Spatial modeling and prediction under stationary non-geometric range anisotropy

For modeling spatial processes, we propose a rich parametric class of stationary range anisotropic covariance structures that, when applied in R², greatly increases the scope of variogram contors. Geometric anisotropy, which provides the most common generalization of isotropy within stationarity, is a special case. Our class is built from monotonic isotropic correlation functions and special cases include the Matérn and the general exponential functions. As a result, our range anisotropic correlation specification can be attached to a second order stationary spatial process model, unlike ad hoc approaches to range anisotropy in the literature. We adopt a Bayesian perspective to obtain full inference and demonstrate how to fit the resulting model using sampling-based methods. In the presence of measurement error/microscale effect, we can obtain both the usual predictive as well as the noiseless predictive distribution. We analyze a data set of scallop catches under the general exponential range anisotropic model, withholding ten sites to compare the accuracy and precision of the standard and noiseless predictive distributions.     

A new synergetic paradigm in environmental numerical modeling: Hybrid models combining deterministic and machine learning components

A new type of environmental numerical models, hybrid environmental numerical models (HEMs) based on combining deterministic modeling and machine learning components, is introduced and formulated. Conceptual and practical possibilities of developing HEM, as an optimal synergetic combination of the traditional deterministic/first principles modeling (like that used for solving PDEs on the sphere representing model dynamics of global climate models) and machine learning components (like accurate and fast neural network emulations of model physics or chemistry processes), are discussed. Examples of developed HEMs (hybrid climate models and a hybrid wind–wave ocean model) illustrate the feasibility and efficiency of the new approach for modeling extremely complex multidimensional systems.     

A Bayesian strategy for combining predictions from empirical and process-based models

We present a strategy for using an empirical forest growth model to reduce uncertainty in predictions made with a physiological process-based forest ecosystem model. The uncertainty reduction is carried out via Bayesian melding, in which information from prior knowledge and a deterministic computer model is conditioned on a likelihood function. We used predictions from an empirical forest growth model G-HAT in place of field observations of aboveground net primary productivity (ANPP) in a deciduous temperate forest ecosystem. Using Bayesian melding, priors for the inputs of the process-based forest ecosystem PnET-II were propagated through the model, and likelihoods for the PnET-II output ANPP were calculated using the G-HAT predictions. Posterior distributions for ANPP and many PnET-II inputs obtained using the G-HAT predictions largely matched posteriors obtained using field data. Since empirical growth models are often more readily available than extensive field data sets, the method represents a potential gain in efficiency for reducing the uncertainty of process-based model predictions when reliable empirical models are available but high-quality data are not.     

Simple Spatial Modeling Tool for Prioritizing Prescribed Burning Activities at the Landscape Scale

Abstract:   Resources for prescribed fire are frequently insufficient to manage public lands for all conservation and resource management objectives, necessitating prioritization of the application of fire across the landscape within any given year. Defining tradeoffs when applying prescribed fire to large landscapes is problematic not only because of the complexity of weighing competing management objectives at the landscape scale, but also because of the difficult nature of independently applying need-to-burn criteria to large areas. We present a case study of a simple modeling process implemented at Eglin Air Force Base in the Florida Panhandle (U.S.A.) to prioritize the application of prescribed fire. In a workshop setting, managers and biologists identified key conservation criteria and landscape management objectives that drive the application of prescribed fire. Remote sensing and other spatial data were developed to directly or indirectly represent all these criteria. Using geographic information system software, managers and biologists weighted each criterion according to its relative contribution to overall burn prioritization, and individual values for the criterion were scored according to how they influence the need to burn. Subsequently, this process has been validated and modified through ecological monitoring. This modeling process has also been applied to the 77,400-ha Blackwater River State Forest, public land adjacent to Eglin Air Force Base, demonstrating its applicability to lands with varying management priorities. The advantages of this model-based approach for prioritizing prescribed fire include the reliance on accessible, inexpensive software, the development of spatially explicit management objectives, the ease of transferability, and clearly stated assumptions about management that may be tested and reviewed through monitoring and public comment.     

Assessing Land-Use Impacts on Natural Resources

/ Much information is available on changes that occur in natural resources from both spatially-explicit data on environmental conditions and models of the interactions of these conditions and resources with human activities. The strategy for assessing land-use impacts on natural resources developed in this paper provides a framework for using relevant data and models to address questions of how management practices can promote both use and protection of resources. This assessment strategy integrates spatially explicit environmental data using geographic information systems (GIS) with computer models that simulate changes in land cover in response to land-use impacts. The computer models also simulate susceptibility of species to changes in habitat suitability and landscape patterns. The approach is applied to management of limestone barrens on the Oak Ridge Reservation in East Tennessee. Potential limestone barrens habitats are identified by overlaying appropriate soils, geology, slope, and land-use/land-cover conditions. Their validity is tested against known sites containing rare species that occur in these habitats. The location of habitats at risk in the aftermath of human activities is determined by using an available area model that identifies the size and proximity of sites that particular types of species can no longer use as habitat. The resulting risk map can be used in land management planning. The approach uses readily available in situ and remotely sensed data and is applicable to a wide range of locations and land-use scenarios. This approach can be refined based on needs identified by land managers and on the sensitivity of the results to the resolution of available resource information.KEY WORDS: Land management; Assessment; Habitat characterization; Limestone barrens; Ecological modeling; Geographic information systems     

ECONOMIC AND WATER QUALITY EFFECTS OF VARIABLE AND UNIFORM APPLICATION OF NITROGEN1

ABSTRACT: Nitrogen (N) fertilizer rates for achieving optimum crop yields often vary within a field due to spatial variability in soil moisture and nitrogen content and other crop growth factors. When there is substantial within-field variability in these factors, uniform application of N (UAN) may not be economically efficient in terms of maximizing net return because N is likely to be over-applied in some areas and under-applied in other areas of the field. In addition, over-application can adversely affect water quality. A sample of fields in a Midwestern agricultural watershed is used to test for statistically significant differences in N application rates, crop yields, surface and ground water quality and net returns between UAN and variable application of N (VAN) for four cropping systems. Profitability and water quality benefits of VAN are sensitive to the distribution of soil types within a field. Water quality effects and profitability of UAN and VAN vary with cropping systems. VAN is not uniformly superior to UAN in terms of increasing net returns and improving water quality for the farming systems and watershed evaluated in this study.     

REGRESSION MODELS OF HERBICIDE CONCENTRATIONS IN OUTFLOW FROM RESERVOIRS IN THE MIDWESTERN USA, 1992–19931

ABSTRACT: Reservoirs are used to store water for public water supply, flood control, irrigation, recreation, hydropower, and wildlife habitat, but also often store undesirable substances such as herbicides. The outflow from 76 reservoirs in the midwestern USA, was sampled four times in 1992 and four times in 1993. At least one herbicide was detected in 82.6 percent of all samples, and atrazine was detected in 82.1 percent of all samples. Herbicide properties; topography, land use, herbicide use, and soil type in the contributing drainage area; residence time of water in reservoirs; and timing of inflow, release, and rainfall all can affect the concentration of herbicides in reservoirs. A GIS was used to quantify characteristics of land use, agricultural chemical use, climatic conditions, topographic character, and soil type by reservoir drainage basins. Multiple linear and logistic regression equations were used to model mean herbicide concentrations in reservoir outflow as a function of these characteristics. Results demonstrate a strong association between mean herbicide concentrations in reservoir outflow and herbicide use rates within associated drainage basins. Results also demonstrate the importance of including soils and basin hydrologic characteristics in models used to estimate mean herbicide concentrations.     

NONLINEAR MODELING AND PREDICTION OF A RIVER FLOW SYSTEM1

ABSTRACT: Model estimation and prediction of a river flow system are investigated using nonlinear system identification techniques. We demonstrate how the dynamics of the system, rainfall, and river flow can be modeled using NARMAX (Nonlinear Autoregressive Moving Average with eXogenuous input) models. The parameters of the model are estimated using an orthogonal least squares algorithm with intelligent structure detection. The identification of the nonlinear model is described to represent the relationship between local rainfall and river flow at Enoree station (inputs) and river flow at Whitmire (output) for a river flow system in South Carolina.     

VERIFICATION OF THE RHEA-OROGRAPHIC-PRECIPITATION MODEL1

ABSTRACT: Observed April 1 snowpack accumulations within and near the Gunnison River basin in southwestern Colorado are compared with simulations from the Rhea-orographic-precipitation model to determine if the model simulates reliable magnitudes and temporal and spatial variability in winter precipitation for the basin. Twenty simulations of the Rhea model were performed using‘optimal’parameter sets determined for 10-kilometer (km) grids (10-km by 10-km grid cells) through stochastic calibration. Comparisons of Rhea-model simulations of winter precipitation with April 1 snowpack accumulations at 32 snowcourse stations were performed for the years 1972–1990. For most stations and most years the Rhea model reliably simulates the temporal and spatial variability in April 1 snowpack accumulations. However, in general, the Rhea-model underestimates April 1 snowpack accumulations in the Gunnison River basin area, and the underestimation is greatest for locations that receive the largest amount of snow. A significant portion of the error in Rhea-model simulations is due to the calibration of the Rhea model using gauge-catch precipitation measurements which can be as much as 50 percent below actual snowfall accumulations. Additional error in the Rhea-model simulations is a result of the comparison of gridded precipitation values to observed values measured at points.