Locations of the Source Regions in China

Supplemental materials

Locations of the source regions in China

Fig.A.1. Locations of the source regions of the Yellow, Yangtze, and Lantsang rivers in China.

Determination of variations in trends of mobile sandy lands

Trends in the development of mobile sandy lands were determined using a decision tree of classifications, which wasconstructed via a series of binary decisions to place pixels into classes (see S2-Fig. A.1 below). The decision tree was used to perform multistage classifications based on a series of binary decisions to place pixels into classes, which is effective for clear spectrum mechanism such as water bodies and land. The maximum likelihood classification assumes that the statistics for each class in each band are normally distributed and calculates the probability that a given pixel belongs to a specific class, which is effective for irregular spectrum mechanism. We combined these two methods to extract information regarding the mobile sandy lands.

S2-Fig.A.1. Decision tree for mobile sandy land classification.

Note 1: Water Body and Land classifications

Water bodies show seasonal variations inspectral absorption rate, with a relativelyhigh rate in the dry season anda low rate in the flood season. In addition, suspended material,chlorophyll, and other factorsmay disturb the normal spectrum. Water bodies are often confused with other classes such as wetlands, wet fields, and shaded hillsides. Consequently, we used lower thresholds of NDVI and the infrared band of the images to identifywater bodies during drier periods.

Note 2: Vegetation and No Vegetation classifications

The spectral reflectance of vegetation has two peaks in the green and infrared bands, and plants show obvious physiological variations including sprouting, growing, maturity, fallen leaves, and dying/withered. Classifications for vegetated areasincluded forest, shrub, grassland, and artificial vegetation (i.e., farmlands), while classifications for land with no vegetation cover included buildings, bare land (i.e., unvegetated mountains), and glacier/permanent snow. Vegetation states were classified using the higher threshold of accumulated NDVI, maximum NDVI, and the infrared bandduring the growing season.

Note 3: Interpretationof mobile sandy lands

After land areas with no vegetation cover were identified, the supervised classification(maximum likelihood) method was employed to interpret the buildings, glacier/permanent snow, and mobile sandy land. For each class, training pixels (S2-Fig. A.2) were selected in the images, and the maximum likelihood classification was carried out in ENVI4.7. Among the training pixels, the B and C types are bare rock and bare land, respectively. However,during the field investigations we found that mobile sand dunes had usually developed on their surfaces, and therefore, we classified them as mobile sandy lands.

S2-Fig.A.2. Training pixelsfor identifying mobile sandy lands.

Trend change detection in NDVI time series

Following previous studies (e.g., Holben 1986, Stowe et al. 1991), during analysis of the temporal trends in the NDVI the maximum value composites (MVC), processed by month in the growing season(May to September), wereapplied. Toanalyze the changes in NDVI, a linear regression model was used to determine the change trend of every pixel. For each pixel, the linear relationship between NDVI and YEAR was calculatedusing the ordinary least squares (OLS) method:

NDVI = SLOPE × YEAR + b

Here nis the number of the monitored year, and iis a certain year between 1 and n.

The value of SLOPE for the NDVI indicatesthe trend in NDVI (e.g.,Zhenet al.2003), and large values of SLOPE show that there were rapid changes in vegetation productivity.For instance, positive values of SLOPE indicate an increasing trend in NDVI and that vegetation rehabilitation had occurred in that particular region (e.g.,Herrmann et al. 2005, Forkel et al. 2013), and that vegetation productivityhad increased (e.g., Xiao and Moody 2005, Beck et al. 2006).

Determination of actual NPP(net primary productivity) and potential NPPActual NPP

Actual NPP was calculated using the Carnegie–Ames–Stanford Approach (CASA) model, determined from the product of absorbed photosynthetically active radiation (APAR) and light use efficiency (ε):

where xis the spatial location and tisthe timescale.

(1) PAR (photosynthetically active radiation)

Here, PAR is the solar energy from the visible bands (400–700 nm), which is about 45% to 50% of (total surfacesolar radiation).For details see (Zotarelli et al.,2010).

(2) FPAR (fraction of absorbed photosynthetically active radiation)

Where SRmin represents SR for unvegetated land areas and is1.08 for all grid cells(Potter et al., 1993). Independent of vegetation, SRmaxvary between 4.14 and6.17.Herefor evergreen broadleaf forestsis 4.14, for deciduous broadleafforests and broadleaf and needleleaf mixed forestsare 6.17,for evergreenneedleleaf forests and deciduous needleleaf forestsare 5.43, and for broadleaf shrubs, temperate grasslands, savannas, alpinemeadows and tundra, deserts, and cultivationare 5.13 (Piao et al.,2005 ).

(3) ε (Light use efficiency)

Here, is the maximum light use efficiency (e.g., Wang et al. 2010)and it is 0.405 (Potter et al., 1993). Thevalues of may alter NPP values and magnitude trends, butthe relative trendsis not be changed.,, and arestress scalars that reduce . and represent monthly deviations from site-specific optimal temperature and from 20°C, respectively.is the monthly relative soil moisture deficit and is based on the difference between actual and potential evapotranspiration determined from the soil water balance.

Where (optimum temperature) is defined asthemonthlytemperature when the NDVI reaches its maximum of the year, and is monthlymean temperature(˚C).

Where EET is the estimated evapotranspiration (mm)generated by Soil Moisture SubModel, for details see (Potter et al,.1993 & Saxton et al.,1986). PET is potential evapotranspiration (mm) calculated by Thornthwaiteindex with acorrection factor of day-length (CF), for details see(Thornthwaite1948 & Fang and Yoda, 1990).

Potential NPP

Potential NPP was calculated using the model in the same manner as CASA, except for the calculations of FPAR, which was generated from vegetation and meteorological parameters as follows:

where k = 0.5,and the leaf area index (LAI) was obtained by MOD15A2 (Xu et al.,2009).

Significance of temporal trends in actual and potential NPPs

Slope of the potential NPP

Potential NPP refers to the NPP under conditions with no impacts fromhuman activities and with the outputs of NPP controlled by climate change (e.g., Princeet al. 2009). The slope of the potential NPP is expressed as follows:

pNPP = SLOPE × YEAR + b

where nis the number of the monitored year, andiis a certain year between 1 and n. When the value of the slope is positive, thisindicates that climate change has led to an increase invegetationproductivity (e.g., Beck et al. 2006), and vice versa.

Residual

The residual was defined as the difference between the potential NPP and the actual NPP (e.g., Xu et al. 2009), and was used to determinethe impacts of human activity and climate change on the regional ecology and environments (e.g., Evans and Geerken 2004, Geerken and Ilaiwi 2004). Positive values for the residuals indicate that human activity had anegative effect on vegetation rehabilitation (e.g., Wessels et al. 2007), and negative residuals indicate that human activities (e.g., programs to return farmland to forests, and to return forests to grasslands in some arid, semiarid, and semihumid regions)had a positive effect on vegetation rehabilitation.

In addition, as with the NDVI slope, the residual slope of NPP also indicated the environmental and ecological impacts of human activities. For instance, positive values of the slope show that over a relatively long period human activities had a negative effect on the environment, and vice versa (e.g., Wessels et al. 2007).

References

Beck PSA, Atzberger C, Høgda KA, Johansen B, Skidmore AK (2006)Improved monitoring of vegetation dynamics at very high latitudes: A new method using MODIS NDVI. Remote Sens Environ 100: 321–334

Evans J, Geerken R (2004)Discrimination between climate and human-induced dryland degradation. J Arid Environ 57: 535–554

Fang J, Kyoji Y (1990) Climate and vegetation in China III water balance and distribution of vegetation.Ecological Research5: 9-23

Forkel M, Carvalhais N, Verbesselt J, Mahecha MD, Neigh CSR, Reichstein M (2013)Trend change detection in NDVI time series: effects of inter-annual variability and methodology. Remote Sens 5: 2113–2144

Geerken R, Ilaiwi M (2004)Assessment of rangeland degradation and development of a strategy for rehabilitation. Remote Sens Environ 90: 490–504

Herrmann SM, Anyamba A, Tucker CJ (2005)Recent trends in vegetation dynamics in the African Sahel and their relationship to climate.Global EnvironChang 15: 394–404

Holben BN (1986)Characteristics of maximum-value composite images from temporal AVHRR data. IntJ Remote Sens 7: 1417–1434

Piao S, Fang J, ZhouL, ZhuB, TanK, TaoS (2005) Changes in vegetation net primary productivity from 1982 to1999 in China, Global Biogeochem. Cycles, 19, GB2027, doi:10.1029/2004GB002274

Potter CS, Randerson JT, Field CB, Matson PA, Vitousek PM, Mooney HA, Klooster SA (1993) Terrestrial ecosystem production: A process model based on global satellite and surface data.Global Biogeochem. Cycles,7(4): 811-841

Prince SD, Becker-Reshef I, Rishmawi K (2009)Detection and mapping of long-term land degradation using local net production scaling: Application to Zimbabwe. Remote Sens Environ 113: 1046–1057

SaxtonKE, Rawls W, Romberger JS, Papendick R I (1986) Estimating generalized soil-water characteristics from texture.Soil Sci. Soc. AM. J.50(4): 1031-1036

Stowe LL, McClain EP, Carey R, Pellegrino P, Gutman GG, Davis P, Long C, Hart S (1991)Global distribution of cloud cover derived from NOAA/AVHRR operational satellite data. Adv Space Res 11: 51–54

Thornthwaite CW (1948)An approach toward a rational classification of climate. Geogr Rev 38: 55–94

Wang H, Jia G, Fu C, Feng J, Zhao T, Ma Z (2010)Deriving maximal light use efficiency from coordinated flux measurements and satellite data for regional gross primary production modeling. Remote Sens Environ 114: 2248–2258

Wessels KJ, Prince SD, Malherbe J, Small J, Frost PE, van Zyl D (2007)Can human-induced land degradation be distinguished from the effects of rainfall variability? A case study in South Africa. J Arid Environ 68: 271–297

Xiao J, Moody A (2005)Geographical distribution of global greening trends and their climatic correlates: 1982–1998. IntJ Remote Sens 26: 2371–2390

Xu DY, Kang XW, Liu ZL, Zhuang DF, Pan JJ (2009) Assessing the relative role of climate change and human activities in sandy desertification of Ordos region, China. Sci China(Ser D) 52: 855–868

Zotarelli L, Dukes MD, Romero CC, Migliaccio KW, Morgan KT (2010) Step by step calculation of the Penman-Monteith Evapotranspiration (FAO-56 Method).Institute of Food and Agricultural Sciences. University of Florida.

Zhen L, Yan F, Fan X (2003)The variability of NDVI over northwest China and its relation to temperature and precipitation. Geosci Remote SensSymp, IEEE Int 4: 2275–2277