Spatial variation and prediction of forest biomass in a heterogeneous landscape

Large areas assessments of forest biomass distribution are a challenge in heterogeneous landscapes, where variations in tree growth and species composition occur over short distances. In this study, we use statistical and geospatial modeling on densely sampled forest biomass data to analyze the relative importance of ecological and physiographic variables as determinants of spatial variation of forest biomass in the environmentally heterogeneous region of the Big Sur, California. We estimated biomass in 280 forest plots (one plot per 2.85 km2) and measured an array of ecological (vegetation community type, distance to edge, amount of surrounding non-forest vegetation, soil properties, fire history) and physiographic drivers (elevation, potential soil moisture and solar radiation, proximity to the coast) of tree growth at each plot location. Our geostatistical analyses revealed that biomass distribution is spatially structured and autocorrelated up to 3.1 km. Regression tree (RT) models showed that both physiographic and ecological factors influenced biomass distribution. Across randomly selected sample densities (sample size 112 to 280), ecological effects of vegetation community type and distance to forest edge, and physiographic effects of elevation, potential soil moisture and solar radiation were the most consistent predictors of biomass. Topographic moisture index and potential solar radiation had a positive effect on biomass, indicating the importance of topographically-mediated energy and moisture on plant growth and biomass accumulation. RT model explained 35% of the variation in biomass and spatially autocorrelated variation were retained in regession residuals. Regression kriging model, developed from RT combined with kriging of regression residuals, was used to map biomass across the Big Sur. This study demonstrates how statistical and geospatial modeling can be used to discriminate the relative importance of physiographic and ecologic effects on forest biomass and develop spatial models to predict and map biomass distribution across a heterogeneous landscape.     

Mapping the risk of establishment and spread of sudden oak death in California

Sudden oak death, caused by the recently described pathogen Phytophthora ramorum, is an emerging forest disease that has reached epidemic levels in coastal forests of central California. We present a rule-based model of P. ramorum establishment and spread risk in California plant communities. The model, which is being used as a management tool to target threatened forests for early-detection monitoring and protection, incorporates the effects of spatial and temporal variability of multiple variables on pathogen persistence. Model predictions are based on current knowledge of host susceptibility, pathogen reproduction, and pathogen transmission with particular regard to host species distribution and climate suitability. Maps of host species distributions and monthly weather conditions were spatially analyzed in a GIS and parameterized to encode the magnitude and direction of each variable's effect on disease establishment and spread. Spread risk predictions were computed for each month of the pathogen's general reproductive season and averaged to generate a cumulative risk map (Fig. 6a and b). The model identifies an alarming number of uninfected forest ecosystems in California at considerable risk of infection by Phytophthora ramorum. This includes, in particular, a broad band of high risk north of Sonoma County to the Oregon border, a narrow band of high risk south of central Monterey County south to central San Luis Obispo County, and scattered areas of moderate and high risk in the Sierra Nevada foothills in Butte and Yuba counties. Model performance was evaluated by comparing spread risk predictions to field observations of disease presence and absence. Model predictions of spread risk were consistent with disease severity observed in the field, with modeled risk significantly higher at currently infested locations than at uninfested locations (P < 0.01, n = 323). Based on what is known about the ecology and epidemiology of sudden oak death, this model provides a simple and effective management tool for identifying emergent infections before they become established.     

Predicting potential and actual distribution of sudden oak death in Oregon: Prioritizing landscape contexts for early detection and eradication of disease outbreaks

An isolated outbreak of the emerging forest disease sudden oak death was discovered in Oregon forests in 2001. Despite considerable control efforts, disease continues to spread from the introduction site due to slow and incomplete detection and eradication. Annual field surveys and laboratory tests between 2001 and 2009 confirmed a total of 802 infested locations. Here, we apply two invasive species distribution models (iSDMs) of sudden oak death establishment and spread risk to target early detection and control further disease spread in Oregon forests. The goal was to develop (1) a model of potential distribution that estimates the level and spatial variability of disease establishment and spread risk for western Oregon, and (2) a model of actual distribution that quantifies the relative likelihood of current invasion in the quarantine area. Our predictions were based on four groups of primary parameters that vary in space and time: climate conditions, topographical factors, abundance and susceptibility of host vegetation, and dispersal pressure. First, we used multi-criteria evaluation to identify large-scale areas at potential risk of infection. We mapped and ranked host abundance and susceptibility using geospatial vegetation data developed with gradient nearest neighbor imputation. The host vegetation and climate variables were parameterized in accordance to their epidemiological importance and the final appraisal scores were summarized by month to represent a cumulative spread risk index, standardized as five categories from very low to very high risk. Second, using the field data for calibration we applied the machine-learning method, maximum entropy, to predict the actual distribution of the sudden oak death epidemic. The dispersal pressure incorporated in the statistical model estimates the force of invasion at all susceptible locations, allowing us to quantify the relative likelihood of current disease incidence rather than its potential distribution. Our predictions show that 65 km² of forested land was invaded by 2009, but further disease spread threatens more than 2100 km² of forests across the western region of Oregon (very high and high risk). Areas at greatest risk of disease spread are concentrated in the southwest region of Oregon where the highest densities of susceptible host species exist. This research identifies high priority locations for early detection and invasion control and illustrates how iSDMs can be used to analyze the actual versus potential distribution of emerging infectious disease in a complex, heterogeneous ecosystem.     

Pre-impact forest composition and ongoing tree mortality associated with sudden oak death in the Big Sur region; California

Mixed-evergreen forests of central coastal California are being severely impacted by the recently introduced plant pathogen, Phytophthora ramorum. We collected forest plot data using a multi-scale sampling design to characterize pre-infestation forest composition and ongoing tree mortality along environmental and time-since-fire gradients. Vegetation pattern was described using trend surface analysis, spatial autocorrelation analysis and redundancy analysis. Species-environment associations were modeled using non-parametric multiplicative regression (NPMR). Tanoak (Lithocarpusdensiflorus) mortality was analyzed with respect to environmental and biotic factors using trend surface analysis and multivariate regression. Mixed-evergreen forest occurs throughout the Big Sur region but is most widespread in the north, on north facing slopes, at mid-elevations near the coast. Relative basal area of the dominant tree species changes fairly predictably from north to south and from coast to interior in relation to mapped patterns of precipitation, temperature factors and soil characteristics. Most dominant tree species sprout vigorously after fire. The forests experience a mixed-fire regime in this region ranging from low severity understory burns to high severity crown fires, with the latter increasing above the marine inversion layer and at more interior locations. Ceanothus spp. can dominate mixed-evergreen sites for several decades after severe fires. All of the dominant broadleaf evergreen tree species are hosts of P. ramorum, although not all will die from infection. Tanoak mortality decreases from northwest to southeast and is significantly correlated with climate, especially growing degree days and mean annual precipitation, and with basal area of the foliar host bay laurel (Umbellularia californica) in a 0.5–1 ha neighborhood. Adaptive management of mixed-evergreen forest to mitigate P. ramorum impacts in the region will need to consider large local and regional variation in forest composition and the potentially strong interactions between climate, fire, forest composition and disease severity.     

Forecasts of urbanization scenarios reveal trade-offs between landscape change and ecosystem services

Context

Expansion of urban settlements has caused observed declines in ecosystem services (ES) globally, further stressing the need for informed urban development and policies. Incorporating ES concepts into the decision making process has been shown to support resilient and functional ecosystems. Coupling land change and ES models allows for insights into the impacts and anticipated trade-offs of specific policy decisions. The spatial configuration of urbanization likely influences the delivery and production of ES.

Objective

When considering multiple ES simultaneously, improving the production of one ecosystem service often results in the decrease in the provision of other ES, giving rise to trade-offs. We examine the impact of three urban growth scenarios on several ES to determine the degree to which spatial configuration of urbanization and the development of natural land cover impacts these services over 25 years.

Methods

We couple land change and ES models to examine impacts to carbon sequestration, surface water-run off, nitrogen and phosphorus export, organic farming and camping site suitability, to determine trade-offs among the six ES associated with each spatial configuration for western North Carolina.

Results

Consequences of urban configurations are dramatic, with degraded ES across all scenarios and substantial variation depending on urban pattern, revealing trade-offs. Counter-intuitive trade-offs between carbon sequestration and lands available for organic farming and camping were observed, suggesting that no configurations result in mutual benefits for all ES.

Conclusions

By understanding trade-offs associated with urban configurations, decision makers can identify ES critical to an area and promote configurations that enhance those.

     

Multi-temporal trajectories of landscape change explain forest biodiversity in urbanizing ecosystems

Context

Forest loss and fragmentation negatively affect biodiversity. However, disturbances in forest canopy resulting from repeated deforestation and reforestation are also likely important drivers of biodiversity, but are overlooked when forest cover change is assessed using a single time interval.

Objectives

We investigated two questions at the nexus of plant diversity and forest cover change dynamics: (1) Do multitemporal forest cover change trajectories explain patterns of plant diversity better than a simple measure of overall forest change? (2) Are specific types of forest cover change trajectories associated with significantly higher or lower levels of diversity?

Methods

We sampled plant biodiversity in forests spanning the Charlotte, NC, region. We derived forest cover change trajectories occurring within nested spatial extents per sample site using a time series of aerial photos from 1938 to 2009, then classified trajectories by spatio-temporal patterns of change. While accounting for landscape and environmental covariates, we assessed the effects of the trajectory classes as compared to net forest cover change on native plant diversity.

Results

Our results indicated that forest stand diversity is best explained by forest change trajectories, while the herb layer is better explained by net forest cover change. Three distinct forest change trajectory classes were found to influence the forest stand and herb layer.

Conclusions

The influence of forest dynamics on biodiversity can be overlooked in analyses that use only net forest cover change. Our results illustrate the utility of assessing how specific trajectories of past land cover change influence biodiversity patterns in the present.

     

Rapid sampling of plant species composition for assessing vegetation patterns in rugged terrain

Detailed species composition data are rapidly collected using a high-powered telescope from remote vantage points at two scales: site level and patch level. Patches constitute areas of homogeneous vegetation composition. Multiple samples of species composition are randomly located within the patches. These data are used as site-level data and are also aggregated to provide species composition data at the patch level. The site- and patch-level data are spatially integrated with high resolution (10 m), topographically-derived fields of environmental conditions, such as solar radiation, air temperature, and topographic moisture index in order to evaluate the applicability of the sampling method for modeling relationships between species composition and environmental processes.The methodology provides a balance between sampling efficiency and the accuracy of field data. Application of the method is appropriate for environments where terrain and canopy characteristics permit open visibility of the landscape. We evaluate the nature of data resulting from an implementation of the remote sampling methodology in a steep watershed dominated by closed-canopy chaparral. Analyses indicate that there is minimal bias associated with scaling the data from the site level to the patch level, despite variable patch sizes. Analysis of variance and correlation tests show that the internal floristic and environmental variability of patches is low and stable across the entire sample of patches. Comparison of regression tree models of species cover at the two scales indicates that there is little scale-dependence in the ecological processes that govern patterns of species composition between the site level and patch level. High explanatory power of the regression tree models suggests that the vegetation data are characterized at an appropriate scale to model landscape-level patterns of species composition as driven by topographically-mediated processes. Patch-level sampling reduces the influence of local stochasticity and micro-scale processes. Comparison of models between the two scales can be useful for assessing the processes and associated scales of variability governing spatial patterns of plant species.     

Perspectives of spatial scale in a wildland forest epidemic

The challenge of observing interactions between plant pathogens, their hosts, and environmental heterogeneity across multiple spatial scales commonly limits our ability to understand and manage wildland forest epidemics. Using the forest pathogen Phytopthora ramorum as a case study, we established 20 multiscale field sites to analyze how host-pathogen-environment relationships vary across spatial scales of observation in a wildland pathosystem. We developed statistical models of disease intensity across five nested levels of spatial aggregation, from an individual host through four broader spatial extents of observation. Analyses were conducted from two spatial perspectives: a focal view, where disease intensity at one scale was examined as a function of broader-scale landscape conditions, and an aggregate view, where disease intensity and landscape conditions was observed at the same scale of spatial aggregation. For each perspective, separate models were developed to compare direct field measurements of host density versus less expensive remotely sensed estimates of host habitat as predictors of disease in landscape-scale studies. From both perspectives, models using direct measurements of host density performed better than models using remotely sensed estimates of host habitat across all four spatial extents. We found no significant difference in model performance at the individual level. From the focal view, the performance of host density models declined with increasing spatial extent, whereas the performance of host habitat models improved with spatial extent. These results illustrate how the scale of observation – both spatial extent and measurement detail – can influence conclusions drawn from epidemiological models of wildland pathosystems.     

A regional analysis of total nitrogen in an agricultural landscape

Techniques for modeling spatial variability in the loss, gain, and storage of total nitrogen (N) in an agricultural landscape were developed utilizing a geographic information system (GIS) based on the Map Analysis Package (C.D. Tomlin, Yale University). The study area is a well-monitored portion (upper 114.9 km²) of the Little River Watershed, located near Tifton, Georgia, U.S.A. On the basis of measured N in the soil and vegetation, and the gains and losses of N by stream discharge, fertilizer, precipitation, N fixation, crop harvest, etc., it was possible to quantify and map source and sink regions of Total N, and to calculate a mass balance of N for an entire year. Results indicate massive flows of N, especially from anthropogenic sources. However, for the watershed as a whole, the N is virtually in balance with a small accretion occurring mostly in the riparian zones. Stream discharge of total N indicates that this landscape is well-buffered against excessive losses of N despite the large agricultural inputs.