Error assessment in decision-tree models applied to vegetation analysis

Methods were developed to evaluate the performance of a decision-tree model used to predict landscape-level patterns of potential forest vegetation in central New York State. The model integrated environmental databases and knowledge on distribution of vegetation. Soil and terrain decision-tree variables were derived by processing state-wide soil geographic databases and digital terrain data. Variables used as model inputs were soil parent material, soil drainage, soil acidity, slope position, slope gradient, and slope azimuth. Landscapescale maps of potential vegetation were derived through sequential map overlay operations using a geographic information system (GIS). A verification sample of 276 field plots was analyzed to determine: (1) agreement between GIS-derived estimates of decision-tree variables and direct field measurements, (2) agreement between vegetation distributions predicted using GIS-derived estimates and using field observations, (3) effect of misclassification costs on prediction agreement, (4) influence of particular environmental variables on model predictions, and (5) misclassification rates of the decision-tree model. Results indicate that the prediction model was most sensitive to drainage and slope gradient, and that the imprecision of the input data led to a high frequency of incorrect predictions of vegetation. However, in many cases of misclassification the predicted vegetation was similar to that of the field plots so that the cost of errors was less than expected from the misclassification rate alone. Moreover, since common vegetation types were more accurately predicted than rare types, the model appears to be reasonably good at predicting vegetation for a randomly selected plot in the landscape. The error assessment methodology developed for this study provides a useful approach for determining the accuracy and sensitivity of landscape-scale environmental models, and indicates the need to develop appropriate field sampling procedures for verifying the predictions of such models.     

Assessing topographic patterns in moisture use and stress using a water balance approach

Through its control on soil moisture patterns, topography’s role in influencing forest composition is widely recognized. This study addresses shortcomings in traditional moisture indices by employing a water balance approach, incorporating topographic and edaphic variability to assess fine-scale moisture demand and moisture availability. Using GIS and readily available data, evapotranspiration and moisture stress are modeled at a fine spatial scale at two study areas in the US (Ohio and North Carolina). Model results are compared to field-based soil moisture measurements throughout the growing season. A strong topographic pattern of moisture utilization and demand is uncovered, with highest rates of evapotranspiration found on south-facing slopes, followed by ridges, valleys, and north-facing slopes. South-facing slopes and ridges also experience highest moisture deficit. Overall higher rates of evapotranspiration are observed at the Ohio site, though deficit is slightly lower. Based on a comparison between modeled and measured soil moisture, utilization and recharge trends were captured well in terms of both magnitude and timing. Topographically controlled drainage patterns appear to have little influence on soil moisture patterns during the growing season. In addition to its ability to accurately capture patterns of soil moisture in both high-relief and moderate-relief environments, a water balance approach offers numerous advantages over traditional moisture indices. It assesses moisture availability and utilization in absolute terms, using readily available data and widely used GIS software. Results are directly comparable across sites, and although output is created at a fine-scale, the method is applicable for larger geographic areas. Since it incorporates topography, available water capacity, and climatic variables, the model is able to directly assess the potential response of vegetation to climate change.     

Hierarchical relationships between landscape structure and temperature in a managed forest landscape

Management may influence abiotic environments differently across time and spatial scale, greatly influencing perceptions of fragmentation of the landscape. It is vital to consider a priori the spatial scales that are most relevant to an investigation, and to reflect on the influence that scale may have on conclusions. While the importance of scale in understanding ecological patterns and processes has been widely recognized, few researchers have investigated how the relationships between pattern and process change across spatial and temporal scales. We used wavelet analysis to examine the multiscale structure of surface and soil temperature, measured every 5 m across a 3820 m transect within a national forest in northern Wisconsin. Temperature functioned as an indicator – or end product – of processes associated with energy budget dynamics, such as radiative inputs, evapotranspiration and convective losses across the landscape. We hoped to determine whether functional relationships between landscape structure and temperature could be generalized, by examining patterns and relationships at multiple spatial scales and time periods during the day. The pattern of temperature varied between surface and soil temperature and among daily time periods. Wavelet variances indicated that no single scale dominated the pattern in temperature at any time, though values were highest at finest scales and at midday. Using general linear models, we explained 38% to 60% of the variation in temperature along the transect. Broad categorical variables describing the vegetation patch in which a point was located and the closest vegetation patch of a different type (landscape context) were important in models of both surface and soil temperature across time periods. Variables associated with slope and microtopography were more commonly incorporated into models explaining variation in soil temperature, whereas variables associated with vegetation or ground cover explained more variation in surface temperature. We examined correlations between wavelet transforms of temperature and vegetation (i.e., structural) pattern to determine whether these associations occurred at predictable scales or were consistent across time. Correlations between transforms characteristically had two peaks; one at finer scales of 100 to 150 m and one at broader scales of ^>300 m. These scales differed among times of day and between surface and soil temperatures. Our results indicate that temperature structure is distinct from vegetation structure and is spatially and temporally dynamic. There did not appear to be any single scale at which it was more relevant to study temperature or this pattern-process relationship, although the strongest relationships between vegetation structure and temperature occurred within a predictable range of scales. Forest managers and conservation biologists must recognize the dynamic relationship between temperature and structure across landscapes and incorporate the landscape elements created by temperature-structure interactions into management decisions.     

Environmental factors and ecological processes controlling vegetation patterns in boreal forests

An individual tree model of forest dynamics was used to examine the environmental and ecological factors controlling forest vegetation patterns in upland boreal forests of North America. Basic life history traits that characterized the regeneration, growth, and death of individual trees were combined with species-specific responses to important environmental factors. This model simulated forest structure and vegetation patterns in conifer, hardwood, and mixed conifer-hardwood forests and woodlands in several bioclimatic sub-regions of the North American boreal forest zone. Model testing identified the processes and parameters required to understand the ecology of upland boreal forests and weaknesses in our current understanding of these processes. These factors included climate, solar radiation, soil moisture, soil temperature and permafrost, the forest floor organic layer, nutrient availability, forest fires, and insect outbreaks. Model testing also identified which of these factors were important in each bioclimatic sub-region.     

Incorporating spatial non-stationarity of regression coefficients into predictive vegetation models

The results of predictive vegetation models are often presented spatially as GIS-derived surfaces of vegetation attributes across a landscape or region, but spatial information is rarely included in the model itself. Geographically weighted regression (GWR), which extends the traditional regression framework by allowing regression coefficients to vary for individual locations (‘spatial non-stationarity’), is one method of utilizing spatial information to improve the predictive power of such models. In this paper, we compare the ability of GWR, a local model, with that of ordinary least-squares (OLS) regression, a global model, to predict patterns of montane ponderosa pine (Pinus ponderosa) basal area in Saguaro National Park, AZ, USA on the basis of variables related to topography (elevation, slope steepness, aspect) and fire history (fire frequency, time since fire). The localized regression coefficients exhibited significant non-stationarity for four of the five environmental variables, and the GWR model consequently described the vegetation-environment data significantly better, even after accounting for differences in model complexity. GWR also reduced observed spatial autocorrelation of the model residuals. When applied to independent data locations not used in model development, basal areas predicted by GWR had a closer fit to observed values with lower residuals than those from the optimal OLS regression model. GWR also provided insights into fine-scale controls of ponderosa pine pattern that were missed by the global model. For example, the relationship between ponderosa pine basal area and aspect, which was obscured in the OLS regression model due to non-stationarity, was clearly demonstrated by the GWR model. We thus see GWR as a valuable complement to the many other global methods currently in use for predictive vegetation modeling.     

Landscape pattern in topographically complex landscapes: issues and techniques for analysis

Ecological research provides ample evidence that topography can exert a significant influence on the processes shaping broad-scale landscape vegetation patterns. Studies that ignore this influence run the risk of misinterpreting observations and making inappropriate recommendations to the management community. Unfortunately, the standard methods for landscape pattern analysis are not designed to include topography as a pattern-shaping factor. In this paper, we present a set of techniques designed to incorporate the topographic mosaic into analyses of landscape pattern and dynamics. This toolbox includes adjustments to classic landscape indices that account for non-uniform landscape topography, indices that capture associations and directionality in vegetation pattern due to topographic structure, and the application of statistical models to describe relationships between topographic characteristics and vegetation pattern. To illustrate these methods, we draw on examples from our own analysis of landscape pattern dynamics in logged and unlogged forest landscapes in southwestern British Columbia. These examples also serve to illustrate the importance of considering topography in both research and management applications.This revised version was published online in May 2005 with corrections to the Cover Date.     

Topography-mediated controls on local vegetation phenology estimated from MODIS vegetation index

Forest canopy phenology is an important constraint on annual water and carbon budgets, and responds to regional interannual climate variation. In steep terrain, there are complex spatial variations in phenology due to topographic influences on microclimate, community composition, and available soil moisture. In this study, we investigate spatial patterns of phenology in humid temperate forest as a function of topography. Moderate-resolution imaging spectro-radiometer (MODIS) vegetation indices are used to derive local patterns of topography-mediated vegetation phenology using a simple post-processing analysis and a non-linear model fitting. Elevation has the most explanatory power for all phenological variables with a strong linear relationship with mid-day of greenup period, following temperatures lapse rates. However, all other phenological variables show quadratic associations with elevation, reflecting an interaction between topoclimatic patterns of temperature and water availability. Radiation proxies also have significant explanatory power for all phenological variables. Though hillslope position cannot be adequately resolved at the MODIS spatial resolution (250 m) to discern impacts of local drainage conditions, extended periods of greenup/senescence are found to occur in wet years. These findings are strongly supported by previous field measurements at different topographic positions within the study area. The capability of detecting topography-mediated local phenology offers the potential to detect vegetation responses to climate change in mountainous terrain. In addition, the large, local variability of meteorological and edaphic conditions in steep terrain provides a unique opportunity to develop an understanding of canopy response to the interaction of climate and landscape conditions.     

Environmental,historical, and contextual determinants of vegetation cover: a landscape perspective

We formulated and tested models of relationships among determinants of vegetation cover in two agroforested landscapes of eastern North America (Haut Saint-Laurent, Quebec, Canada) that differed by the spatial arrangement of their geomorphic features and intensity of agricultural activities. Our landscape model compared the woody plots of each landscape in terms of the relative influence of environmental attributes, land use history (1958 – 1997), and spatial context (i.e., proximity of similar or contrasting land cover). Our vegetation model evaluated the relative contribution of the same sets of variables to the distributions of herbs, trees, and shrubs. Relationships were assessed using partial Mantel tests and path analyses. Significant environmental and contextual differences were found between the vegetation plots of the two landscapes, but disturbance history was similar. Our vegetation model confirms the dominant effect of historical factors on vegetation patterns. Whereas land-use history overrides environmental and contextual control for trees, herbaceous and shrub species are more sensitive to environmental conditions. Context is determinant only for understory species in older, less-disturbed plots. Results are discussed in relevance to vegetation dynamics in a landscape perspective that integrates interactions between environmental and human influences.     

Soil saturation effects on forest dynamics: scaling across a southern boreal/northern hardwood landscape

Patch modeling can be used to scale-up processes to portray landscape-level dynamics. Via direct extrapolation, a heterogeneous landscape is divided into its constituent patches; dynamics are simulated on each representative patch and are weighted and aggregated to formulate the higher level response. Further extrapolation may be attained by coarsening the resolution of or lumping environmental data (e.g., climatic, edaphic, hydrologic, topographic) used to delimit a patch.Forest patterns at the southern boreal/northern hardwood transition zone are often defined by soil heterogeneity, determined primarily by the extent and duration of soil saturation. To determine how landscape-level dynamics predicted from direct extrapolation compare when coarsening soil parameters, we simulated forest dynamics for soil series representing a range of drainage classes from east- central Maine. Responses were aggregated according to the distribution of soil associations comprising a 600 ha area based on local- (1:12,000), county- (1:120,000) and state- (1:250,000) scale soil maps. At the patch level, simulated aboveground biomass accumulated more slowly in poorer draining soils. Different soil series yielded different communities comprised of species with various tolerances for soil saturation. When aggregated, removal of waterlogging caused a 20–60% increase in biomass accumulation during the first 50 years of simulation. However, this early successional increase and the maximum level of biomass accumulation over a 200 year period varied by as much as 40% depending on the geospatial data. This marked discrepancy suggests caution when extrapolating with forest patch models by coarsening parameters and demonstrates how rules used to rescale environmental data need to be evaluated for consistency.     

Gradient modeling of conifer species using random forests

Landscape ecology often adopts a patch mosaic model of ecological patterns. However, many ecological attributes are inherently continuous and classification of species composition into vegetation communities and discrete patches provides an overly simplistic view of the landscape. If one adopts a niche-based, individualistic concept of biotic communities then it may often be more appropriate to represent vegetation patterns as continuous measures of site suitability or probability of occupancy, rather than the traditional abstraction into categorical community types represented in a mosaic of discrete patches. The goal of this paper is to demonstrate the high effectiveness of species-level, pixel scale prediction of species occupancy as a continuous landscape variable, as an alternative to traditional classified community type vegetation maps. We use a Random Forests ensemble learning approach to predict site-level probability of occurrence for four conifer species based on climatic, topographic and spectral predictor variables across a 3,883 km² landscape in northern Idaho, USA. Our method uses a new permutated sample-downscaling approach to equalize sample sizes in the presence and absence classes, a model selection method to optimize parsimony, and independent validation using prediction to 10% bootstrap data withhold. The models exhibited very high accuracy, with AUC and kappa values over 0.86 and 0.95, respectively, for all four species. The spatial predictions produced by the models will be of great use to managers and scientists, as they provide vastly more accurate spatial depiction of vegetation structure across this landscape than has previously been provided by traditional categorical classified community type maps.     

Rangeland biodiversity assessment using fine scale on-ground survey,time series of remotely sensed ground cover and climate data: an Australian savanna case study

Savanna rangelands are undergoing rapid environmental change and the need to monitor and manage landscape health is becoming increasingly an imperative of government agencies and research organizations. Remotely sensed ecological indicators of disturbance offer a potential approach, particularly in the context of issues of scale required to assess and monitor extensive rangeland areas. The objective of this research is to analyse the potential of spatially explicit ecological indicators of disturbance to explain the spatial variability in species diversity and abundance (including introduced flora species) in rangelands. For two mapped rangeland ecosystem types in northern Australia, regression analysis was used to explore the relationships between species diversity and abundance, and remotely sensed ground cover time series statistics, foliage projective cover, and a precipitation deficit index. It was assumed that the ecosystem types used had been mapped to represent uniform vegetation units and consequently predictors of environmental heterogeneity were not used in the regression analysis. It was found that the predictor variables performed well in explaining the variation in species diversity and abundance for the more open, homogenous and less topographically complex basalt ecosystem type and less effectively for the more structurally complex, more wooded and less disturbed metamorphic ecosystem type. The results indicate that, for mapped ecosystem types with low heterogeneity and topographic complexity, ground cover temporal mean and variance are potentially useful indicators of disturbance to species diversity and abundance, provided the local spatial variability in the climate signal is accounted for.     

Predicting Southern Appalachian overstory vegetation with digital terrain data

Vegetation in mountainous regions responds to small-scale variation in terrain, largely due to effects on both temperature and soil moisture. However there are few studies of quantitative, terrain-based methods for predicting vegetation composition. This study investigated relationships between forest composition, elevation, and a derived index of terrain shape, and evaluates methods for predicting forest composition. Trees were measured on 406 permanent plots within the boundaries of the Coweeta Hydrologic Lab, located in the Southern Appalachian Mountains of western North Carolina, USA. All plots were in control watersheds, without human or major natural disturbance since 1923. Plots were 0.08 ha and arrayed on transects, with approximately 380 meters between parallel transects. Breast-height diameters were measured on all trees. Elevation and terrain shape (cove, ridge, sideslope) were estimated for each plot. Density (trees/ha) and basal area were summarized by species and by forest type (cove, xeric oak-pine, northern hardwoods, and mixed deciduous). Plot data were combined with a digital elevation data (DEM), and a derived index of terrain shape at two sampling resolutions: 30 m (US Geological Survey), and 80 m (Defense Mapping Agency) sources. Vegetation maps were produced using each of four different methods: 1) linear regression with and without log transformations against elevation and terrain variables combined with cartographic overlay, 2) kriging, 3) co-kriging, and 4) a mosaic diagram. Predicted vegetation was compared to known vegetation at each of 77 independent, withheld data points, and an error matrix was determined for each mapping method.We observed strong relationships between some species and elevation and/or terrain shape. Cove and xeric oak/pine species basal areas were positively and negatively related to concave landscape locations, respectively, while species typically found in the mixed deciduous and northern hardwood types were not. Most northern hardwood species occurred more frequently and at higher basal areas as elevation increased, while most other species did not respond to elevation. The regression and mosaic diagram mapping approaches had significantly higher mapping accuracies than kriging and co-kriging. There were significant effects of DEM resolution on map accuracy, with maps based on 30 m DEM data significantly more accurate than those based on 80 m data. Taken together, these results indicate that both the mapping method and terrain data resolution significantly affect the resultant vegetation maps, even when using relatively high resolution data. Landscape or regional models based on 100 m or lower resolution terrain data may significantly under-represent terrain-related variation in vegetation composition.     

Prediction of total and rare plant species richness in agricultural landscapes from satellite images and topographic data

The diversity of future landscapes might depend on our ability to predict their potential species richness. The predictability of patterns of vascular plant species richness in a Finnish agricultural river landscape was studied using generalized linear modeling, floristic records from fifty-three0.25-km grid squares in the “core” study area, and environmental variables derived from Landsat TM images and a digital elevation model. We built multiple regression models for the total number of plant species and the number of rarities, and validated the accuracy of the derived models with a test set of 52 grid squares. We tentatively extrapolated the models from the core study area to the whole study area of 601 km² and produced species richness probability maps using GIS techniques. The results suggest that the local ‘hotspots’ of total flora (grid squares with > 200species) are mainly found in river valleys, where habitat diversity is high and a semi-open agricultural-forest mosaic occurs. The ‘hotspots’ of rare species (grid squares with > 4 rare species) are also found in river valleys, in sites where extensive semi-natural grasslands and herb-rich deciduous forests occur on steep slopes. We conclude that environmental variables derived from satellite images and topographic data can be used as approximate surrogates of plant species diversity in agricultural landscapes. Modeling of biological diversity based on satellite images and GIS can provide useful information needed in land use planning. However, due to the potential pitfalls in processing satellite imagery and model-building procedures, the results of predictive models should be carefully interpreted. This revised version was published online in July 2006 with corrections to the Cover Date.     

Small-scale effects of historical land use and topography on post-cultural tree species composition in an Alpine valley in southern Switzerland

Investigations of spatial patterns in forest tree species composition are essential in the understanding of landscape dynamics, especially in areas of land-use change. The specific environmental factors controlling the present patterns, however, vary with the scale of observation. In this study we estimated abundance of adult trees and tree regeneration in a Southern Alpine valley in Ticino, Switzerland. We hypothesized that, at the present scale, spatial pattern of post-cultural tree species does not primarily depend on topographic features but responds instead to small-scale variation in historical land use. We used multivariate regression trees to relate species abundances to environmental variables. Species matrices were comprised of single tree species abundance as well as species groups. Groups were formed according to common ecological species requirements with respect to shade tolerance, soil moisture and soil nutrients. Though species variance could only be partially explained, a clear ranking in the relative importance of environmental variables emerged. Tree basal area of formerly cultivated Castanea sativa (Mill.) was the most important factor accounting for up to 50% of species’ variation. Influence of topographic attributes was minor, restricted to profile curvature, and partly contradictory in response. Our results suggest the importance of biotic factors and soil properties for small-scale variation in tree species composition and need for further investigations in the study area on the ecological requirements of tree species in the early growing stage.     

Changes in large-scale patterns of plant biodiversity predicted from environmental economic scenarios

In the industrialized world large sums of money are spent on measures to preserve biodiversity by improving environmental quality. This creates a need to evaluate the effectiveness of such measures. In response we devel oped a model, NTM, that links plant biodiversity to abiotic variables that are under human control. These vari ables are: vegetation management, and the soil variables groundwater level, and nitrogen availability. We used species richness and the criteria of the Red Lists, i.e., the rarity and decline per species as measure for potential changes in biodiversity. NTM uses a statistical approach, and models potential plant biodiversity based on the above criteria as a non-linear function of the three soil variables. The regression model is calibrated on a data set consisting of 33,706 vegetation relevés. Because field data of vegetation combined with measurements of soil variables are insufficiently available, we used the average of Ellenbergs indicator values of the species in each relevé as a proxy. NTM was subjected to both validation and uncertainty analysis. The validation was car ried out by comparison with an independent data set. The uncertainty analysis showed that uncertainty in abso lute biodiversity values is large, but that comparative scenario studies can be carried out with an acceptable uncertainty. As an example we show the evaluation of the impact of three European economic scenarios on po tential plant biodiversity in the Netherlands. Although there were differences per vegetation type and per region, potential plant biodiversity had a tendency to increase, with the highest increase for the scenario with the highest reduction in atmospheric deposition of nitrogen and acidity.This revised version was published online in May 2005 with corrections to the Cover Date.     

An invasive species’ relationship with environmental variables changes across multiple spatial scales

To further our understanding of invasive species?? novel distributions, knowledge of invasive species?? relationships with environmental variables at multiple spatial scales is paramount. Here, we investigate which environmental variables and which spatial scales best explain the invasive mute swan??s (Cygnus olor) distribution in southern Ontario (Canada). Specifically we model mute swan distribution changes according to ecologically-relevant spatial scales: average territory size radius, 140?m; median dispersal distance of cygnets, 3,000?m; and average activity distance of males, 8,000?m. For individual spatial scales, global models using variables measured at each particular scale result in the highest Akaike weights, AUC, and Cohen??s Kappa values. Yet composite models (models combining variables measured at different scales) elicit the best models, as determined by higher Akaike weights and high AUC and Cohen??s Kappa values. Overall, percent water, waterbody perimeter density, temperature, precipitation, and road density are positively correlated with mute swan distribution, while percent forest and elevation are negatively correlated at all scales of analysis. Only percent water and annual precipitation are more influential in determining mute swan distribution at the 3,000 and 8,000?m zone scales than the territory scale. While most species distribution models are performed at a single scale, the results of our study suggest that composite models reflecting a species?? ecological needs provide models of better fit with similar, if not better, predictive accuracy. When analyzing species distributions, we also recommend that ecologists consider the scale of the underlying landscape processes and the effect that this may have on their modelling outcomes.     

Longitudinal- and transverse-scale environmental influences on riparian vegetation across multiple levels of ecological organization

Riparian vegetation is distinct from adjacent upland terrestrial vegetation and its distribution is affected by various environmental controls operating at the longitudinal scale (along the river) or transverse scale (perpendicular to the river). Although several studies have shown how the relative importance of transverse or longitudinal influences varies with the scale of observation, few have examined how the influences of the two scales vary with the level of ecological organization. We modeled vegetation-environment relationships at three hierarchically nested levels of ecological organization: species, plant community, and vegetation type. Our hierarchically structured analyses differentiated the spatial extent of riparian zones from adjacent upland vegetation, the distribution of plant community types within the riparian zone, and the distribution of plant species within community types. Longitudinal gradients associated with climate and elevation exerted stronger effects at the species level than at the community level. Transverse gradients related to lateral surface water flux and groundwater availability distinguished riparian and upland vegetation types, although longitudinal gradients of variation better predicted species composition within either riparian or upland communities. We concur with other studies of riparian landscape ecology that the relative predictive power of environmental controls for modeling patterns of biodiversity is confounded with the spatial extent of the study area and sampling scheme. A hierarchical approach to spatial modeling of vegetation-environment relationships will yield substantial insights on riparian landscape patterns.     

Modelling tree canopy cover and evaluating the driving factors based on remotely sensed data and machine learning

Quantifying urban tree cover is important to ensure sustainable urban ecosystem. This study calculates urban percent tree cover (PTC) for Bursa city, Turkey from Sentinel-2 data and evaluates the driving factors of PTC using an Artificial Neural Network-Multi Layer Perception (ANN-MLP) approach. For the PTC calculation, a Regression Tree (RT) analysis was performed using several vegetation indices (NDVI, LAI, fCOVER, MSAVI2, and MCARI) to improve accuracy. Socio-economic, topographic, and biophysical variables were incorporated into the ANN-MLP approach to evaluate the factors that drive urban PTC. A PTC prediction map was generated with an accuracy of 0.95 and a coefficient of determination of 0.87. The ANN-MLP training process yielded a correlation coefficient value of 0.71 and an R-square of 0.82 was achieved between the predicted ANN-MLP and observed tree cover maps. A priority tree cover map was generated considering statistical relationships between the factors and the ANN-MLP prediction map in addition to visual interpretations at the urban scale. Results demonstrate that, unlike other urban forms, PTC has a statistically negative relationship with the gross dwelling density (R2 =0.31). Topographic variables including slope and DEM were positively correlated with PTC with the R2 value of 0.80 and 0.72 respectively. The integration of remote sensing data with vegetation indices and driving factors yielded accurate prediction for identifying and evaluating the variability in the urban PTC.     

Road Density and Landscape Pattern in Relation to Housing Density,and Ownership,Land Cover,and Soils

Roads are conspicuous components of landscapes and play a substantial role in defining landscape pattern. Previous studies have demonstrated the link between roads and their effects on ecological processes and landscape patterns. Less understood is the placement of roads, and hence the patterns imposed by roads on the landscape in relation to factors describing land use, land cover, and environmental heterogeneity. Our hypothesis was that variation in road density and landscape patterns created by roads can be explained in relation to variables describing land use, land cover, and environmental factors. We examined both road density and landscape patterns created by roads in relation to suitability of soil substrate as road subgrade, land cover, lake area and perimeter, land ownership, and housing density across 19 predominantly forested counties in northern Wisconsin, USA. Generalized least squares regression models showed that housing density and soils with excellent suitability for road subgrade were positively related to road density while wetland area was negatively related. These relationships were consistent across models for different road types. Landscape indices showed greater fragmentation by roads in areas with higher housing density, and agriculture, grassland, and coniferous forest area, but less fragmentation with higher deciduous forest, mixed forest, wetland, and lake area. These relationships provide insight into the complex relationships among social, institutional, and environmental factors that influence where roads occur on the landscape. Our results are important for understanding the impacts of roads on ecosystems and planning for their protection in the face of continued development.