Appendix: Analysis of driving mechanisms for LUCC
Selection of variables for LUCC
Elevation and slope, which were derived from a regional digital elevation model (DEM), were chosen to describe the general configuration of the surface. Climatic factors, especially precipitation and temperature, have a large impact on terrestrial vegetation and human land-use activities. The climatic records from weather stations around the Dongjiang River watershed from 1987 to 2010 were used to evaluate the influence of rainfall and temperature on LUCC. Aside from natural factors, human activities have also greatly influenced LUCC (Su et al. 2011; Weng 2002). Thus, population density and per capita GDP data were used to evaluate the influence of population mobility and socioeconomic growth on LUCC. In addition, proximity to national highways, provincial highways, cities, and rivers are four factors that contribute to the impact of human accessibility on LUCC (Patarasuk and Binford 2012). Hence, four variables (e.g., distance to national highway, distance to provincial highway, distance to city and distance to river) were chosenso that the impact of these factors on land use/land cover changes could be determined ( see in Table 1).
Acquisitions of variables for LUCC
The detailed extraction process of the selected driving forces was as follows. The grid maps of rainfall and temperature for the four periods (1994–1999, 1999–2004, 2004–2009, 1994–2009), derived from 11 meteorological stations (Figure 1), were generated using aKriging method. Elevation and slope were derived from a DEM. The vector map of national and provincial highways in 2005 was obtained from the traffic map of GuangdongProvince at a scale of 1:500,000. The vector map of the river and city were selected from the DEM and local administrative maps, respectively. The four distance variables (distance to national highways, distance to provincial highways, distance to rivers, and distance to cities) were calculated by a GIS spatial analysis method. The vector maps of the mean values of population and GDP for the four periods, derived from statistical yearbooks of 17 local counties, were attached to the corresponding map data of the 17 local counties using ArcGIS 9.3.
The RDA model establishment
In this case, RDA was used as an explanatory model to quantitatively determine the relationships between the dominant components of LUCC and biophysical and socio-economic driving forces in each of the grid cells, using climate, landform, population, GDP, distance to highways, distance to rivers, and distance to cities as explanatory variables. The response variable was swap changes of each LULC type as a result of map overlays in the four periods (1994–1999, 1999–2004, 2004–2009 and 1994–2009). According to Pontius et al (2004), the amount of swap changes for each LULC type was twice as much as the minimum of the gain and loss. Therefore, maps of the dominant LUCC for the four periods were selected and generated from the gross gain and loss of each kind of LULC type (Table 2).
Before applying RDA, the ten variables for each study period (1994–1999, 1999–2004, 2004–2009 and 1994–2009) were converted into raster data at a resolution of 30×30 m. The study region was divided into 1215 zones with an area of 5×5 km,which was subsequently used as the base statistical unit. The formula of thee response variables (ΔLij) is calculated using Eq. (1):
(1)
whereΔLj is the area of swap change of LUCC andjis between 1 and 7 (1: cropland, 2: garden land, 3: grassland, 4: woodland, 5: built-up land, 6: wetland and 7: bare land). Lkj is the swap change area of the typej and it was located in zonek (k = 1, 2, 3, ……, 1215). Tj is the total area of typej in the study region. Before statistical analysis, all of the dependent and independent variables were arcsine-square root and “min-max” normalized, respectively.
RDA resultswere presented graphically on thebiplots, on which swap changesfor each kind of LUCC and driving forceswere shown with arrows and their abbreviated names; the formerwere presented in blue and the latter were indicated in red. The length of the arrows indicated the significance of the correlationsbetween the distribution of the changed land use types andthe drivingforces. A longer arrow indicated astronger correlation between them(Sadyś et al. 2015).The vector of relationships (directly proportional or inversely proportional) was interpreted based on the position of the changed land use types relative to the end of the arrows. If the land use types were closeto the end of the arrow, the correlation was positive. If they were at the opposite side from the driving forces, the correlation was negative (Sadyś et al. 2015).
Table A1 Transition matrix of land use/land cover change in the Dongjiang River watershed from 1994 to 2009
Periods / Types / Crop / Garden / Grass / Wood / Built-up / Wet / Bare / Total1994–1999 / Crop / 1138.0 / 505.0 / 163.1 / 33.4 / 265.7 / 2.8 / 195.8 / 2303.8
Garden / 30.5 / 769.8 / 1177.7 / 2153.9 / 158.2 / 0.8 / 0.3 / 4291.1
Grass / 29.0 / 1341.3 / 1182.3 / 219.6 / 65.8 / 0.0 / 7.1 / 2845.1
Wood / 173.3 / 1200.5 / 615.4 / 11649.7 / 167.4 / 12.4 / 73.3 / 13891.9
Built-up / 215.6 / 19.9 / 95.1 / 34.5 / 1411.3 / 64.6 / 12.9 / 1853.9
Wet / 0.1 / 0.0 / 0.0 / 6.0 / 12.2 / 594.5 / 0.5 / 613.2
Bare / 571.0 / 41.1 / 221.8 / 2.1 / 32.3 / 0.0 / 391.1 / 1259.3
Total / 2157.3 / 3877.6 / 3455.4 / 14099.1 / 2112.9 / 675.0 / 681.0 / –
1999–2004 / Crop / 590.8 / 480.2 / 205.7 / 303.5 / 442.7 / 45.9 / 88.1 / 2157.3
Garden / 515.5 / 986.6 / 288.5 / 1729.3 / 252.9 / 9.7 / 94.9 / 3877.6
Grass / 471.2 / 891.2 / 277.0 / 1511.1 / 228.5 / 19.5 / 56.5 / 3455.4
Wood / 574.1 / 964.3 / 452.3 / 11506.5 / 441.0 / 35.6 / 122.7 / 14099.1
Built-up / 293.9 / 327.2 / 170.1 / 319.3 / 826.7 / 122.3 / 50.5 / 2112.9
Wet / 58.9 / 28.7 / 6.8 / 24.9 / 51.3 / 492.3 / 12.0 / 675.0
Bare / 204.5 / 160.0 / 69.3 / 78.7 / 100.3 / 21.0 / 47.2 / 681.0
Total / 2709.1 / 3838.9 / 1470.0 / 15477.7 / 2343.9 / 746.3 / 472.0 / –
2004–2009 / Crop / 695.6 / 105.6 / 631.5 / 767.4 / 347.4 / 30.8 / 130.4 / 2709.1
Garden / 729.9 / 285.3 / 1018.3 / 1403.9 / 321.8 / 17.3 / 61.9 / 3838.9
Grass / 291.8 / 88.7 / 357.9 / 544.8 / 144.9 / 4.2 / 37.6 / 1470.0
Wood / 740.2 / 237.1 / 1170.3 / 12597.7 / 611.3 / 14.5 / 103.4 / 15477.7
Built-up / 358.0 / 77.7 / 355.3 / 535.1 / 929.1 / 26.8 / 61.7 / 2343.9
Wet / 24.0 / 4.2 / 10.3 / 41.0 / 224.0 / 427.3 / 15.4 / 746.3
Bare / 138.7 / 8.7 / 67.3 / 128.0 / 63.8 / 5.9 / 59.5 / 472.0
Total / 2978.9 / 807.4 / 3611.3 / 16021.0 / 2642.9 / 526.8 / 469.9 / –
1994–2009 / Crop / 669.5 / 70.7 / 571.2 / 550.9 / 374.2 / 12.9 / 54.2 / 2303.8
Garden / 332.9 / 261.3 / 644.2 / 2720.3 / 260.7 / 12.7 / 58.7 / 4291.1
Grass / 439.6 / 139.3 / 746.3 / 1323.2 / 151.6 / 5.8 / 39.3 / 2845.1
Wood / 946.1 / 181.4 / 1050.0 / 10963.1 / 599.9 / 14.0 / 135.5 / 13891.9
Built-up / 259.3 / 51.2 / 205.5 / 303.9 / 903.4 / 54.0 / 74.9 / 1853.9
Wet / 26.2 / 2.6 / 13.7 / 19.5 / 119.4 / 415.5 / 16.4 / 613.2
Bare / 304.9 / 100.9 / 380.1 / 137.3 / 233.4 / 11.9 / 90.9 / 1259.3
Total / 2978.9 / 807.4 / 3611.3 / 16021.0 / 2642.9 / 526.8 / 469.9 / –
Note: units (km2)