser Hinggan, and the Changbai mountain ranges in the northeast, although both of these regions are covered by forests. As mentioned above, forests in the first group of areas have higher vegetation status values than those in the second group of areas, which leads to the differences in the equivalent values of the ecosystem services between the northeastern forestland and the southern forestland in China. The other difference lies in the transitional zones between the different land use/cover types, such as the transitional zones between the grassland and forestland in eastern Inner Mongolia and the eastern Qinghai–Tibet Plateau. There is no transitional zone, but only a steep change in values in the left map; however, many smoothly changing transitional zones exist in the right map. As seen in Fig. 6 (right), the equivalent values of the ecosystem services gradually increase from unused land to grassland to forestland in the eastern Qinghai–Tibet Plateau. In particular, the equivalent values of ecosystem services in grasslands are different. Closer to unused land, the values of grasslands are smaller, and closer to forestland, the values are larger, showing the natural spatial variation trend in the vegetation status and ecosystem services. Overall, the distribution trend in the equivalent values of ecosystem services in Fig. 6 (right) appears similar to a combination of those in Figs. 1, 4, which shows the differences not only in the different land use/cover types but also in the same land use/cover type. The clear distinction in the latter can better exhibit the distribution trend in the equivalent values of ecosystem services in some study areas that contain multiple climatic zones, geographic topographies, and a variety of forest and vegetation types.
Figure 5. Vegetation status differences in cultivated land, forestland, grassland, wetland, and unused land.