Estimating Forest Species Composition Using a Multi-Sensor Approach
The magnitude, duration, and frequency of forest disturbance caused by the spruce budworm and forest tent caterpillar has increased over the last century due to a shift in forest species composition linked to historical fire suppression, forest management, and pesticide application that has fostered the increase in dominance of host tree species. Modeling approaches are currently being used to understand and forecast potential management effects in changing insect disturbance trends. However, detailed forest composition data needed for these efforts is often lacking. Here, we used partial least squares (PLS) regression to integrate satellite sensor data from Landsat, Radarsat-1, and PALSAR, as well as pixel-wise forest structure information derived from SPOT-5 sensor data (Wolter et al. 2009), to estimate species-level forest composition of 12 species required for modeling efforts. C-band Radarsat-1 data and L-band PALSAR data were frequently among the strongest predictors of forest composition. Pixel-level forest structure data were more important for estimating conifer rather than hardwood forest composition. The coefficients of determination for species relative basal area (RBA) ranged from 0.57 (white cedar) to 0.94 (maple) with RMSE of 8.88 to 6.44 % RBA, respectively. Receiver operating characteristic (ROC) curves were used to determine the effective lower limits of usefulness of species RBA estimates which ranged from 5.94 % (jack pine) to 39.41 % (black ash). These estimates were then used to produce a dominant forest species map for the study region with an overall accuracy of 78 %. Most notably, this approach facilitated discrimination of aspen from birch as well as spruce and fir from other conifer species which is crucial for the study of forest tent caterpillar and spruce budworm dynamics, respectively, in the Upper Midwest. Thus, use of PLS regression as a data fusion strategy has proven to be an effective tool for regional characterization of forest composition within spatially heterogeneous forests using large-format satellite sensor data.
Spatial radiation environment in a heterogeneous oak woodland using a three-dimensional radiative transfer model and multiple constraints from observations
B15: Remote Characterization of Vegetation Structure: Including Research to Inform the Planned NASA DESDynI and ESA BIOMASS Missions Title: Spatial radiation environment in a heterogeneous oak woodland using a three-dimensional radiative transfer model and multiple constraints from observations Hideki Kobayashi, Youngryel Ryu, Susan Ustin, and Dennis Baldocchi Abstract Accurate evaluations of radiation environments of visible, near infrared, and thermal infrared wavebands in forest canopies are important to estimate energy, water, and carbon fluxes. Californian oak woodlands are sparse and highly clumped so that radiation environments are extremely heterogeneous spatially. The heterogeneity of radiation environments also varies with wavebands which depend on scattering and emission properties. So far, most of modeling studies have been performed in one dimensional radiative transfer models with (or without) clumping effect in the forest canopies. While some studies have been performed by using three dimensional radiative transfer models, several issues are still unresolved. For example, some 3D models calculate the radiation field with individual tree basis, and radiation interactions among trees are not considered. This interaction could be important in the highly scattering waveband such as near infrared. The objective of this study is to quantify the radiation field in the oak woodland. We developed a three dimensional radiative transfer model, which includes the thermal waveband. Soil/canopy energy balances and canopy physiology models, CANOAK, are incorporated in the radiative transfer model to simulate the diurnal patterns of thermal radiation fields and canopy physiology. Airborne LiDAR and canopy gap data measured by the several methods (digital photographs and plant canopy analyzer) were used to constrain the forest structures such as tree positions, crown sizes and leaf area density. Modeling results were tested by a traversing radiometer system that measured incoming photosynthetically active radiation and net radiation at forest floor and spatial variations in canopy reflectances taken by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). In this study, we show how the model with available measurements can reproduce the spatially heterogeneous radiation environments in the oak woodland.
Comparative Influence of Terrain Slope and Canopy Closure on Lidar DEM Accuracy
The use of Lidar (Light Detection and Ranging) technology is becoming one of the most effective and reliable means of collecting a variety of terrain and vegetation data. Most Lidar based estimates come from the creation of digital elevation models (DEM). As a result of the DEM’s importance in using Lidar data as a management tool, it is necessary to understand the variables that influence the DEM accuracy. Two of these variables, terrain slope and percent canopy closure, were investigated in a mixed conifer forest and woodland area in central Idaho. Following the creation of a DEM from the last return Lidar points, a series of 54 fixed radius plots were stratified by 3 slope classes from 0% to 45% and 3 canopy closures classes from 0% to 95% and surveyed using a laser total station and differential GPS for DEM accuracy analysis. Within each tenth acre plot a grid consisting of 80 ground points was collected along with a point for each tree and each 1,000 hour fuel, further data was collected to classify the canopy and the fuel loading. Each plot was processed to create a DEM for comparison to the Lidar derived DEM. These results will have implications in the development and use of high-resolution DEM models derived from Lidar data for natural resource managers.
Biomass Accumulation Rates of Amazonian Secondary Forest and Biomass of Old-Growth Forests from Landsat Time Series and GLAS
We estimate the age of humid lowland tropical forests in Rondônia, Brazil, from a somewhat densely spaced time series of Landsat images (1975-2003) with an automated procedure, the Threshold Age Mapping Algorithm (TAMA), first described here. We then estimate a landscape-level rate of aboveground woody biomass accumulation of secondary forest by combining forest age mapping with biomass estimates from the Geoscience Laser Altimeter System (GLAS). Though highly variable, the estimated average biomass accumulation rate of 8.4 Mg ha-1 yr-1 agrees well with ground-based studies for young secondary forests in the region. In isolating the lowland forests, we map land cover and general types of old-growth forests with decision tree classification of Landsat imagery and elevation data. We then estimate aboveground live biomass for seven classes of old-growth forest. TAMA is simple, fast, and self-calibrating. By not using between-date band or index differences or trends, it requires neither image normalization nor atmospheric correction. In addition, it uses an approach to map forest cover for the self-calibrations that is novel to forest mapping with satellite imagery; it maps humid secondary forest that is difficult to distinguish from old-growth forest in single-date imagery; it does not assume that forest age equals time since disturbance; and it incorporates Landsat Multispectral Scanner (MSS) imagery. Variations on the work that we present here can be applied to other forested landscapes. Applications that use image time series will be helped by the free distribution of coregistered Landsat imagery, which began in December 2008, and of the Ice Cloud and land Elevation Satellite (ICESat) Vegetation Product, which simplifies the use of GLAS data. Finally, we demonstrate here for the first time how the optical imagery of fine spatial resolution that is viewable on Google Earth provides a new source of reference data for remote sensing applications related to land cover. Reference: Helmer, E. H., M. A. Lefsky and D. A. Roberts. 2009. Biomass accumulation rates of Amazonian secondary forest and biomass of old-growth forests from Landsat time series and the Geoscience Laser Altimeter System. Journal of Applied Remote Sensing 3:033505.
Flash Lidar Data Processing
Late last year, a prototype Flash LIDAR instrument flew on a series of airborne tests to demonstrate its potential for improved vegetation measurements. The prototype is a precursor to the Electronically Steerable Flash LIDAR (ESFL) currently under development at Ball Aerospace and Technology Corp. with funding from the NASA Earth Science Technology Office. ESFL may soon significantly expand our ability to measure vegetation and forests and better understand the extent of their role in global climate change and the carbon cycle - all critical science questions relating to the upcoming NASA DESDynI and ESA BIOMASS missions. In order to more efficiently exploit data returned from the experimental Flash Lidar system and plan for data exploitation from future flights, Ball funded a graduate student project (through the Ball Summer Intern Program, summer 2009) to develop and implement algorithms for post-processing of the 3-Dimensional Flash Lidar data. This effort included developing autonomous algorithms to resample the data to a uniform rectangular grid, geolocation of the data, and visual display of large swaths of data. The resampling, geolocation, surface hit detection, and aggregation of frame data are implemented with new MATLAB code, and the efficient visual display is achieved with free commercial viewing software. These efforts directly support additional tests flights planned as early as October 2009, including possible flights over Niwot Ridge, CO, for which there is ICESat data, and a sea-level coastal area in California to test the effect of higher altitude (above ground level) on the divergence of the beams and the beam spot sizes.
Above-ground Forest Biomass Estimation by ALOS/PALSAR over Boreal Forest in Alaska Accompanied with Ground-based Forest Survey
For the better understanding of the carbon cycle in the global ecosystem, investigations on the spatio-temporal variation of the carbon stock which is stored as vegetation biomass is important. The L-band microwave radar “PALSAR (Phased Array type L-band Synthetic Aperture Radar)” of the satellite “ALOS (Advanced Land Observing Satellite)” provides the information which can be used for the above-ground forest biomass (AGFB) estimation. An attempt to map the AGFB distribution over an ecotone region in Alaska was carried out based on ALOS/PALSAR data. In July 2007, a ground-based forest survey was executed in the south-north transect (about 500 km long) along a trans-Alaska pipeline which profiles the ecotone from boreal forest to tundra in Alaska. 29 forests along the transect were targeted for the survey, and their AGFB were measured. Consequently, it was revealed that the AGFB ranges from 5 to 100 ton/ha (dried matter). These ground-based AGFB measurements at 29 forests were compared with the signal (digital number) in 20 scenes of ALOS/PALSAR (HV polarization mode) that cover the 29 forests in July or August 2007. In addition, 16 areas of grassland in the images were picked for the reference value of the zero AGFB. The result showed a positive strong (r = 0.84) and linear relationship between them, demonstrating a feasibility of ALOS/PALSAR for the mapping of the AGFB. Based on the linear relationship, the AGFB was estimated and mapped over the ecotone region in Alaska. Generally, there is a south to north gradient in AGFB that reflects the AGFB gradient from southern forest-rich region to northern forest-sparse region in the ecotone. The AGFB in some regions in southern part reaches 100 Mg/ha.
Accuracy of DESDynI Biomass Estimates using Lidar and Data Fusion Methods
DESDynI (Deformation, Ecosystem Structure and Dynamics of Ice) is a NASA satellite mission that will provide global estimates of aboveground biomass at a maximum spatial resolution of 500 m and an accuracy of ±10 Mg C ha-1 (up to 20% of the total) after 5 years. This will be accomplished by 1) developing algorithms to predict biomass from full waveform lidar returns; 2) sub-sampling grid cells along contiguous tracks with an orbiting multi-beam lidar; and 3) fusing lidar observations with radar and multispectral data to upscale biomass estimates within grid cells. Uncertainties exist at each stage of this process, and the goal of this study was to perform an accuracy assessment and to identify steps where improvements might be needed. Summer field campaigns were conducted in 2003 and 2009 to collect ground-based measurements and remote sensing data for study areas in the boreal transition zone near Howland, Maine, USA. Model parameters and uncertainties associated with biomass predictions from lidar data were estimated using Bayesian methods, and information criterion were used to select models based on simplicity and goodness of fit. DESDynI orbits were simulated and used to subsample airborne scanning lidar data the Laser Vegetation Imaging Sensor (LVIS), and up-scaling to 500 m grid cells was accomplished with ASTER multispectral data and L-band radar available from satellite and airborne platforms (ALOS/PALSAR and UAVSAR, respectively). Methods developed for this case study will be applied to other DESDynI field campaign sites to obtain accuracy assessments for other forest ecosystems.
On the correct estimation of effective leaf area index: does it reveal information on clumping effects?
Effective leaf area index is routinely quantified with optical instruments that measures gap fraction through the probability of beam penetration of sunlight through the vegetation. However, there have been few efforts to obtain theoretically consistent effective leaf area index from those measurements. To apply Beer-Lambert law, multiple gap fraction measurements may be averaged in two ways: 1) take logarithm on individual gap fraction and then average the logarithms and 2) average gap fraction and take logarithm. Based on a theoretical model and gap fraction measurements from 41 sites, we report that effective leaf area index must be quantified using the second approach. The first approach is implemented in the LAI-2000 instrument and considers clumping effects from 35 to 100% when comparing with independent clumping index estimates. Thus, the combination of the first approach with independent clumping index estimates will overestimate leaf area index. Clumping effects accounted for by the LAI-2000 instrument, called as “apparent” clumping index, were dependent on canopy cover, crown shape, and canopy height. We show that apparent clumping index is a useful quantity to constrain the true clumping index and to investigate spatial and temporal variation of clumping effects. Such information would be useful to evaluate global canopy structure products such as the planned NASA DESDynI and ESA BIOMASS Missions.
LiDAR canopy height and shape measurements in a sagebrush-steppe ecosystem
Airborne LiDAR has the potential for estimating canopy characteristics at a range of scales appropriate for landscape assessments. Separating LiDAR returns for use in determining canopy height and shape in low-height vegetation is difficult because the vegetation canopy return is often close to the ground return in time and space. In addition, height underestimation is likely exacerbated in sparsely vegetated shrub ecosystems. This study will compare LiDAR point-cloud data to sagebrush canopy characteristics measured in the field. Comparisons will be made across differing scales (individual points to 5 m rasterized surface). If sagebrush is not adequately represented with the point density of the existing datasets (5 points/ m2), simulations will be performed to identify the appropriate point density needed to characterize sagebrush canopy and shape. Results will be used to understand the efficacy of making biomass measurements in this community.
Estimation of Tropical Forest Structure Using the Full Waveform Lidar from ICESat
The Amazon basin contains the world’s largest continuous tropical forest constituting 40% of the remaining area for this ecotype and is made up of heterogeneous canopies and forest communities with unique assemblages of tree species, complex vegetation dynamics and history, and high biodiversity. Forest structural components include canopy geometry and tree architecture, size distributions of trees, and are closely linked with ecosystem functioning. The dynamic processes of growth and disturbance are reflected in the structural components of forest. Large footprint lidar has been used to estimate biomass in tropical and temperate forests, primarily through the correlation with field measured height, basal area, and plot biomass estimates. However, in tall-stature forests height loses much of its correlation with basal area, so the height-biomass curve becomes asymptotic and is associated with greater error at large biomass values. Use of lidar in such an analysis also does not include estimations of other stand level structural properties. We used full lidar waveforms from ICESat GLAS to estimate forest stand structure. We developed a 3D canopy model that uses trunk or crown diameter distributions and allometric equations of associated crown depth and canopy height to generate a synthetic canopy. Using geometric series of tree size distributions, we generated thousands of synthetic vegetation profiles. These synthesized forest canopy profiles were rapidly and efficiently compared with lidar waveforms and matches identified using least squared difference. Using GLAS lidar waveforms, we identified patterns of forest structure across Amazonia. . Landscape level estimates of q-values derived from lidar estimates are similar to estimates of q-values from field based data from a 400 ha area in Tapajos National Forest, approximately q=1.7, with a range of 1.69 to 1.82 per 100 ha plot. Estimates comparing field data collected in areas associated specifically with a GLAS footprint were found to be similar.
Retrieving Biome Types from Multi-angle Spectral Data
Many studies have been conducted to demonstrate the ability of multi-angle spectral data to discriminate plant dominant species. Most have employed the use of empirically based techniques, which are site specific, requires some initial training based on characteristics of known leaf and/or canopy spectra and therefore may not be extendable to operational use or adapted to changing/unknown land cover. An ancillary objective of the MISR LAI/FPAR algorithm is classification of global vegetation into biome types. The algorithm is based on the 3D radiative transfer equation. Its performance suggests that is has valid LAI retrievals and correct biome identification in about 20% of the pixels. However with a probability of about 70%, uncertainties in LAI retrievals due to biome misclassification do not exceed uncertainties in the observations. In this poster we present an approach to improve reliability of the distribution of biomes and dominant species from multi angle spectral data. The radiative transfer theory of canopy spectral invariants underlies the approach, which facilitates parameterization of the canopy bidirectional reflectance factor in terms of the leaf spectrum and two spectrally invariant and structurally varying variables - recollision and directional escape probabilities. Theoretical and empirical analyses of ground and airborne data acquired by AVIRIS, AirMISR over two sites in New England and CHRIS/PROBA over BARAX site in Spain suggest that the canopy spectral invariants convey information about canopy structure at both the macro and micro scales. These properties allow for the natural separation of biome classes based on the location of points on the total escape probability vs the proportional escape ratio log-log plane.
Biometric Properties Estimated from High Resolution Imagery in the Amazon and the Cerrado Regions
The Amazon and Cerrado regions are unique ecotypes with complex and varied forest and vegetation structure. Forest structure reveals the dual influences of disturbance and growth. Because these two tropical regions have and are undergoing rapid change due to human encroachment, understanding the forests structure in these ecotypes aids in efforts to quantify carbon dynamics on both regional and global scales. Analysis of data from literature found that canopy cover and biomass are highly correlated in the Cerrado (r2=.86), more so than other structural variables. This indicates that use of radar and lidar to estimate biomass in savannah ecotypes with sparse and clumpy tree cover might be prone to error. Literature also suggests that lidar and radar saturate in high biomass forests. Remote sensing of forest canopy structure estimation has greatly advanced to due the aid of high resolution satellite images. We estimated forest structure using high resolution image data from IKONOS using textural methods such as lacunarity, semivariance, power spectrum, entropy, and a crown characterization algorithm for 11,014 image tiles or sections (1 square km each) extracted from 300 IKONOS images. Our preprocessing of this data calculated top-of-atmosphere reflectance based on metadata from IKONOS image acquisition. A user-trained five category landuse classification was used to determine which areas within an IKONOS tile would be analyzed using textural methods.We compare results with available field measured forest biometric data. We used an Index of Translational Homogeneity (ITH) calculated from our lacunarity results. ITH is an index of average crown width and we estimated an average of 8.1 m +/- 7.7 SD. Our estimate of the range based on semivariance was an average of 11.4 m +/- 7.3 SD. Our crown characterization algorithm estimated average crown width to be 12.5 m +/- 4.0 SD. The average entropy of each tile was 5.7 +/- 0.5 SD. We associated each IKONOS tile with one of thirteen vegetation classes developed by PROBIO. We found significant differences of textural measurements between some vegetation classes indicating that vegetation structure is able to be discerned using textural methods and that this structure is able to used to differentiate vegetation types. Finally, we associated high resolution forest structure estimates with coarser scale MODIS pixel values in an effort to scale across the Amazon and Cerrado regions. Our analytical methods for such scaling used both multivariate linear methods and Bayesian nonlinear regressions to match derived canopy characteristics from high resolution images (one set of variables for each tile) with spectral and angular moderate resolution reflectance data.
On the characteristics of Terra and Aqua MODIS LAI subsets at six FLUXNET sites
The MODIS sensor onboard the satellites Terra and Aqua launched within the framework of NASA's EOS (Earth Observing System) program has proven itself a key sensor in remote sensing of ecosystem dynamics and land surface processes. Together with FLUXNET, a network of tower measurements of ecosystem fluxes and climate variables, MODIS data provide a powerful framework for analyzing and predicting ecosystem behaviour. The interplay between the two data sources is supported by MODIS subsets which provide MODIS land products resampled around continuous field measurement sites such as the flux towers of FLUXNET to generate 7x7 pixels with a resolution of 1 km centered at the tower. A main product of these subsets is MOD15A2 providing estimates of the leaf area index (LAI). This product is used with increasing frequency; the ways of its usage, however, differ strongly. In this study, we analyze explicitly the effects of some fundamental post-processing methods exemplarily at six FLUXNET sites with three different vegetation classes. Specifically, we consider the combination of Aqua and Terra subsets, illustrate the impact of the application of several quality filters and compare the consequences of averaging over different window sizes around the tower pixels on the magnitude, temporal variability and consistency of LAI time series.
Development of an Integrated Hyperspectral Imager and 3D-Flash Lidar for Terrestrial Characterization
The characterization of terrestrial ecosystems using remote sensing technology has a long history with using multi-spectral imagers for vegetation classification indices, ecosystem health, and change detection. Traditional multi-band imagers are now being replaced with more advanced hyperspectral imagers, which offer finer spectral resolution and more specific characterization of terrestrial reflectances. Recently, 3-dimensional (3D) imaging technologies, such as radar interferometry and scanning laser rangers, have added a vertical dimension to the characterization of ecosystems. The combination of hyperspectral imagery with 3D Lidar allows for detailed analysis of terrestrial biomass, health and species identification. Recognizing the need, and the technical feasibility of this type of environmental assessment, the National Research Counsel has advocated two future NASA satellite missions to measure terrestrial ecosystem health and structure, the DESDynI and HyspIRI missions. These programs will orbit synthetic aperture radar, Lidar and a hyperspectral imager. Northrop Grumman has integrated a hyperspectral vis-IR imager and 3D-flash Lidar and flown it on a twin-otter aircraft platform for measurements of terrestrial ecology. The goal of the system is to demonstrate an integrated system design, similar to that flown by Asner et al. on the Carnegie Airborne Observatory, but with a design path to high altitude systems that could offer pathfinders for an operational satellite system. Lidar systems are typically limited to either low altitude small-footprint sampling or higher altitude broad pixel resolution. The Northrop Grumman system goals are to be able to image terrestrial ecosystems at small horizontal resolutions from high altitude, while maintaining a relatively broad swath capability. Performance of the integrated system during collections from the twin-otter will be discussed, as well as design performance for a dual sensor system for high altitude platforms that could offer early development of space based systems.
The Impact of Spatial Resolution on Classification Accuracy of Hyperspatial Imagery
Our research investigated approaches to classification of imagery having ground spatial resolutions from 0.5m to 1mm, testing the relative importance on classification accuracy of spatial resolution. The classifications involved only spectral information and the results underscore the increasing importance of spatial, as compared to spectral information as the spatial resolution in the imagery changes from 0.5m to 1mm.
Determining change in vegetation structure using small-footprint, waveform-resolving lidar
The Experimental Advanced Airborne Research Lidar (EAARL) is a raster-scanning, waveform-resolving, green-wavelength (532 nm) lidar designed to map near-shore bathymetry, topography, and vegetation structure simultaneously. For each laser pulse, the EAARL sensor records the time history of the return waveform within a small footprint (20-cm diameter at nominal flying altitude of 300 m), enabling characterization of vegetation canopy structure and “bare Earth” topography beneath vegetation. A collection of individual waveforms combined within a synthesized large footprint was used to define three metrics to quantify vegetation structure: canopy height (CH), canopy reflection ratio (CRR), and height of median energy (HOME). The appropriate size of the synthesized footprint was based on an approximate stand size of 5-m diameter. The methodology was applied to the Naval Live Oaks Area Preserve (NLO) in Gulf Islands National Seashore, where an EAARL survey was conducted in October 2005 and May 2007. Lidar-based vegetation metrics at NLO from the 2005 and 2007 surveys were used to define patches of vegetation communities that had undergone significant change. The presence/absence of vegetation change was evaluated using high-resolution digital aerial imagery acquired during the survey. The Gulf Coast Network National Park Service Inventory & Monitoring program is using this change analysis as part of its’ vegetation monitoring protocol.
Modeling Avian Richness Patterns with Texture Measures of Remotely Sensed Imagery
Avian biodiversity is under great threat, primarily from human influences. With limited resources for habitat conservation, the accurate identification of high-value bird habitat is crucial. One major factor known to influence an area’s avian biodiversity is habitat structure. While biodiversity maps exist, most cover small extents or employ coarse resolution. While habitat structure can be measured in the field, methods are time consuming and impractical for use over very large extents. Thus, in order to more easily identify high-value habitat for conservation, new methods are needed to characterize habitat structure at a fine spatial scale over broad extents. We evaluated the performance of image texture measures derived from remotely sensed data as a measure of habitat structure for the prediction of avian diversity. A suite of moving window based texture measures were calculated from 114 Landsat TM scenes circa 2000 covering the Midwestern United States. Average avian species richness over the period 1998-2002 was calculated for multiple avian guilds on approximately 800, 39 km long routes of the North American Breeding Bird Survey. Circular buffers of 19.8 km radius were created to encompass each BBS route. Within these buffers, summary statistics were calculated for each texture measure. Univariate and multivariate analysis was used to evaluate the ability of texture measures to explain variation of species richness in several functional guilds across the Midwestern United States. Measures of texture derived from Landsat imagery were able to explain some of the variation in avian species richness. For Neotropical migrants, the within-buffer mean of mean DN values yielded R2 values of 0.25, 0.25, and 0.22 for Landsat TM bands 1, 2, and 3 respectively. The performance of texture measures of TM band 4 and NDVI performed comparatively poorly. The ability of texture measures to explain species richness was much higher for some guilds, such as Neotropical migrants, than others, such as permanent residents. Texture measures derived from remotely sensed imagery are a useful method for characterizing habitat structure. However, measures of image texture vary in their explanatory power according to the avian guilds and the spectral bands used. We were especially surprised that measures of TM band 4 performed so poorly since this band is closely associated with live vegetation, an important component of avian habitat.
Forest Volume and Biomass estimation from SAR/LIDAR/Optical Fusion in Chile
The paper reports on research to investigate ALOS/PALSAR L-band radar and optical time series data in conjunction with airborne lidar datasets to develop advanced data fusion algorithms for biomass and ecosystem structure measurements in support of the NASA DESDynI mission. The research is based on the acquisition of ALOS/PALSAR time series data beginning in 2007 and the timely confluence of these acquisitions with other highly relevant remote sensing and ground reference data sets in forested areas in Chile. Through collaboration with Digimapas Chile, the project has access to 75,000 km2 of 1-meter resolution full-waveform small footprint lidar (SFPL) data and 0.5 m resolution digital orthophoto imagery covering the commercial forests of Arauco, one of the largest cellulose producers in Latin America. Field inventory data from Arauco are used to test terrain and environmental influences on biomass estimation from empirical regression tree based data fusion approaches. The SAR data acquisitions available from PALSAR during the project time frame will span a five year period from 2007 to 2011, allowing investigations into how L-band time series data, similar to that expected from the DESDynI SAR (backscatter and interferometric coherence), can be used to build (1) the DESDynI biomass map product to be produced at the end of the “designed mission life” (i.e., 3 and/or 5/5+ years) and (2) annual maps of aboveground biomass change.
Lidar modeling with the 3D DART model
A new lidar waveform module is introduced in the optical-thermal model DART (Discrete Anisotropic Radiative Transfer), a 3D radiative transfer model in the optical and thermal domains. It takes advantage of the DART detailed representation of the landscape that can be natural and urban, with the possibility to include atmosphere and topography. It relies on a Monte Carlo Ray tracing approach that was previously used for checking the accuracy of the iterative ray tracing approach of DART. A major point Is that it inherits major features of the DART model; possibility to combine different vegetation species, etc. It permits to calculate the resulting lidar waveform, using 1st order of scattering or including multiple scattering. It is suited for both small and large footprints with an emphasis on large footprint lidar. For validation of the DART lidar model, comparisons took place with the 3D lidar waveform model of Sun and Ranson. Due to the flexibility of the DART model, the two lidar models can share common input parameters. Moreover, they are based on the same physical principles. Then, the potential of the model was further assessed using LVIS data acquired in the northern experimental forest (Maine, USA). Finally, the DART lidar model was used for investigating the importance of multiple scattering within canopy.
Assessing the use of Geoscience Laser Altimeter System data to quantify forest structure change resultant from large-scale forest disturbance events- Case Study Hurricane Katrina
The biodiversity, structure, and functioning of forest systems in most areas are strongly influenced by disturbances. Forest structure can both influence and help indicate forest functions such as the storage and transfer of carbon between the land surface and the atmosphere. A 2007 report published by the National Research Council states that ‘Quantifying changes in the size of the [vegetation biomass] pool, its horizontal distribution, and its vertical structure resulting from natural and human-induced perturbations, such as deforestation and fire, and the recovery processes is critical for measuring ecosystem change.’ This study assessed the use of the Geoscience Laser Altimeter System (GLAS) to detect and quantify changes in forest structure caused by Hurricane Katrina. Data from GLAS campaigns for the year proceeding and following Katrina were compared to wind speed, forest cover, and damage maps to analyze sensor sampling, and forest structure change over a range of spatial scales. Results showed a significant decrease in mean canopy height of 4.0 m in forested areas experiencing category two winds, a 2.2 meter decrease in forests experiencing category one winds, and a 0.6 meter change in forests hit by tropical storm winds. Changes in structure were converted into carbon estimates using the Ecosystem Demography (ED) model to yield above ground carbon storage losses of ~30Tg over the domain. Although the greatest height loss was observed in areas hit by category two winds, these areas only contributed to a fraction (~3Tg) of the estimated above ground carbon storage losses resultant from Katrina, highlighting that small disturbance spread over a large area can account for as much as or more damage than intense disturbance over smaller areas. This finding stresses the importance of detecting and measuring the full extent of storm damage. While results highlighted the potential use of space-born Lidar in damage detection and quantification, they also emphasize limitations on the scope and scale at which current data can quantify hurricane related changes. Season of data acquisition was shown to influence calculations of mean canopy height and change. Limited sampling hindered our ability to make reliable estimates of height change and standing biomass loss at one degree resolution and smaller across the domain. These results have implications for sampling requirements in upcoming missions, such as DESDnyI, that will improve our ability to detect and quantify forest structure changes from disturbance events.
Alternate spatial sampling approaches for ecosystem structure inventory using spaceborne lidar
Current and proposed spaceborne lidar sensors sample the land surface using observations along transects in which consecutive observations in the along-track dimension are either contiguous (e.g. VCL, DESDynI, Livex) or spaced (ICESat, ICESat-2). In contrast, vegetation inventories distribute field observations either in regular grids or within patches that are delineated to represent uniform conditions. In the context of supporting large scale inventories of ecosystem structure, a transect sampling pattern is inefficient because multiple observations are made of a spatially autocorrelated phenomenon while large areas of the landscape are left unsampled. This results in higher uncertainty in estimates of average ecosystem structure across landscapes than would be obtained using sampling in regular grids. As the pulses generated by a spaceborne laser are a valuable resource to be conserved, any strategy that decreases the number of observations required to develop large scale inventories with a given level of confidence should be pursued. Data fusion between lidar data and a spatially complete data source (e.g. polarmetric or interferrometric SAR) will also benefit from a spatially distributed sample of lidar as the average distance between any point and a lidar observation is greatly reduced. This study demonstrates that more flexible spatial arrangements of observations can result in estimates of average landscape height that have as little as one-third of the uncertainty of estimates made with an equal number of observations along transects. The method of sampling described here can be implemented by a technology, Electronically Steerable Flash Lidar, that can distribute observations in the patterns described here and simultaneously support transect sampling.