B43C-1440
Recent Signal of Vegetation Change in Siberia Using Satellite Data
Global vegetation index data based on the observation of the satellite NOAA from 1980's enabled us to analyze the vegetation change over continental-scale regions for 20 years and more. An increasing trend of vegetation in boreal forests has been reported by some previous studies. Since the signal of such trend in the vegetation is probably apparent in the ecotone, detailed investigation should be required in the region of ecotone. This study targeted the ecotone from the boreal forest to the tundra in eastern Siberia, and examined the 19-year (1982 to 2000) trend of the satellite-derived vegetation index by using two kinds of parameters, i.e., the annual accumulative NDVI (ΣNDVI) and the annual maximum NDVI (MaxNDVI). As a result, we confirmed that ΣNDVI showed an increasing trend in the boreal forest as previous studies pointed out. By contrast, an increasing trend of MaxNDVI was found over the region to the north of the boreal forest (transitional zone between the boreal forest and the tundra). This suggests that signals due to the biomass increase or the vegetation change occurred in the region are detected.
B43C-1441
Validation of an Individual-Based Gap Model of the Eurasian Boreal Forest With Remote Sensing Imagery Analysis
The boreal forests of the Earth provide a significant source of renewable natural resources and are capable of sequestering and storing the carbon emitted by human industrial activities. This storage is estimated to be in the range of nearly 31 X 1012 kg of carbon in the trees alone (Kuusela, 1990). Annual storage estimates for the boreal forest suggest between 1-2 gigatons of carbon are captured in pools in the boreal regions, which amounts to nearly 30% of all annual anthropogenic carbon emissions (Bousquet et al. 1990). These systems are of critical importance to the climate system (Bonan et al. 1992) and for mitigating the effects of human activities on the Earth's climate. These boreal systems are likely to receive an enhanced warming effect due to global climate change (Smith et al. 1998). The result will be a shift in traditional ecotone distribution and a release of stored carbon (Harmon, Ferrell 1990). Scientists believe that these changes could result in a positive feedback loop of warming to the system. Thus, mapping current biomass stores are valuable for assessing the scale of future warming. Modeling efforts are underway to accurately map the Earth's boreal forests. The individual-based gap model FAREAST can accurately model many components of the boreal forest including carbon storage and species composition. Using this mechanism to model the boreal forest carbon system seems to be a viable methodology, one that is more spatially acute than other methods due to the geographic scale of the gaps it creates. With an expansion of the FAREAST model to a regional scale, it is possible for much of the Eurasian boreal forest to be accurately modeled and mapped. Large scale validation of the model is the remaining step in the modeling process. High resolution remote sensing data can be used to examine forest ecosystem characteristics with a minimum of field validation. This paper investigates the use of remotely sensed data to validate the gap model FAREAST in the Russian and Chinese boreal forest by predicting the location of species transition zones. In pristine landscapes, FAREAST predictions and remote sensing imagery align significantly. In disturbed landscapes, FAREAST is not capable of accurately predicting landscapes compared to remote sensing analysis. The efficacy of remote sensing imagery analysis to reveal necessary changes in the model suggests that imagery analysis is a viable solution to ecosystem model validation and also indicates the ability for high resolution sensors to extract fine ecosystem characteristics.
B43C-1442
Remote Sensing of Boreal Forest Biophysical and Inventory Parameters: A Review
Vegetation makes up nearly 70 % of the Earth's terrestrial surface and products from vegetated systems are vitally important for human populations. The growing need to manage vegetation resources at regional and global spatial scales has led to the increased use of remote sensing technologies among forestry scientists and managers for use in their investigation and supervision of forested landscapes. With a panoply of extant and developing airborne and satellite sensors, as well as multiple analysis techniques, there is a need to discern the most acceptable methods in which to examine remotely sensed imagery for forest ecosystem parameters. This includes both biophysical and inventory indicators. This study investigates the methods used to examine plant parameters in the boreal forest, and attempts to derive the most appropriate methods for extracting information regarding plant structure and stand information. A suggested methodology is constructed for use by remote sensors and forest managers. Specifically, we reviewed the literature on the remote sensing of boreal forests that featured airborne and satellite optical, passive and active radar, and lidar systems in order to determine if common frameworks for monitoring and assessing change in forest biophysical and inventory parameters could be developed. Other important remote sensing techniques such as change detection and land cover identification were also examined. Our review considered the purpose of each study, the type of sensor(s) used [e.g., Landsat or Lidar], where the study occurred, the methods used, including what vegetation and soil parameters or processes were considered, and the remote sensing indicator developed to measure this parameter [e.g., the normalized difference vegetation index (NDVI) is a surrogate for phytomass, LAI, land cover type, and other plant parameters]. We also investigated how the measured indicators were calibrated and validated as well as the limitations of the sensors that were expressed in conducting a particular study. The most effective methodology for each parameter and/or variable was described. Additionally, we produced and posted in the environment and conservation section of the Google Earth Community (http://bbs.keyhole.com) an annotated spatial bibliography of fifty of these studies.
B43C-1443
A Multivariate Approach for Using Satellite Imagery to Map the Composition and Structure of Forests Susceptible to Insect Disturbance: Application to the Simulation of Carbon Dynamics in Northern Minnesota and Ontario
Compared to other forest disturbances, insects and disease influence the largest area of forests in both the U.S. and Canada, affecting an estimated 50 million acres in the U.S. with economic costs over $1.5 billion. The successful understanding and modeling of ecosystem impacts of insect disturbances (especially for carbon dynamics) requires good knowledge of the spatial distribution, density and structure of host species on the landscape. In this study, we mapped the distribution of host species for the spruce budworm ( Choristoneura fumiferana) to facilitate landscape scale planning and modeling of outbreak dynamics. Spruce budworm is one of the most destructive indigenous pests in sub-boreal and boreal spruce-fir forests in the United States and Canada. Although periodic outbreaks are part of the natural cycle in these forests, traditional forest management practices may be responsible for increasing the frequency and severity of outbreaks. Currently, accurate spatially explicit forest structure data for such endeavors remains a persistent challenge and considerable research has focused on using remote sensing to identify methodologies to facilitate accurate estimation of stand volume and/or biomass. We used multi-temporal, multi-seasonal Landsat data and over 230 ground truth plots (and 220 additional validation plots) to map basal area (BA), for over two million hectares of forest in northern Minnesota and neighboring Ontario. BA was mapped both overall and for two spruce budworm host tree species ( Picea glauca and Abies balsamea) using partial least squares (PLS) regression applied to raw spectral bands, various spectral derivatives, and ground truth data. Results of the PLS regression yielded reasonable estimates of overall forest BA with an adjusted R2 of 0.62 and RMSE 4.67 m2 ha-1. White spruce relative BA had an adjusted R2 of 0.88 (RMSE 12.57 m2ha-1) and balsam fir relative BA had an adjusted R2 of 0.64 (RMSE 6.08 m2ha-1). The method also produced estimates for proportional cover of deciduous and evergreen species, with each having adjusted R2 values of 0.86 (RMSE 9.89 and 9.78 m2ha- 1, respectively). Because ground based measurements were placed largely in forest stands containing spruce and fir, modeled results show considerable confusion with non-target conifers, such as pines and cedar. Research is currently aimed at improving results by expanding ground-based measurements to include more non-target forest types to strengthen models. PLS regression has proven to be an effective data fusion tool for regional mapping of forest structure within spatially heterogeneous forests. Ongoing research is aimed at including other large-format sensors, such as Radarsat, to expand the capacity for modeling regional forest structure.
B43C-1444
Estimation of Tree Height, Biomass, and Standing Carbon in Miombo Woodlands Using Radar Interferometry
Savannas and woodlands are a major component of the world's vegetation covering one-sixth of the global land surface and one-half of the African continent. They account for about 30% of the primary production of all terrestrial vegetation. The southern African savannas cover 54% of the sub-continent with a plant diversity of approximately 8500 species and approximately 50% endemism. Miombo covers about two thirds of Mozambique and estimations of its biomass are critical because ecosystem services provided include food, fiber, and fuel for 39 million rural peoples and another 15 million urban dwellers in southern Africa. The Shuttle Radar Topography Mission (SRTM) C-band derived digital terrain model (DTM) can be used to estimate tree height by subtracting a base-level digital elevation model (DEM) from the calibrated SRTM. SRTM C-band's wavelength is such that there is partial penetration of the tree canopy before scattering which results in an underestimate of tree height. Consequently, mean tree height data from 50 30-m x 30-m random-stratified field plots in Niassa Reserve were used to bias the SRTM data up to average tree height and thus calibrate. However, DEMs in developing countries, particularly Africa, are not usually present and have to be developed either from field survey, orthophotography, or topographic maps. We derived a bare-ground binary mask from a land cover map of Niassa Reserve in northern Mozambique. The land cover map was generated from a Landsat Enhanced Thematic Mapper (ETM+) scene and the binary mask was overlaid against the SRTM to derive ground elevations from the SRTM. The resulting point map of elevations was spatially interpolated using thin plate spines with tension to derive a base-level DEM. The DEM was then subtracted from the calibrated SRTM to get tree heights. Secondly we explored the derivation of an independent base elevation DEM using the last return of the NASA Geoscience Laser Altimeter System (GLAS) and compared this to the bare-ground mask approach. Tree heights in conjunction with plant allometric equations were then converted to estimates of biomass and standing carbon in Miombo woodlands. Future studies will focus on the demographic relationship between tree age and height using tree ring data to facilitate backward and foreword reconstructions of vegetation history.
B43C-1445
A Comparison Of Tree Crown Recognition Techniques
The derivation of tree crown maps from high resolution remote sensing yields critical information on forest structure that can be used, via allometric relationships, to predict per-tree biomass, crown bulk density, and leaf area index. These relationships have the potential to be much more accurate at predicting biomass than coarse scale passive and active approaches since they do not rely on the direct retrieval of signals from the deep canopy, and rather use the areal crown extent linked with allometric relationships. However, techniques for mapping tree crowns are still being developed, and carry their own set of assumptions and errors, namely in their ability to accurately detect trees as individual scene objects (omission/commission errors), and the ability to accurately predict crown extent. We compare several published tree crown mapping algorithms for their ability to predict the presence/absence of trees and their individual crown area using photointerpreted subscenes and field data for validation.
B43C-1446
Deconvolution of multi-peak ICESat/GLAS waveforms
Although primarily designed for cryosphere studies, data from ICESat/GLAS currently provide the only source of global vegetation height mapping. The objective of this research is to examine the methodological techniques and accuracy of lidar waveform analysis for 3d vertical structure using ICESat/GLAS. This research will investigate the ranging techniques and methods (deconvolution and decomposition) for discriminating various features or reflecting surfaces within each returned waveform. The returned waveform energy detected by the digitizer is a function of the scattering elements within the energy path and the impulse response of the system. By knowing the impulse response of the system, this signal can be removed from the return waveform to improve separability between targets along the laser path. Heights derived from the deconvolution methodology will be assessed against heights derived from Gaussian decomposition of the returned waveform. Here, the assumption that the return waveform is a modeled composite of Gaussian distributions from multiple scatters falling along the laser path. This is currently the technique (up to 6 Gaussians) that is implemented on ICESat/GLAS processing. The White Sands Missile Range (WSMR) Space Harbor area in New Mexico is used as a precision calibration and validation site for ICESat, with experiments operated and maintained by the University of Texas at Austin Center for Space Research (UTCSR). The returned waveforms from an array of corner-cube reflectors placed on poles of known heights located at the WSMR will be used to evaluate the deconvolution and decomposition results.
B43C-1447
Interpretations of GLAS LiDAR for the Tapajos National Forest, Brazil
LiDAR remote sensing has proven to be a valuable source of information for characterization of forest structure. We conducted a study at the Tapajós National Forest (TNF) in the state of Pará, Brazil (centered at 3.56S 55.06W) to understand how forest structural properties interpreted from GLAS derived forest heights compared to a more traditional forest classification. The vegetation classification map was based on forest surveys, topography, soils, and interpretation of Landsat data. The original map groups TNF into 16 vegetation classifications. Using approximately 1500 GLAS waveform height predictions (Lefsky, ICESat Vegetation Product, heights ver.0.2) we calculated the 10th and 90th percentile values of distributions. These were interpreted as signals of disturbance and potential forest stand height respectively. We found no clear agreement on an area by area basis though general coherent patterns were observed. Areas close to human populations and those with high water-table depths showed a lower 10th percentile signal indicative of frequent recent disturbance. High plateau areas on clay soils had the greatest 90th percentile values. Our data suggests that statistical interpretation of GLAS may be valuable for comprehensive analyses of forest structure.
B43C-1448
Species Identification From Spatial Analysis of Digital Imagery
Our research tested whether spatial-spectral analysis techniques, when applied to digital camera imagery of wetlands, allow discrimination of plant species. We collected oblique imagery of emergent wetlands species as well as nearby upland vegetation at more than 100 sites in the Sacramento-San Joaquin River Delta. We limited our analysis to those portions of each oblique image in which the spatial resolution was calculated to be between 1 mm and 10 cm and applied the texture analysis techniques developed by RM Haralick and available in the ENVI software. Our results suggest that plant species can be discriminated with reasonable accuracy in oblique photography with this range of spatial resolutions.
B43C-1449
Leaf Area Index Modeling and Mapping in Conifer Forests using Multispectral and Lidar Data
Leaf area index (LAI; the ratio of half the total needle surface area per unit ground area) is an essential conifer forest structural characteristic for quantifying biosphere-atmosphere carbon and water fluxes because it largely controls the fraction of photosynthetically active radiation absorbed by vegetation. Most previous attempts to estimate LAI from remotely sensed data have relied on empirical relationships between field-measured observations and various spectral vegetation indices (VIs) derived from optical imagery or the inversion of canopy radiative transfer models. However, as biomass within an ecosystem increases, accurate LAI estimates are difficult to quantify with VI/empirical methods due to the asymptotic relationships of VIs and LAI. Traditional methods are also complicated by complex topography, variable species composition, and heterogeneous spatial structure of forest canopies. Here, we identify and compare the functional relationships between field-measured LAI quantities and various SPOT satellite image-derived VIs, height distribution metrics derived from small footprint, discrete-return lidar, and the extent to which integration of both lidar and spectral datasets can accurately estimate LAI over a broad range of coniferous forest stand conditions in the northern Rocky Mountains. We find that LAI models derived from lidar metrics are only incrementally improved with the inclusion of multispectral VI data.
B43C-1450
Examination of Tropical Forest Structure Using Field Data and High Spatial Resolution Image Data
Structural properties of tropical forests are an important component in ecological studies, yet they are difficult to quantify. Remote sensing of forest canopy structure estimation has greatly advanced to due the aid of high resolution satellite images. Field based methods of canopy structure have also improved due to the involvement of handheld laser range finders, which aid in gauging height, width, and depth of tree canopies. Using a handheld laser rangefinder we estimated canopy depth and generated canopy profiles from this data. Previously, we developed a crown characterization algorithm that uses high resolution satellite image data and have applied this algorithm in undisturbed tropical forests with good results. In this work we have further developed the algorithm to examine canopy depth using two allometric equations, developed from field data, that relate crown width to the top of the canopy and bottom of the canopy. Modification of our original algorithm also involved the incorporation of site specific allometric equations developed from field based measurements. Automated analysis of IKONOS imagery was used to estimate the distribution of canopy elements at various heights and their spatial locations. A comparison between the field based data and the estimates derived from remotely sensed images was conducted at four sites throughout Amazonia. We further compared our estimates of canopy structure with results from large footprint LIDAR data from GLAS. Ability to estimate canopy profiles and forest structural properties in vast areas of the Brazilian Amazon using high resolution imagery will help us to understand the regional carbon balance.
B43C-1451
The use of Soil Reflectance Database and NDVI Time Series to compute an Adjusted Green Vegetation Fraction
The green vegetation fraction (Fg) is an important climate and hydrologic model parameter. Common methods to calculate Fg are simple linear mixing models between two NDVI end-members: bare soil NDVI (NDVIo) and full vegetation NDVI (NDVI∞). A common assumption is that NDVIo is close to zero and as a result it is generally chosen from the lowest observed NDVI values. However, the mean soil NDVI computed from 2906 samples is much larger (NDVI=0.21) and is highly variable (standard deviation=0.1). Because underestimation of NDVIo yields overestimations of Fg that are higher than 0.2 for any pixel with 0.2<NDVI<0.4, it is important to evaluate alternative methods to estimate NDVIo when no information on local soils is available. Here we describe and test a method to improve Fg estimation by combining a global soil database and time series of NDVI derived from MODIS Terra data. Fg is computed for each pixel using a subset of the soils database that respects the linear mixing model condition NDVIo ≤NDVIh, where NDVIh is the historical minimum for that pixel. Using in situ measurements of soil NDVI, from sites that span a range of land cover types, we show that this method improves the temporal Fg estimates in areas with no information on soil reflectance. We also estimate the spatial distribution of Fg errors by employing our method for the conterminous U.S. and comparing the new Fg with that derived using the standard method to calculate Fg.
B43C-1452
Retrieving canopy structure from synergy of multi-angle spectral and Lidar data
Recent empirical studies have shown that multi-angle spectral data are useful for predicting canopy height, forest cover density, fractional area distributions, stand basal area, tree height and biomass, but the physical reason for this correlation was not understood. We found that the canopy spectral invariants can explain the physics behind the observed correlation, where the wavelength independent directional escape probability is the variable that imbues the sensitivity of multi-angle spectral data to canopy structure. Spectral information is essential to extract the spectral invariant parameters from the measured signal. We also discuss how our finding can be used to retrieve vertical and horizontal dimensions of trees from synergistic analysis of multi-angle, hyperspectral and lidar data.
B43C-1453
Water Hyacinth Identification Using CART Modeling With Hyperspectral Data in the Sacramento-San Joaquin River Delta of California
Water hyacinth (Eichhornia crassipes) is an invasive aquatic weed that is causing severe economic and ecological impacts in the Sacramento-San Joaquin River Delta (California, USA). Monitoring its distribution using remote sensing is the crucial first step in modeling its predicted spread and implementing control and eradication efforts. However, accurately mapping this species is confounded by its several phenological forms, namely a healthy vegetative canopy, flowering canopy with dense conspicuous terminal flowers above the foliage, and floating dead and senescent forms. The full range of these phenologies may be simultaneously present at any time, given the heterogeneity of environmental and ecological conditions in the Delta. There is greater spectral variation within water hyacinth than between any of the co-occurring species (pennywort and water primrose), so classification approaches must take these different phenological stages into consideration. We present an approach to differentiating water hyacinth from co-occurring species based on knowledge of relevant variation in leaf chlorophyll, floral pigments, foliage water content, and variation in leaf structure using a classification and regression tree (CART) applied to airborne hyperspectral remote sensing imagery.
B43C-1454
Estimation of Leaf Area Index (LAI) Through the Acquisition of Ground Truth Data in Yosemite National Park
Leaf area index (LAI) is an important indicator of ecosystem health. Remote sensing offers the only feasible method of estimating LAI at global and regional scales. Land managers can efficiently monitor changes in vegetation by using NASA data products such as the MODIS LAI 1km product. To increase confidence in use of the MODIS LAI product in Yosemite National Park, we investigated the accuracy of remotely sensed LAI data and created LAI maps using three optical in-situ instruments: the LAI-2000 instrument, digital hemispheric photography (DHP), and the Tracing Radiation and Architecture of Canopies (TRAC) instrument. We compared our in-situ data with three spectral vegetation indices derived from Landsat Thematic Mapper imagery: Reduced Simple Ratio (RSR), Simple Ratio (SR), and Normalized Difference Vegetation Index (NDVI) to produce models which created LAI maps at 30m and 1km resolution. The strongest correlations occurred between DHP LAI values and RSR. Pixel values from the 1km LAI map were then compared to pixel values from a MODIS LAI map. A strong correlation exists between our in-situ data and MODIS LAI values which confirms its accuracy for use by the National Park Service as a decision support tool in Yosemite. The MODIS LAI product is particularly useful because of its high temporal resolution of 1-2 days and can be used to monitor current and future vegetation changes. The model created using the in-situ data can also be applied to Landsat data to provide thirty years of historical LAI values.
B43C-1455
Mapping Tropical Cyclone Damage to Mangrove Habitats: An Example from South Florida
The impacts of two hurricanes (Katrina and Wilma) were assessed on 44,390 ha of protected mangroves in southwest Florida using a series of 20m multispectral SPOT and 1-km MODIS images. Established empirical relationships between mangrove leaf area index (LAI) and the normalized difference vegetation index (NDVI) were used to generate four different LAI maps before and after the hurricanes. These maps were compared to gridded LAI data (MOD 15A2) derived from the Moderate Resolution Imaging Spectrometer (MODIS) on board the Terra satellite. A semi-empirical approach based on the GeoSAIL radiative transfer model was also developed to estimate both pre- and post-Hurricane LAI based on SPOT band 2 (red). The results indicated modest agreement (r = 0.53) between one empirical formula and MODIS LAI pre-hurricane; while the GeoSAIL approach produced good agreement between post-hurricane MODIS LAI and SPOT LAI (r = 0.69) among the different methods employed. The results suggest that in the absence of detailed field data a relatively simple radiative transfer model linked to multispectral satellite imagery may provide an effective way to monitor the damage and subsequent recovery of mangrove ecosystems following major disturbances such as tropical cyclones. http://www.as.miami.edu/geography/Climatology/
B43C-1456
Severe storms and blow-down disturbances in the Amazon forest
Large natural disturbances (> 1 ha) in old-growth tropical forests are caused by a variety of processes such as landslides, fires, wind, and cyclonic storms. We analyzed the pattern of large forest disturbances apparently caused by severe winds (blow-downs) in a mostly unmanaged portion of the Brazilian Amazon using a longitudinal transect of Landsat images (27 scenes between 6°43'W 68°50'S and 2°16'W 51°51'S) and daily precipitation estimates based on NOAA satellite data. We found 170 blow-downs with an average area of 3 km2. Most blow-down disturbances occurred in the Western Amazon between 67°W and 58°W. A map of heavy rainfall (> 20 mm d-1) showed that the maximum frequency of heavy daily rainfall (~80 days y-1) occurred around 63°W in our study region. We found a close relationship between the frequency of heavy storms and the occurrence of blow-down disturbances events. This, in turns, suggests a close connection between severe weather and the rate of forest turnover caused by blow-down disturbances. The forest turnover time calculated for these disturbances within 9 Eastern Landsat scenes studied was almost 9000 years whereas for the 18 scenes in the Western Amazon, turnover time was closer to 1200 year. Large disturbances may have a significant influence on the spatial pattern of forest dynamics and productivity of the Amazon.
B43C-1457
Integrating eddy covariance estimates of GPP, LUE, and airborne lidar estimates of fPAR for local to regional scaling and assessment of GPP from MODIS
The adequate representation of within-pixel heterogeneity is essential for the extrapolation of CO2 and water fluxes from the canopy to regional and global scales. In this study, we use small-footprint, discrete-pulse-return airborne light detection and ranging (lidar) data to estimate spatial and temporal variation of gross primary productivity (GPP) across a post-harvest jack pine chronosequence in Saskatchewan, Canada. GPP is estimated from lidar-derived fraction of photosynthetically active radiation absorbed by the canopy (fPAR), measured incoming PAR radiation, and two light use efficiency (LUE) terms. The first GPP model uses maximum LUE based on species type and age and the second is based on the relationship between vegetation fractional cover from lidar and decreasing LUE. Both are controlled using meteorological driving mechanisms. The results are compared with eight-day products from the Moderate Resolution Imaging Spectroradiometer (MODIS) at the flux tower MODIS pixel level. In order to understand better some of the biases associated with scaling, we use GPP estimates from lidar to scale from pixel resolutions of 1 m to 25 m, 250 m, 500 m, and 1000 m, within both homogeneous and heterogeneous areas. Species and age-based LUE models are an improvement over biome-based LUE models used within the MODIS GPP algorithm. Differentiation between cloudy periods with diffuse shortwave radiation and sunny days with direct shortwave radiation also improve GPP estimates using species and age-based maximum LUE. Where maximum LUE models are not available per species and age class, increasing fractional cover of vegetation can be used to reduce LUE. When compared with measured GPP from EC for pixels containing towers, MODIS tends to over-estimate GPP by less than 30% at the mature jack pine site, but becomes more biased with decreasing age and increasing site heterogeneity. Increasing site heterogeneity creates biases when scaling GPP from 1 m to 1 km resolutions, especially as a result of land cover patch edges and differences (e.g. patches of forest and patches of cleared land within a single MODIS pixel).
B43C-1458
Predicting tree-level forest structure from LiDAR data
This research evaluates the efficacy of statistical imputation models (e.g., the k-nearest neighbor algorithm) incorporating LiDAR data to predict and map tree-level forest structure data (individual tree height, diameter at breast height, and species) across an 88,000 ha study area in Northern Idaho, USA. The primary objective is to provide spatially explicit data to parameterize the Forest Vegetation Simulator (FVS), a forest growth model that operates at the individual tree level, so that forest growth can be modeled across the entire study area. In addition to FVS parameterization, the imputed forest structure data could be used for many purposes including forest commodity assessment, carbon accounting, wildlife habitat modeling, etc. The final imputation models utilize LiDAR derived intensity and height measurements as well as LiDAR DEM topographic variables, to predict tree- level forest structure data. The imputed forest structure data are compared to independent ground-based forest inventory data via statistical equivalence models. Initial results indicate that the imputed data are equivalent (± 15 %) to the independent forest inventory data. Further development of this method will provide tree- level forest inventory data across large spatial extents.
B43C-1459
A Machine Learning Approach to Modeling Old-Growth Using Spectral and Topographic Data
Forest land managers and researchers are faced with a myriad of ecological, wildlife, management, and legal issues directly tied to old-growth vegetation structures. However, lack of adequate old-growth inventories has hindered decision making. Spectral remote sensing has previously proved inadequate in filling this gap. Presented is machine learning approach that leverages both spectral and topographic data to describe old- growth niches and accurately predict presence/absence. Two sources of forest inventory plot data were utilized; an operationally derived and a targeted sample. Results show that although the targeted sample provided more accurate results, a strong bias is evident making landscape inference erroneous. The operational sample contained insufficient information to adequately describe the range of old-growth across the study area. The best results were provided by a combination of the data. We provide not only a new modeling framework for old-growth but also recommendations on sample design and utilization of preexisting data to avoid sampling bias and improve predictions.
B43C-1460
Modeling the Impact of Vegetation Structure on Canopy Radiative Transfer for a Global Vegetation Dynamic Model
The transmission of light through plant canopies results in vertical profiles of light intensity that affect the photosynthetic activity and gas exchange of plants, their competition for light, and the canopy energy balance. The accurate representation of the canopy light profile is then important for predicting ecological dynamics. The study presents a simple canopy radiative transfer scheme to characterize the impact of the horizontal and vertical vegetation structure heterogeneity on light profiles. Actual vertical foliage profile and a clumping factor which are functions of tree geometry, size and density and foliage density are used to characterize the vertical and horizontal vegetation structure heterogeneity. The simple scheme is evaluated using the ground and airborne lidar data collected in deciduous and coniferous forests and was also compared with the more complex Geometric Optical and Radiative Transfer (GORT) model and the two-stream scheme currently being used to describe light interactions with vegetation canopy in most GCMs. The simple modeled PAR profiles match well with the ground data, lidar and full GORT model prediction, it performs much better than the simple Beer's&plaw used in two stream scheme. This scheme will have the same computation cost as the current scheme being used in GCMs, but provides better photosynthesis, radiative fluxes and surface albedo estimates, thus is suitable for a global vegetation dynamic model embedded in GCMs.
B43C-1461
Evaluating the Potential of Waveform Lidar and Hyperspectral Data Fusion for Species Level Biomass Mapping.
Many studies have demonstrated the ability of waveform lidar to map forest structural metrics such as canopy height, canopy cover and above ground biomass with high accuracies over different forest cover and types. Hyperspectral data provides forest attributes complementary to lidar such as vegetation stress, moisture content and land cover at species level. This study explores and evaluates the combined potential of waveform lidar (LVIS) and AVIRIS hyperspectral imagery for species level biomass mapping in the Sierra Nevada spotted owl habitat.LVIS quartile heights and canopy cover along with spectral metrics and endmember fractions from AVIRIS were compared with field biomass measures using linear and stepwise regression.Water band indices and shade fractions from AVIRIS show moderate to strong correlation with LVIS canopy height and biomass for certain species. LVIS variables were found to be consistently good predictors of total biomass as well as species level biomass. The inclusion of AVIRIS metrics in combination with lidar added little explanatory value for biomass. However, biomass prediction at species level lowered residual error by 20% or more in comparison to total biomass estimates, suggesting that its main value for biomass mapping is through species-level stratification.
B43C-1462
Latitudinal Variation In Mangrove Height And Biomass in East Africa Using SRTM Elevation Data
One of the foundations of biogeography states that productivity and biomass are highest at the equator and decrease as latitude increases. This relationship was confirmed for mangrove ecosystems by Sanger and Snedaker in 1993, based on a review of mangrove biomass and litterfall studies. Recent advances in remote sensing technology have permitted the estimation of mangrove biomass using landcover maps and data from the Shuttle Radar Topography Mission (SRTM). A test of this paradigm of mangrove ecology, using Mozambique as a case study, did not confirm the relationship between mangrove biomass and latitude. In this study, we produced a height and biomass map of mangrove forests of the whole Western Indian Ocean (WIO) region, which spans from Somalia in the North (12N) to South Africa in the S (28S), including Madagascar. Landsat ETM+ data was used to produce landcover maps and SRTM elevation data to produce height and biomass maps for the WIO region. Based on our results, we did not find a significant correlation between latitude and height/biomass in mangrove forests. Our results suggest that that freshwater input and type of geographical setting (lagoons, bays, deltas and open coast) are more important in determining mangrove height and biomass.
B43C-1463
Analysis of Tropical Forest Structural Dynamics Using Medium-footprint Lidar
As a forest canopy recovers from a disturbance event, the vertical structure passes through various stages of biomass distribution until reaching an age at which it approximates the vertical structure of old-growth forest. Quantifying and mapping rates of biomass accumulation and distribution in the forest canopy has important implications for understanding carbon stocks and fluxes.The La Selva Biological Station in Costa Rica contains numerous sections of secondary forest at different stages of recovery from disturbance. We explore the vertical canopy structure of these forests at two different years and the progression of biomass distribution in the canopy over time using canopy information collected by the Laser Vegetation Imaging Sensor (LVIS). LVIS, a medium- footprint airborne scanning lidar, collected vegetation data over La Selva in March of 1998 and March of 2005. Waveforms and waveform-derived metrics are used to obtain canopy heights and vertical biomass distribution patterns and dynamics. We assess the potential of using medium-footprint lidar to determine successional status. The ability to remotely detect and map successional status can greatly improve carbon modeling and management.
B43C-1464
Remote Sensing of Coastal Marsh Vegetation Structure using Multi-angle Imaging
This study explores using a remote sensing technique based on imagery from multiple viewing angles that retrieves leaf area index (LAI) for coastal marsh vegetation. The method uses data from spaceborne instruments with high spatial resolution (~20 m) and multi-angular imaging, specifically the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) aboard the NASA spacecraft Terra or the Compact High Resolution Imaging Spectometer (CHRIS) aboard the Project for On-Board Autonomy (Proba). Off-nadir views are shown to provide information about the marsh canopy structure, which is mostly comprised of grasses, sedges, and rushes in moderate to high saline regimes. Water strongly affects the bi-directional reflectance distribution function (BRDF) for wetlands and cannot be handled by inversion of conventional canopy radiative transfer models. The first objective of this study was to extend current methods to account for the non-Lambertian, aquatic substrate. The methodology was then validated and tied to remote sensing instruments (e.g., ASTER and CHRIS) through ground truth. Initial study targeted sites in the Chesapeake Marshlands National Wildlife Refuge Complex. However, it is also likely that the techniques and products of this and future work will be widely applicable to assessment and monitoring of marshes globally and is the first step to develop models for other types of wetland.
B43C-1465
ICESat Estimates of Forest Canopy Height Loss for Post-Hurricane Timber Damage Detection and Assessment Decision Support
Along the Gulf Coast and Atlantic Seaboard, tropical storms and hurricanes annually cause defoliation and deforestation amongst coastal forests. Following a severe storm, there is an urgent need to assess the impact on timber growth so resources can be targeted to assist in recovery. It is also important to identify these damaged areas due to their increased risk of fire and susceptibility to invasive species. Current methods of detection involve assessment through ground-based field surveys, aerial surveys, computer modeling, space- borne remote sensing, and Forest Inventory and Analysis field plots. This project focuses on a need for methods that are at once more synoptic than field surveys and more closely linked to the phenomenology of tree loss and damage than passive remote sensing methods. The primary concentration is on the utilization of Ice, Cloud, and land Elevation Satellite (ICESat) data products to detect changes in forest canopy height as an indicator of post- hurricane forest disturbances. ICESat is a NASA spaceborne lidar mission that utilizes green and infrared light to determine land surface vertical structure in 70m elliptical footprints. While created to primarily measure polar ice sheet mass and cloud property information, it has proven successful in measuring forest canopy height. By analyzing ICESat data over areas affected by Hurricane Katrina, this study demonstrates that ICESat may serve as a useful indicator of a storm's direct effects as well as its long term consequences. http://develop.larc.nasa.gov
B43C-1466
Analysis of Early Forest Regrowth in the Eastern US using IceSAT/GLAS-LiDAR and Landsat- spectral data.
Forest-cover conversion, disturbance, and recovery have been proposed as key mechanisms for transferring carbon between the land surface and the atmosphere, yet the area and timing of these processes are still poorly quantified. Combining remote sensing products such as LiDAR and Landsat data can help quantify the amount, area and timing of forest disturbance along with estimated rate of recovery. This study examines the use of NASA's Geoscience Laser Altimeter System (GLAS) for assessing post-disturbance forest re-growth rates via "space for time" substitution –GLAS observations from a single year combined with 20+ year Landsat disturbance record. Landsat image time series from three locations in the Eastern US (Maine, Virginia, Mississippi) were analyzed to obtain the timing and magnitude of major disturbance events for the 1984-2003 period. GLAS waveforms from 2003 were extracted for these patches and heights were determined via visual inspection of the waveform. Only "high magnitude" (stand clearing) disturbance events were selected, and only from regions of low topographic relief (< 5 degrees). Height Measurements of forest stands undisturbed over the last 20 years were also obtained along the latitudinal transect. Results show a progression in stand height from youngest to oldest stand. Regrowth rates vary with ecoregion and climate, from 0.6 m/yr (Maine) to 1.0-1.2 m/yr (Virginia - Mississippi). The latter rates compare favorably with known values for southeastern loblolly pine. Although the precision of an individual GLAS-derived height is relatively low, this study demonstrates that by combining multiple space-for-time observations, we can measure landscape-scale growth rates on order of ~1 m/yr. Decreasing the diameter of the lidar footprint in future land missions may help to increase the accuracy of forest structure measurements.