Uncertainty analysis of predicted disturbance from off-road vehicular traffic in complex landscapes at fort hood

The US Army Engineering Research Development Center (ERDC) uses a modified form of the Revised Universal Soil Loss Equation (RUSLE) to estimate spatially explicit rates of soil erosion by water across military training facilities. One modification involves the RUSLE support practice factor (P factor), which is used to account for the effect of disturbance by human activities on erosion rates. Since disturbance from off-road military vehicular traffic moving through complex landscapes varies spatially, a spatially explicit nonlinear regression model (disturbance model) is used to predict the distribution of P factor values across a training facility. This research analyzes the uncertainty in this model's disturbance predictions for the Fort Hood training facility in order to determine both the spatial distribution of prediction uncertainty and the contribution of different error sources to that uncertainty. This analysis shows that a three-category vegetation map used by the disturbance model was the greatest source of prediction uncertainty, especially for the map categories shrub and tree. In areas mapped as grass, modeling error (uncertainty associated with the model parameter estimates) was the largest uncertainty source. These results indicate that the use of a high-quality vegetation map that is periodically updated to reflect current vegetation distributions, would produce the greatest reductions in disturbance prediction uncertainty.     

Spatial and Temporal Assessment of Cumulative Disturbance Impacts Due to Military Training,Burning, Haying,and Their Interactions on Land Condition of Fort Riley

The effects of military training activities on the land condition of Army installations vary spatially and temporally. Training activities observably degrade land condition while also increasing biodiversity and stabilizing ecosystems. Moreover, other anthropogenic activities regularly occur on military lands such as prescribed burns and agricultural haying—adding to the dynamics of land condition. Thus, spatially and temporally assessing the impacts of military training, prescribed burning, agricultural haying, and their interactions is critical to the management of military lands. In this study, the spatial distributions and patterns of military training-induced disturbance frequency were derived using plot observation and point observation-based method, at Fort Riley, Kansas from 1989 to 2001. Moreover, spatial and variance analysis of cumulative impacts due to military training, burning, haying, and their interactions on the land condition of Fort Riley were conducted. The results showed that: (1) low disturbance intensity dominated the majority of the study area with exception of concentrated training within centralized areas; (2) high and low values of disturbance frequency were spatially clustered and had spatial patterns that differed significantly from a random distribution; and (3) interactions between prescribed burning and agricultural haying were not significant in terms of either soil erosion or disturbance intensity although their means and variances differed significantly between the burned and non-burned areas and between the hayed and non-hayed areas.     

US army land condition-trend analysis (LCTA) program

The US Army Land Condition-Trend Analysis (LCTA) program is a standardized method of data collection, analysis, and reporting designed to meet multiple goals and objectives. The method utilizes vascular plant inventories, permanent field plot data, and wildlife inventories. Vascular plant inventories are used for environmental documentation, training of personnel, species identification during LCTA implementation, and as a survey for state and federal endangered or threatened species. The permanent field plot data documents the vegetational, edaphic, topographic, and disturbance characteristics of the installation. Inventory plots are allocated in a stratified random fashion across the installation utilizing a geographic information system that integrates satellite imagery and soil survey information. Ground cover, canopy cover, woody plant density, slope length, slope gradient, soil information, and disturbance data are collected at each plot. Plot data are used to: (1) describe plant communities, (2) characterize wildlife and threatened and endangered species habitat, (3) document amount and kind of military and nonmilitary disturbance, (4) determine the impact of military training on vegetation and soil resources, (5) estimate soil erosion potential, (6) classify land as to the kind and amount of use it can support, (7) determine allowable use estimates for tracked vehicle training, (8) document concealment resources, (9) identify lands that require restoration and evaluate the effectiveness of restorative techniques, and (10) evaluate potential acquisition property. Wildlife inventories survey small and midsize mammals, birds, bats, amphibians, and reptiles. Data from these surveys can be used for environmental documentation, to identify state and federal endangered and threatened species, and to evaluate the impact of military activities on wildlife populations. Short- and long-term monitoring of permanent field plots is used to evaluate and adjust land management decisions.     

Estimating the Variability and Confidence of Land Use and Imperviousness Relationships at a Regional Scale1

Abstract: Impervious cover is a commonly used metric to help explain or predict anthropogenic impacts on aquatic resources; often it is used as a surrogate for intensity of human impacts when evaluating effects on aquatic resources. The most common way to estimate imperviousness is based on relationships with land use. Few studies have evaluated how the relationship between impervious surface and land use varies among geographies with different levels of development and between types of imagery used to assign land use type. In this study, we assess variability in estimates of imperviousness based on two locally available land use datasets: one based on aerial imagery (2‐m resolution) and another based on satellite imagery (30‐m resolution). The ranges and variability in imperviousness within land use categories were assessed at several spatial scales, including within counties, between counties, and between watersheds. Results indicate that there was considerable variability for all developed land use types. Estimated impervious cover often varied over a range of 20‐40% points within a land use category. Furthermore, there were clear spatial patterns both between and within counties, with impervious cover for a given land use type being higher near the urban centers and lower at the margins of development. Estimates of imperviousness for 12 study watersheds indicated that variability increased with increasing watershed development, making it difficult to confidently set management or regulatory targets based on impervious cover. This study suggests that locally derived, high resolution satellite or aerial imagery should be used to estimate imperviousness when a high level of accuracy and precision is required for regulatory or management decisions. Furthermore, the error associated with impervious land use relationships should be accounted for when using impervious cover in runoff or water quality models, or when making management decisions regarding stream health.     

A Sensitivity and Error Analysis Framework for Lake Eutrophication Modeling1

ABSTRACT: A framework for sensitivity and error analysis in mathematical modeling is described and demonstrated. The Lake Eutrophication Analysis Procedure (LEAP) consists of a series of linked models which predict lake water quality conditions as a function of watershed land use, hydrolgic variables, and morphometric variables. Specification of input variables as distributions (means and standard errors) and use of first-order error analysis techniques permits estimation of output variable means, standard errors, and confidence ranges. Predicted distributions compare favorably with those estimated using Monte-Carlo simulation. The framework is demonstrated by applying it to data from Lake Morey, Vermont. While possible biases exist in the models calibrated for this application, prediction variances, attributed chiefly to model error, are comparable to the observed year-to-year variance in water quality, as measured by spring phosphorus concentration, hypolimnetic oxygen depletion rate, summer chlorophyll-a, and summer transparency in this lake. Use of the framework provides insight into important controlling factors and relationships and identifies the major sources of uncertainty in a given model application.     

Forecasts of habitat loss and fragmentation due to urban growth are sensitive to source of input data

The conversion of natural habitat to urban settlements is a primary driver of biodiversity loss, and species' persistence is threatened by the extent, location, and spatial pattern of development. Urban growth models are widely used to anticipate future development and to inform conservation management, but the source of spatial input to these models may contribute to uncertainty in their predictions. We compared two sources of historic urban maps, used as input for model calibration, to determine how differences in definition and scale of urban extent affect the resulting spatial predictions from a widely used urban growth model for San Diego County, CA under three conservation scenarios. The results showed that rate, extent, and spatial pattern of predicted urban development, and associated habitat loss, may vary substantially depending on the source of input data, regardless of how much land is excluded from development. Although the datasets we compared both represented urban land, different types of land use/land cover included in the definition of urban land and different minimum mapping units contributed to the discrepancies. Varying temporal resolution of the input datasets also contributed to differences in projected rates of development. Differential predicted impacts to vegetation types illustrate how the choice of spatial input data may lead to different conclusions relative to conservation. Although the study cannot reveal whether one dataset is better than another, modelers should carefully consider that geographical reality can be represented differently, and should carefully choose the definition and scale of their data to fit their research objectives.     

Effects of Land Use Changes on the Ecosystem Service Values of Coastal Wetlands

Changes in the coastal landscape of Southern Sinaloa (Mexico), between 2000 and 2010, were analyzed to relate spatial variations in wetlands extent with the provision and economic value of the ecosystem services (ES). Remote sensing techniques applied to Landsat TM imagery were used to evaluate land use/land cover changes while the value transfer method was used to assess the value of ES by land cover category. Five wetland types and other four land covers were found as representative of the coastal landscape. Findings reveal a 14 % decrease in the saltmarsh/forested mangrove area and a 12 % increase in the area of shrimp pond aquaculture (artificial wetland) during the study period. ES valuation shows that the total value flow increased by 9 % from $215 to $233 million (2007 USD) during the 10-year period. This increase is explained as result of the high value worldwide assigned to saltmarsh. We recognize limitations in the transfer-based approach in quantifying and mapping ES values in the region, but this method provides with value estimates spatially defined, and also provides some guidance in the preliminary screening of policies and projected development in the context of data-scarce regions.     

Application of Likelihood Ratio and Logistic Regression Models to Landslide Susceptibility Mapping Using GIS

For landslide susceptibility mapping, this study applied and verified a Bayesian probability model, a likelihood ratio and statistical model, and logistic regression to Janghung, Korea, using a Geographic Information System (GIS). Landslide locations were identified in the study area from interpretation of IRS satellite imagery and field surveys; and a spatial database was constructed from topographic maps, soil type, forest cover, geology and land cover. The factors that influence landslide occurrence, such as slope gradient, slope aspect, and curvature of topography, were calculated from the topographic database. Soil texture, material, drainage, and effective depth were extracted from the soil database, while forest type, diameter, and density were extracted from the forest database. Land cover was classified from Landsat TM satellite imagery using unsupervised classification. The likelihood ratio and logistic regression coefficient were overlaid to determine each factors rating for landslide susceptibility mapping. Then the landslide susceptibility map was verified and compared with known landslide locations. The logistic regression model had higher prediction accuracy than the likelihood ratio model. The method can be used to reduce hazards associated with landslides and to land cover planning. Note: This version was published online in June 2005 with the cover date of August 2004.     

A Comparison of Spatial and Spectral Image Resolution for Mapping Invasive Plants in Coastal California

We explored the potential of detecting three target invasive species: iceplant (Carpobrotus edulis), jubata grass (Cortaderia jubata), and blue gum (Eucalyptus globulus) at Vandenberg Air Force Base, California. We compared the accuracy of mapping six communities (intact coastal scrub, iceplant invaded coastal scrub, iceplant invaded chaparral, jubata grass invaded chaparral, blue gum invaded chaparral, and intact chaparral) using four images with different combinations of spatial and spectral resolution: hyperspectral AVIRIS imagery (174 wavebands, 4 m spatial resolution), spatially degraded AVIRIS (174 bands, 30 m), spectrally degraded AVIRIS (6 bands, 4 m), and both spatially and spectrally degraded AVIRIS (6 bands, 30 m, i.e., simulated Landsat ETM data). Overall success rates for classifying the six classes was 75% (kappa 0.7) using full resolution AVIRIS, 58% (kappa 0.5) for the spatially degraded AVIRIS, 42% (kappa 0.3) for the spectrally degraded AVIRIS, and 37% (kappa 0.3) for the spatially and spectrally degraded AVIRIS. A true Landsat ETM image was also classified to illustrate that the results from the simulated ETM data were representative, which provided an accuracy of 50% (kappa 0.4). Mapping accuracies using different resolution images are evaluated in the context of community heterogeneity (species richness, diversity, and percent species cover). Findings illustrate that higher mapping accuracies are achieved with images possessing high spectral resolution, thus capturing information across the visible and reflected infrared solar spectrum. Understanding the tradeoffs in spectral and spatial resolution can assist land managers in deciding the most appropriate imagery with respect to target invasives and community characteristics.     

Hydrologic Modeling Uncertainty Resulting From Land Cover Misclassification1

Abstract: A stochastic, spatially explicit method for assessing the impact of land cover classification error on distributed hydrologic modeling is presented. One‐hundred land cover realizations were created by systematically altering the North American Landscape Characterization land cover data according to the dataset’s misclassification matrix. The matrix indicates the probability of errors of omission in land cover classes and is used to assess the uncertainty in hydrologic runoff simulation resulting from parameter estimation based on land cover. These land cover realizations were used in the GIS‐based Automated Geospatial Watershed Assessment tool in conjunction with topography and soils data to generate input to the physically‐based Kinematic Runoff and Erosion model. Uncertainties in modeled runoff volumes resulting from these land cover realizations were evaluated in the Upper San Pedro River basin for 40 watersheds ranging in size from 10 to 100 km² under two rainfall events of differing magnitudes and intensities. Simulation results show that model sensitivity to classification error varies directly with respect to watershed scale, inversely to rainfall magnitude and are mitigated or magnified by landscape variability depending on landscape composition.     

Review of Urban Stormwater Quality Models: Deterministic,Stochastic, and Hybrid Approaches1

Abstract: The growing impact of urban stormwater on surface‐water quality has illuminated the need for more accurate modeling of stormwater pollution. Water quality based regulation and the movement towards integrated urban water management place a similar demand for improved stormwater quality model predictions. The physical, chemical, and biological processes that affect stormwater quality need to be better understood and simulated, while acknowledging the costs and benefits that such complex modeling entails. This paper reviews three approaches to stormwater quality modeling: deterministic, stochastic, and hybrid. Six deterministic, three stochastic, and three hybrid models are reviewed in detail. Hybrid approaches show strong potential for reducing stormwater quality model prediction error and uncertainty. Improved stormwater quality models will have wide ranging benefits for combined sewer overflow management, total maximum daily load development, best management practice design, land use change impact assessment, water quality trading, and integrated modeling.     

Estimation of biogenic emissions with satellite-derived land use and land cover data for air quality modeling of Houston-Galveston ozone nonattainment area

The Houston-Galveston Area (HGA) is one of the most severe ozone non-attainment regions in the US. To study the effectiveness of controlling anthropogenic emissions to mitigate regional ozone nonattainment problems, it is necessary to utilize adequate datasets describing the environmental conditions that influence the photochemical reactivity of the ambient atmosphere. Compared to the anthropogenic emissions from point and mobile sources, there are large uncertainties in the locations and amounts of biogenic emissions. For regional air quality modeling applications, biogenic emissions are not directly measured but are usually estimated with meteorological data such as photo-synthetically active solar radiation, surface temperature, land type, and vegetation database. In this paper, we characterize these meteorological input parameters and two different land use land cover datasets available for HGA: the conventional biogenic vegetation/land use data and satellite-derived high-resolution land cover data. We describe the procedures used for the estimation of biogenic emissions with the satellite derived land cover data and leaf mass density information. Air quality model simulations were performed using both the original and the new biogenic emissions estimates. The results showed that there were considerable uncertainties in biogenic emissions inputs. Subsequently, ozone predictions were affected up to 10 ppb, but the magnitudes and locations of peak ozone varied each day depending on the upwind or downwind positions of the biogenic emission sources relative to the anthropogenic NOx and VOC sources. Although the assessment had limitations such as heterogeneity in the spatial resolutions, the study highlighted the significance of biogenic emissions uncertainty on air quality predictions. However, the study did not allow extrapolation of the directional changes in air quality corresponding to the changes in LULC because the two datasets were based on vastly different LULC category definitions and uncertainties in the vegetation distributions.     

Integrating Satellite Imagery with Simulation Modeling to Improve Burn Severity Mapping

Both satellite imagery and spatial fire effects models are valuable tools for generating burn severity maps that are useful to fire scientists and resource managers. The purpose of this study was to test a new mapping approach that integrates imagery and modeling to create more accurate burn severity maps. We developed and assessed a statistical model that combines the Relative differenced Normalized Burn Ratio, a satellite image-based change detection procedure commonly used to map burn severity, with output from the Fire Hazard and Risk Model, a simulation model that estimates fire effects at a landscape scale. Using 285 Composite Burn Index (CBI) plots in Washington and Montana as ground reference, we found that an integrated model explained more variability in CBI (R ² = 0.47) and had lower mean squared error (MSE = 0.28) than image (R ² = 0.42 and MSE = 0.30) or simulation-based models (R ² = 0.07 and MSE = 0.49) alone. Overall map accuracy was also highest for maps created with the Integrated Model (63 %). We suspect that Simulation Model performance would greatly improve with higher quality and more accurate spatial input data. Results of this study indicate the potential benefit of combining satellite image-based methods with a fire effects simulation model to create improved burn severity maps.     

Sensitivity and uncertainty analysis of agro-ecological modeling for saffron plant cultivation using GIS spatial decision-making methods

The main objective of this research is to model the uncertainty associated with GIS-based multi-criteria decision analysis (MCDA) for crop suitability assessment. To achieve this goal, an integrated approach using GIS-MCDA in association with Monte Carlo simulation (MCS) and global sensitivity analysis (GSA) were applied for Saffron suitability mapping in East-Azerbaijan Province in Iran. The results of this study indicated that integration of MCDA with MCS and GSA could improve modeling precision by reducing data variance. Results indicated that applying the MCS method using the local training data leads to computing the spatial correlation between criteria weights and characteristics of the study area. Results of the GSA method also allow us to obtain the priority of criteria and identify the most important criteria and the variability of outputs under uncertainty conditions for model inputs. The findings showed that, commonly used primary zoning methods, without considering the interaction effects of variables, had significant errors and uncertainty in the output of MCDA-based suitability models, which should be minimized by the advanced complementarity of sensitivity and uncertainty analysis.     

Assessing state-wide biodiversity in the Florida Gap analysis project

The Florida Gap (Fl-Gap) project provides an assessment of the degree to which native animal species and natural communities are or are not represented in existing conservation lands. Those species and communities not adequately represented in areas being managed for native species constitute 'gaps' in the existing network of conservation lands. The United States Geological Survey Gap Analysis Program is a national effort and so, eventually, all 50 states will have completed it. The objective of Fl-Gap was to provide broad geographic information on the status of terrestrial vertebrates, butterflies, skippers and ants and their respective habitats to address the loss of biological diversity. To model the distributions and potential habitat of all terrestrial species of mammals, breeding birds, reptiles, amphibians, butterflies, skippers and ants in Florida, natural land cover was mapped to the level of dominant or co-dominant plant species. Land cover was classified from Landsat Thematic Mapper (TM) satellite imagery and auxiliary data such as the national wetlands inventory (NWI), soils maps, aerial imagery, existing land use/land cover maps, and on-the-ground surveys. Wildlife distribution models were produced by identifying suitable habitat for each species within that species' range. Mammalian models also assessed a minimum critical area required for sustainability of the species' population. Wildlife species richness was summarized against land stewardship ranked by an area's mandates for conservation protection.     

A multi-scale remote sensing approach for monitoring northern peatland hydrology: present possibilities and future challenges

Remote sensing has the potential to provide quantitative spatially explicit hydrological information across northern peatland complexes. This paper details a multi-scale remote sensing approach for assessing the use of Sphagnum mosses as proxy indicators of near-surface hydrology. Several spectral indices developed from the near infra-red (NIR) and shortwave infra-red (SWIR) liquid water absorption bands, as well as a biophysical index can be correlated with measures of near-surface moisture in the laboratory, in the field and from airborne imagery. Data from all platforms revealed similar patterns in the spectral indices in relation to changes in moisture although the strength of correlations was reduced as the spatial scale increased. The rapid collection of temporally and spatially explicit hydrological data means that the technique has potential practical application for environmental managers and peatland scientists at the local scale. The task of up-scaling the technique for use in operational peatland hydrological monitoring to the global scale is challenging but achievable, and requires further investigation into the heterogeneity of near-surface moisture across Sphagnum patches and the application of novel image processing techniques to improve the spatial resolution of currently available satellite imagery.     

Urbanization and the Loss of Resource Lands in the Chesapeake Bay Watershed

We made use of land cover maps, and land use change associated with urbanization, to provide estimates of the loss of natural resource lands (forest, agriculture, and wetland areas) across the 168,000 km² Chesapeake Bay watershed. We conducted extensive accuracy assessments of the satellite-derived maps, most of which were produced by us using widely available multitemporal Landsat imagery. The change in urbanization was derived from impervious surface area maps (the built environment) for 1990 and 2000, from which we estimated the loss of resource lands that occurred during this decade. Within the watershed, we observed a 61% increase in developed land (from 5,177 to 8,363 km²). Most of this new development (64%) occurred on agricultural and grasslands, whereas 33% occurred on forested land. Some smaller municipalities lost as much as 17% of their forest lands and 36% of their agricultural lands to development, although in the outlying counties losses ranged from 0% to 1.4% for forests and 0% to 2.6% for agriculture. Fast-growing urban areas surrounded by forested land experienced the most loss of forest to impervious surfaces. These estimates could be used for the monitoring of the impacts of development across the Chesapeake Bay watershed, and the approach has utility for other regions nationwide. In turn, the results and the approach can help jurisdictions set goals for resource land protection and acquisition that are consistent with regional restoration goals.     

Effect of the Spatial Variability of Land Use,Soil Type,and Precipitation on Streamflows in Small Watersheds1

Abstract: The spatial variability of the data used in models includes the spatial discretization of the system into subsystems, the data resolution, and the spatial distribution of hydrologic features and parameters. In this study, we investigate the effect of the spatial distribution of land use, soil type, and precipitation on the simulated flows at the outlet of “small watersheds” (i.e., watersheds with times of concentration shorter than the model computational time step). The Soil and Water Assessment Tool model was used to estimate runoff and hydrographs. Different representations of the spatial data resulted in comparable model performances and even the use of uniform land use and soil type maps, instead of spatially distributed, was not noticeable. It was found that, although spatially distributed data help understand the characteristics of the watershed and provide valuable information to distributed hydrologic models, when the watershed is small, realistic representations of the spatial data do not necessarily improve the model performance. The results obtained from this study provide insights on the relevance of taking into account the spatial distribution of land use, soil type, and precipitation when modeling small watersheds.     

Spatial Variation of Soil Salinity in the Mexicali Valley, Mexico: Application of a Practical Method for Agricultural Monitoring

The degradation of irrigated lands through the process of soil salinization, or the buildup of salts in the soil, has hampered recent increases in agricultural productivity and threatens the sustainability of large-scale cultivation in critical agricultural regions of the world. Rapid detection of soil salinity on a regional basis has been identified as key for effective mitigation of such land degradation. The ability to detect regional patterns of soil salinity at an accuracy sufficient for regional-scale resource management is demonstrated using Landsat 5 Thematic Mapper (TM) imagery. A case study of the Mexicali Valley of Baja California, Mexico was selected due to the region's agricultural significance and concern for future soil salinity increases. Surface soil salinity was mapped using georeferenced field measurements of electrical conductivity (EC), collected concurrently with Landsat 5 TM imagery. Correlations between EC measurements and common indices derived from the satellite imagery were used to produce a model of soil salinity through regression analysis. Landsat band 7, TNDVI, PCA 1, Tasseled Cap 3 and Tasseled Cap 5 were found to offer the most promising correlations with surface soil salinity. Generally low levels of soil salinity were detected, however, distinct areas of elevated surface salinity were detected at levels potentially impacting sensitive crops cultivated within the region. The difficulty detecting low levels of salinity and the mid-range spatial resolution of Landsat 5 TM imagery restrict the applicability of this methodology to the study of broad regional patterns of degradation most appropriate for use by regional resource managers.