2.1. Study area
Haikou city is the capital of Hainan Province, located in the north of Hainan Province, China, with a latitude range of 19° 31' 32"N ~ 20° 04' 52" N and a longitude range of 110° 07' 22" E ~ 110° 42' 32" E (Fig. 1). Haikou City belongs to a tropical Marine monsoon climate, dry and less rain in winter and spring, many typhoons and rainfall in summer and autumn. The annual average sunshine hours are more than 2000 hours, the annual average wind speed is 3.4 m/s, the annual average precipitation is 1721.10 mm, and the annual average temperature is 24.13 ℃. Haikou City is rich in wetland resources, including rivers, lakes, reservoirs, tidal flats, mangroves and other types. However, with the acceleration of urbanization in recent years, wetlands have changed dramatically.
2.2. Data sources and processing
ENVI 5.3 image processing software and ArcGIS 10.5 software were used for data processing. Data on change of wetlands during 1959–2018 were acquired from historical remote sensing data. The land-cover data for 1959 are derived from the topographic map (scale = 1:25,000). The image of 1976 came from commercial purchase with a resolution of 5m. Images in 1985, 1995, 2005, 2018 were obtained from Google Earth Engine (GEE), with a resolution of 19m except for the 2018 resolution of 0.6m.
For research needs, we refer to the Ramsar Convention and China's Wetland Classification (GB/T 24708 − 2009), combined with the actual situation of Haikou wetland and the identification of wetland types on remote sensing images, and formulate the wetland classification system (Table 2). The wetland classification system is divided into two levels: the primary wetland includes coastal wetlands, inland wetlands, and constructed wetlands, and the secondary wetland is divided into 15 types. The non-wetland land is divided into construction land and other land.
The remote sensing images were pre-processed by radiometric correction, atmospheric correction, geometric correction, projection, and then trimmed with the scope map of the study area to obtain the 6-phase remote sensing images of the study area. Referring to the wetland classification system in Table 1, the wetland vector data of the six phases of Haikou City were obtained through human-computer interaction and visual interpretation, and then were compared, revised and improved according to the land-use status map of neighbor year. Then, 200 patches were randomly selected in each year, confirming the accuracy of interpretation. Sampling validation results show that the interpretation accuracy is higher than 90% per year.
2.3. Wetland temporal and spatial change
ArcGIS and Excel software were used to count the wetland area, and transitions between land types were analyzed using the transfer matrix. Spatial data overlay analysis was used to explore the spatial and temporal patterns of wetlands in Haikou City, and their area change was calculated. On this basis, the transition matrix (1959–2018) was generated from excel, and the Sankey plot of land type changes was drawn by Python.
2.4. Wetland ESV change
Based on the global ESV evaluation method proposed by Costanzaet (1997), Xie (2003, 2008, 2015) revised the evaluation method of equivalent factors per unit area value, which is simple and practical. Wetland ESV in Haikou was evaluated according to Xie's equivalent factor method and related literature (Daily 1997; Sun et al., 2018; Zhang et al., 2022). Ecosystem services are divided into four categories: supply services, regulatory services, support services, and cultural services, and they are further subdivided into food production, materials provision, gas regulation, climate regulation, hydrological regulation, waste degradation, soil conservation, biodiversity conservation and the aesthetic landscape (Table 1). Xie (2008) defined the ecosystem service value with an equivalent coefficient as 1/7 of the annual output market value of 1 hectare of farmland. Due to the long time span of the study, grain yields and comparable prices fluctuated greatly across years. To be comparable, 1,058.42 yuan / ha was calculated based on the 2018 grain output (2,465.52 kg / ha) and grain price (3.005 yuan / kg), as the benchmark equivalent to the value of ecosystem services. The ESV was calculated as follows: (1) and (2):
Where ESV is the total service value of the ecosystem in the study area, ESVi is the service value of i-type of ecosystem in the study area, Ai is the area of i-type of ecosystem (ha), En is equal to the ecological service value (1058.42 yuan / ha), and Vij is the j-type service value of the i-type ecosystem.
Table 1
Ecosystem service equivalent value per unit area in Haikou City
Land type
|
Supply services
|
Regulation services
|
Support services
|
Cultural services
|
Food production
|
Materials provision
|
Gas regulation
|
Climate regulation
|
Hydrological regulation
|
Waste degradation
|
Soil conservation
|
Biodiversity conservation
|
Aesthetic landscaping
|
shallow sea
|
21.09
|
17.2
|
15.7
|
31.75
|
27.14
|
34.46
|
0
|
36.7
|
6.1
|
saltwater lake/swamp
|
17.62
|
15.02
|
17.67
|
31.22
|
27.33
|
26.5
|
0
|
32.81
|
13.37
|
beach
|
20.14
|
13.33
|
20.28
|
28.56
|
27.53
|
25.89
|
15.81
|
36.77
|
9.34
|
mangroves
|
13.2
|
16.32
|
23.5
|
35.2
|
30.2
|
56.7
|
20.32
|
40.35
|
14.64
|
inland tidal flats/deltas
|
0.44
|
5.72
|
6.92
|
13.61
|
17.47
|
3.67
|
2.02
|
0.81
|
8
|
freshwater lake
|
0.6
|
0.25
|
8.42
|
17.1
|
56.35
|
6.75
|
0.1
|
2.31
|
4.52
|
river
|
0.65
|
0.26
|
9.45
|
18.3
|
60.45
|
8.95
|
0.33
|
2.41
|
4.31
|
freshwater swamp
|
0.5
|
0.2
|
8.21
|
6.2
|
31.2
|
12.3
|
0.5
|
1.4
|
2.1
|
freshwater farm
|
25.6
|
0
|
0.52
|
0.72
|
0.2
|
-12.5
|
0
|
0
|
0
|
mariculture
|
37.5
|
0
|
0.63
|
0.81
|
0.3
|
-16.3
|
0
|
0
|
0
|
reservoir
|
0.63
|
0.27
|
8.32
|
6.53
|
50.31
|
6.52
|
0
|
0.02
|
11.39
|
ditch
|
0
|
2.1
|
3.26
|
4.72
|
11.2
|
3.2
|
0
|
0
|
0
|
farmland
|
1
|
0.27
|
0.48
|
0.98
|
0.51
|
0.9
|
1.56
|
1.01
|
0.13
|
pond
|
0
|
1.2
|
0.62
|
0.81
|
1.52
|
1.62
|
0
|
0
|
0
|
other constructed wetlands
|
0
|
0
|
1.21
|
0.56
|
0.32
|
0.1
|
0
|
0
|
0
|
other land
|
0.62
|
2.1
|
2.3
|
3
|
2.68
|
1.65
|
4.86
|
3.73
|
1.58
|
construction land
|
0
|
0
|
0
|
0
|
-5.79
|
-1.88
|
0
|
0
|
0.19
|
2.5. ESV driving factors statistical analysis
Wetland ESV is constantly changing under the dual effects of nature and social economy. We collected annual meteorological data (http://data.cma.cn) and social and economic data (from Hainan province and Haikou statistical yearbooks). Precipitation (X1), wind speed (X2), temperature (X3), water pressure (X4), and relative humidity (X5) are used to reflect the climate of Haikou City. We select land use degree (X6), GDP (X7), population (X8) represent the social and economic characteristics. Aquaculture output (X9), aquaculture area (X10), farmland area (X11) and grain output (X12) are selected to reflect the development of agriculture.
The Pearson correlation analysis (2-tailed significance test) was used to test the relationship between total wetland ESV (ESVw), costal wetland ESV (ESVcw), terrestrial wetland ESV (ESViw), and artificial wetland ESV (ESVaw). Principal component analysis (PCA) selects fewer important variables through multiple linear transformation.
2.6. Wetland ESV spatial autocorrelation
Global Moran’s I was proposed by Moran in 1950. The four quadrants of the Moran scatter plots represent the four types of spatial relations between the spatial units and their neighborhoods. Among them, the high-high(HH) and low-low(LL) spatial units have a positively correlated spatial autocorrelation with the surrounding units, with less degree of spatial difference. However, the low-high(LH) and high-low(HL) are negative spatial autocorrelation relationships, indicating the large degree of spatial difference between the spatial units and the surrounding elements. Anselin proposed the local spatial autocorrelation problem in 1995 (Anselin, 1995). Local indicators of spatial association (LISA) are measured by the local Moran’s I statistic. Global Moran’s I and local Moran's I are calculated as shown in (3) and (4):
Where I is the Global Moran’s I; Ii is the local Moran's I index; n is the number of space units; xi and xj are the respective observations; is the average of the observations; and wij is the spatial weight matrix.
The study was divided into 500 m× 500 m-size grids, and the ESV change was used to explore the differences in the spatial change of the wetland. Meanwhile, the binary spatial autocorrelation analysis of land use degree (LUD), digital elevation model (DEM) and population was conducted using ESV, respectively. The distribution of all wetland or land use types is as follows, with shallow and saltwater lake / swamp distribution of 1. Beaches, mangroves, inland tidal flats / deltas, freshwater lakes, rivers, freshwater marshes, and other lands were assigned values of 2. The reservoir was designated as 3. Fresh water farms, mariculture, ditches, farmland, pond surface and other constructed wetlands assigned 4, and construction land value was 5. Other lands consists mainly of woodland and grassland. The mean DEM for each grid was calculated using the region statistics in ArcGIS.