Performing Department
Soil, Water & Environmental Science
Non Technical Summary
Arid and semiarid ecosystems represent critical agricultural and economic components of southwestern U.S. ecosystems (Mitchell, 2000). Semiarid ecosystems provide critical ecosystem services such as watershed and groundwater recharge, carbon sequestration, and maintenance of open space and plant diversity. Semiarid ecosystem productivity is limited by lack of consistent precipitation (Noy Meir, 1973), and as such, there is a need to understand the potential response of these ecosystems to changes in climate or precipitation patterns. In particular, coupling the controlling effects of variability in soil-water dynamics with plant production/diversity will facilitate improved management strategies and land use practices in these climatically sensitive ecosystems. Variability in the quantity and temporal distribution of precipitation and soil moisture are critical to semiarid ecosystem productivity and health (Whitford 2002; Knapp et al., 2002; Huxman et al., 2004). Of particular importance is the interaction between soil properties (e.g., clay content, and subsurface diagnostic horizons) and their impacts on the variability and availability of soil-water (McAuliffe, 1994). Recent drought and the introduction of non-native plant species have significantly reduced the ability of many semiarid ecosystems to support the traditional ecosystem services, such as livestock production, because of reduction and greater interannual variability in aboveground primary productivity. The ability of these lands to return to a previous productive state is unclear, and as such, land managers may need to transition from traditional land use practices and manage for alternative ecosystem services such as open space, carbon sequestration and/or groundwater recharge. A key feature to understanding semiarid ecosystem function are the dynamic land-vegetation-atmosphere interactions that control water availability and partitioning of water to recharge, evapotranspiration and primary production. A fundamental knowledge gap to understanding these interactions is an accurate, high resolution representation of soil properties. Recently, the field of "digital soil mapping" has emerged to fill this knowledge gap. Digital soil mapping includes techniques such as remote sensing, digital terrain modeling, Geographic Information Systems (GIS), and statistical methods such as principal component analysis and empirical pedotransfer functions, and may be used to quantitatively predict soil type and soil physical property distribution at high resolution across large areas. This approach to predicting soil data has the potential to provide a key component necessary for understanding how climate variability impacts the semiarid ecosystems of southern Arizona. The proposed research will quantify how variability in soil properties and climate forcing interact to control aboveground primary productivity and ecosystem function. Specifically, the research will utilize a combination of high-technology techniques, such as digital soil mapping, to link spatial and temporal patterns of soil and climate with aboveground vegetation dynamics
Animal Health Component
0%
Research Effort Categories
Basic
75%
Applied
25%
Developmental
(N/A)
Goals / Objectives
The overall objective is to integrate digital soil mapping techniques with measures of soil physical properties, soil-water dynamics, and ecosystem production to characterize semi-arid ecosystem response to climate forcing over larges areas of southern Arizona. The overall hypothesis is that soil physical properties modulate ecosystem response to climate variability through control of soil-water dynamics and availability. Thus quantifying in detail the spatial patterns of soil physical properties such as clay or rock content provides a first order means to constrain potential ecosystem response to climate forcing. The proposed research involves two research components: Determine spatial patterns of soil physical properties important to quantifying soil-water availability using a combination of digital soil mapping, field based soil survey and characterization, and pedotransfer functions; Characterize the temporal and spatial variation in aboveground vegetation productivity as it relates to soil physical properties using a combination of remote sensing and modeling techniques.
Project Methods
Study Areas: The study will take a multi-scale approach to addressing the research objectives. This will include: § Regional scale approach quantifying the relation of soil physical properties, climate, and vegetation productivity across the southern Basin and Range physiographic province (Fig. 3) that encompasses over XX km2; § Soil Survey Area - Aravaipa and western Arizona § First order catchment scale - 5-10 ha Data Inputs: The data inputs, or environmental covariates representing the CORPAN parameters, will be derived from a combination of remotely sensed data and digital elevation models. For the soil survey and catchment scale work LandSat 7 ETM+ data (sourced prior to 2001 to avoid issues with image data striping) will be used to quantify surface reflectance and specific band ratios that relate to specific surface properties (Scull et al., 2003). Most Landsat bands record data at a 30-m resolution. The 30-m resolution data will be refined to 15-m using a "pan-sharpening" statistical function derived between individual bands and the 15-m resolution panchromatic band. Reflectance indices representing soil, geology, and vegetation will be calculated from the Landsat data and elevation data will be used to derive topographic indices (e.g., slope, solar radiation, wetness index). Topographic variables will be derived from the 10-m resolution digital elevation model (DEM) data available from the USGS. Derived topographic variables will include slope, aspect, wetness index, and a moving window of variation in surface elevation. Prior to topographic variable derivation, we will perform a low pass filter of the DEM data and rescale to the 15-m resolution LandSat data. The 15-m resolution data represents an appropriate resolution for mapping over an area this size at provides information at a scale comparable to 1:24,000 or larger. Predictive Data Layers: The landscape will initially be segregated based on landform, i.e., mountainous upland and Quaternary sediments (alluvial fans, fan terraces, and basin floor material) using a combination of available USGS geology maps and aerial imagery following Nauman (2008). Following this supervised classification, an iterative data reduction technique will be implemented to determine the environmental covariates that account for the greatest amount of variation across for each landscape class. These environmental covariates will comprise the predictive data layers used for the field sample design, soil spatial modeling, and derivation of soil map units. A data reduction technique that involves an iterative principal component analysis (iPCA) to reduce the number of input data layers will be used to determine those layers contributing most to the spatial variance in soil-landscape relationships (Nauman, 2008; Levi, in preparation). This systematic approach utilizes principal component outputs (eigenmatrix and eigenvalues) to calculate loading factors of each input band to quantitatively determine the importance of each data layer. Field Sampling and Analysis The environmental covariates determined by the iPCA routine will be used to develop a field sampling scheme that best captures the range and variance of the predictive data layers. We will utilize a conditioned Latin Hypercube Sampling (cLHS) technique that is designed to represent the maximum variability in the predictive data layers determined by iPCA (surface reflectance and topographic indices). The cLHS design is a stratified random technique that represents the multivariate distributions of the chosen environmental covariate data (Minasny and McBratney, 2006). The cLHS technique is essentially an optimization routine where n number of points (sample locations) are selected from N number of data (predictive data layers) so that n forms a Latin hypercube of N. This technique provides full coverage of the range of each environmental covariate and has been found to be an efficient way to sample the distribution of predictive data layers (Minasny and McBratney, 2006). Effective coverage of predictive data layer range and distribution is critical when using these layers in a soil spatial prediction framework. In the cLHS routine the location of sample points is determined using an iterative process to yield randomly stratified sample locations in the variable space defined by the predictive data layers. The number of points is determined by the user and involves running the cLHS routine for a range of n sample locations and determining the sample number the best represents the distribution of the predictive data layers. Choosing the number of sample locations may also be limited somewhat subjectively by the number of samples that can realistically be collected and analyzed for a survey area. In the interest of maintaining a manageable number of locations within the time and funding limitations, we will determine ten sample locations within each replicate mapunit using the cLHS routine. These sample locations may be complemented with any NRCS point/transect data collected in the selected mapunits during mapping activities. Sample Collection and Analysis Soil morphological description and field-site characterization will be performed at all the sample locations determined from the cLHS analysis following NCSS standards (Schoenberger et al., 2002). Field description will consist of excavating a soil pit on the order of 1 m2 and 1 m deep (or to the depth of refusal) and documenting field site characteristics, soil morphological and physical properties, and classification of soil genetic horizons and taxonomic class. Soil samples will be collected from each genetic horizon described in the field. Research Component 2: Spatiotemporal Patterns in Aboveground Productivity We will quantify vegetative cover as a proxy for aboveground productivity using a combination of remotely-sensed and ground-based data collection at both the plot and regional spatial scales. We will use the normalized difference vegetation index (NDVI) data derived from the MODIS platform (integrated over 7 or 16-day interval) to derive a time series of vegetative cover for each block for the Spring and Monsoon growing seasons. NDVI is the most widely used index for local and regional studies (e.g., Goward et al., 1985; Tucker and Sellers, 1986; Malingreau et al., 1989; Loveland et al., 1991; Townshend et al., 1994) and is a routine MODIS data product. At the regional scale the magnitude and variance of NDVI among soil groups determined from the digital soil mapping exercise and at the specific points selected from the NSSL database will be compared with soil physical and hydraulic properties to empirically identify differences among sites.