2. El green experimental
3.4. Salidas por drenaje (G)
3.4.1. Entradas y drenaje
In the Xiangxi catchment legacy data is rarely available and access to terrain is limited or restricted due to missing roads and other logistic reasons. Available data vary in spatial coverage, spatial resolution, and temporal consistency (Tables 1 and 2). Thus, the most suitable database to assess relevant variables covering large areas is provided by space-borne RS data.
In particular, information from Digital Elevation Models (DEM) from the Shuttle Radar Topographic Mission (SRTM) are widely applied in soil erosion studies (e.g., HANCOCK ET AL., 2006; VRIELING, 2007; VRŠČAJ ET AL., 2007) since they provide presentations of landforms and allow for derivation of topographic information for grid-based analysis.
For poorly mapped regions, such as the Xiangxi catchment, SRTM-DEM data can offer large benefits for catchment studies. Contrary to SRTM-DEMs in a resolution of 1-arc-second (30 m cell size), those in a coarser resolution of 3-arc-second (90 m cell size) are freely available for most regions of the world (ZANDBERGEN, 2008). However, topographic features representing the geomorphology and hydrology relevant for soil erosion modeling (e.g., slope, aspect, flow accumulation flow path length, contributing area) cannot always be adequately captured at the coarse scale of 90 m and thus, produce artifacts (e.g., LYNN USERY ET AL., 2004). Since the SRTM-DEM in 3-arc-second were generated from the 1-arc-second data by sub-sampling using the center values of each 3×3 cells (KEERATIKASIKORN and TRISIRISATAYAWONG, 2008) for regions outside the United States, refinement of the DEM data to a finer resolution than 90 m using different interpolation techniques, such as bicubic polynomical interpolation, kriging, spline, IDW, and natural neighbor, proved as successful without a loss of significance and, thus, as beneficial in terms of adequate landscape representation (GROHMANN, 2006, VALERIANO ET AL., 2006; HER and HEATWOLE, 2008; KEERATIKASIKORN and TRISIRISATAYAWONG, 2008). Because of this, the spatial resolution of the available SRTM-DEM covering the Xiangxi catchment was improved using bilinear re-sampling as no impact of the re-sampling on the analysis but benefit from the higher spatial resolution was expected. The 'target' resolution was set to 45 m × 45 m as this cell size adequately captures the environmental indicators by avoiding artifacts resulting from a few pixels along a slope compared to the original resolution of 90 m. Moreover, this cell size allows for reasonable computation time.
Since terracing as a key soil conservation technology in the Xiangxi catchment (c.f., Section 3.2.5) is also linked to specific terrain conditions, digital terrain analysis can further provide indicators, which show effects on terrace conditions. Rather than providing indicators describing the local terrace design which is limited by the spatial resolution of the DEM, terrain analysis aim at providing information on the natural landscape surface characteristics without terraces. This allows determining effects relevant for differentiating terrace conditions, since the general flow paths and velocities of runoff still follows the natural terrain above the terrace plot scale. For instance, if the
contributing area increases, the water pressure by infiltration and interflow on the terrace plot is assumed to increase, too. Additionally, terrain can be used as a proxy for other environmental covariates such as climate or parent material (BEHRENS ET AL., 2010a, b). A typical example is the
elevation as a spatially explicit covariate on the rainfall (c.f., Section 3.2.2). Thus, the DEM (Table 1) served as most powerful data source and has been mostly used throughout the studies for digital terrain analyses such as the derivation of erosion-relevant properties of the terrain (e.g., erosive slope length and slope angle).
Table 1 Available data sources for soil erosion modeling in the Xiangxi catchment and their spatial resolution and applicability as well as references of use in the studies conducted within the framework
of the present thesis.
Type Data* Purpose/Applicability Resolution**/
Recording
Reference
Remote sensing data
DEM based on SRTM data version 4 (JARVIS ET AL., 2008)
Digital terrain analyses 90 m × 90 m / 45 m × 45 m
Manuscripts 1,2,3, and 4
Landsat-5-TM (Thematic Mapper), path 125, row 38, 2005 (2005-09-09), 2006 (2006-09-12), and 2007 (2007-09-15
Land cover analyses, derivation of
environmental covariates on the crop and
management factors
30 m × 30 m / 45 m × 45 m
Manuscripts 2 and 4
Land use classification from 1987 and 2007 (SEEBER ET AL., 2010)
Model parameterization, random agricultural plot selection for the terrace survey
30 m × 30 m / 45 m × 45 m
Manuscripts 1, 2, and 4
SPOT5 image from 2007 (2007-09-21) and Google Earth data (varying dates)
Reference base, basis for field mapping and for deriving data on the road network, settlements, and the shoreline of the Xiangxi River after the impoundment
5 m × 5 m Manuscript 4
Maps Soil types based on the Second National Soil Survey in China (SHI ET AL., 2010)
Soil classification, calculation of the soil erodibility 1:180,000 and 1:160,000, 126 soil profiles Manuscript 1, 2 Point data
Soil profiles and topsoil samples (0 - 20 cm) roperties)
Analysis of erosion- relevant soil physical and chemical properties, plausibility check 245 soil profiles and topsoil samples Manuscript 1, 2 Poyline data
Road network Plausibility check for the
road networks digitized from SPOT5 More than 386 kilo- meters of road Manuscript 4 Point data
Rainfall data from the National Meteorological Information Centre of China Determination of rainfall characteristics and calculation of rainfall erosivity 6 stations, daily recording Manuscript 2
* all data were projected to UTM WGS 1984, Zone 49N, ** original/re-sampled
Data on the land cover, and vegetation cover, respectively, is based on Landsat-5 (Thematic Mapper) images. In total, three Landsat-TM scenes showing the lowest cloud and haze cover
throughout the observations years were used for gridded analyses (Table 1). They served as data source for the derivation of environmental covariates on the crop and management cover on the catchment scale and as available input for the analyses of a possible influence of the land cover on the terrace degradation over time. Internal and external errors of these RS images, primarily due to atmospheric and topographic effects of rugged and steep sloping areas (JENSEN, 2000; RICHARDS and JIA, 2006), were reduced by means of image preprocessing (atmospheric, radiometric, and geometric corrections) using ATCOR for ERDAS IMAGINE® according to NEUBERT and MEINEL (2005) and RICHTER (2010).
Additionally, two land use classifications based on Landsat-TM images from 1987 and 2007 (Table 1), conducted for the Xiangxi catchment by SEEBER ET AL. (2010), were applied for parameterization of the soil erosion model and for random plot selection of agricultural land for the field mapping of terrace conditions. According to SEEBER ET AL. (2010), the land use classification from 1987 presents the time before the political enforcement of the TGD, its first feasibility study, and the environmental impact assessment (c.f., Annex V). In contrast to the pre-construction land use classification, the one from 2007 presents the post-construction phase referring to the time after the establishment of resettlement schemes as well as completion of dam construction (c.f., Annex V).
An ortho-rectified, panchromatic, and geo-referenced SPOT image (Table 1) was used as master scene for geo-referencing multi-spectral datasets and for data conducted within field surveys such as GPS-tracks of roads (Table 1). In combination with available, adequate Google Earth data covering the study area, both datasets further served as basis for field mapping and for deriving anthropogenic indicators by digitizing (Table 1). For instance, since neither precise information nor spatial data on the human activities in the TGA were available, both datasets were used to derive proxy indicators. According to INBAR and LLERENA (2000), the distance from a village is an important component in the erosion-relevant process of terrace degradation. Hence, spatial Euclidian distance transforms (e.g., of the digitized settlement areas) served as one indicator on the human influence. The same influence was assumed for the distance to roads and to the river network. Due to the river impoundment, the Xiangxi River and its major tributaries in the Backwater area became navigable. Consequently, comparable to the road network, Euclidian distance transforms to different orders of the river network system served as proxies on the accessibility of terraces and thus human activity.
Besides RS data, available legacy data on the soil erosion factors 'rainfall' and 'soil' were checked for their applicability and accordingly pre-processed.
For the derivation of the rainfall erosivity from precipitation data, climate data from the National Meteorological Information Centre of China (NMIC) were used. As daily precipitation data were only available for one station (Xingshan station) in the central Xiangxi catchment, precipitation data from five other available climate stations lying closest to the study area were included. According
to the NMIC, these stations are Badong, Shennongjia, Yichang, Yichangxian, and Zigui (Table 2). Despite for Yichangxian station, rainfall erosivity was calculated for each station as the requirement on long-term recording is fulfilled allowing for a comparison of long-term average observation and climate trends according to the recommendation by the World Meteorological Organization (WMO).
Table 2 Geographical position (UTM WGS 1984, Zone 49 N; X = Northing, Y = Easting), elevation, and length of daily rainfall records of the climate stations in the Xiangxi catchment and its
surrounding area. Station X Y Altitude (m a.s.l.) Start date of record (day- month-year) End date of record (day- month-year) Length of daily record (year/month/day) Badong 442491 3437891 295 01-07-1952 31-12-2007 45/06/00 Shennongjia 469423 3512773 950 01-01-1975 31-12-2007 32/11/29 Xingshan 477772 3455484 275 01-01-1958 31-12-2007 50/00/00 Yichang 528850 3396689 134 01-08-1951 31-12-2007 56/07/00 Yichangxian 530978 3396689 116 01-01-2003 31-12-2007 05/00/00 Zigui 469765 3429644 151 01-04-1959 31-12-2007 48/09/00
Soil data originates from soil maps and descriptions based on the SNSS (c.f., Section 3.2.4). In the Xiangxi catchment, the soils were surveyed at a scale of 1:160,000 to 1:180,000. The available data were pre-processed in terms of geo-referencing and digitizing. An additional in situ soil survey was conducted at a scale of 1:5,000 to 1:10,000. The focus of this soil survey was to determine soil types according to the WRB and to get erosion-relevant soil physical and chemical properties (i.e., grain sizes, bulk density, carbonate, soil organic matter). In total, 245 soil profiles, respectively, topsoil samples (0 - 20 cm) from forest and farmland plots were analyzed and used for a plausibility check of the soil map from the SNSS after personal communication with Prof. Dr. X. Shi from the Institute of Soil Science in Nanjing, Chinese Academy of Sciences. Based on this, the gridded soil map covering the whole Xiangxi catchment and linked soil profile descriptions from the SNSS were used as input in the soil erosion modeling.
Considering the DEM as most relevant base for digital terrain, all further available, conducted, and computed data were converted to exact the same spatial resolution of 45 m × 45 m and same grid origin. This provides exact support of field data and computed data for modeling to avoid negative, random influences from finer scales.