Search

Model-assisted, two-stage forest survey sampling designs provide a means to combine airborne remote sensing data, collected in a sampling mode, with field plot data to increase the precision of national forest inventory estimates, while maintaining important properties of design-based inventories, such as unbiased estimation and quantification of uncertainty. In this study, we present a comprehensive set of model-assisted estimators for domain-level attributes in a two-stage sampling design, including new estimators for densities, and compare the performance of these estimators with standard poststratified estimators. Simulation was used to assess the statistical properties (bias, variability) of these estimators, with both simple random and systematic sampling configurations, and indicated that 1) all estimators were generally unbiased. and 2) the use of lidar in a sampling mode increased the precision of the estimators at all assessed field sampling intensities, with particularly marked increases in precision at lower field sampling intensities. Variance estimators are generally unbiased for model-assisted estimators without poststratification, while model-assisted estimators with poststratification were increasingly biased as field sampling intensity decreased. In general, these results indicate that airborne remote sensing, collected in a sampling mode, can be used to increase the efficiency of national forest inventories.

A two-stage hierarchical Bayesian model is developed and implemented to estimate forest biomass density and total given sparsely sampled LiDAR and georeferenced forest inventory plot measurements. The model is motivated by the United States Department of Agriculture (USDA) Forest Service Forest Inventory and Analysis (FIA) objective to provide biomass estimates for the remote Tanana Inventory Unit (TIU) in interior Alaska. The proposed model yields stratum-level biomass estimates for arbitrarily sized areas. Model-based estimates are compared with the TIU FIA design-based post-stratified estimates. Model-based small area estimates (SAEs) for two experimental forests within the TIU are compared with each forest's design-based estimates generated using a dense network of independent inventory plots. Model parameter estimates and biomass predictions are informed using FIA plot measurements, LiDAR data that are spatially aligned with a subset of the FIA plots, and complete coverage remotely detected data used to define landuse/landcover stratum and percent forest canopy cover. Results support a model-based approach to estimating forest variables when inventory data are sparse or resources limit collection of enough data to achieve desired accuracy and precision using design-based methods.

Geolocation error in spaceborne sampling light detection and ranging (LiDAR) measurements of forest structure can compromise forest attribute estimates and degrade integration with georeferenced field measurements or other remotely sensed data. Data integration is especially problematic when geolocation error is not well quantified. We propose a general model that uses airborne laser scanning (ALS) data to quantify and correct geolocation error in spaceborne sampling LiDAR. To illustrate the model, LiDAR data from NASA Goddard's LiDAR Hyperspectral & Thermal Imager (G-LiHT) was used with a subset of LiDAR data from NASA's Global Ecosystem Dynamics Investigation (GEDI). The model accommodates multiple canopy height metrics derived from a simulated GEDI footprint kernel using spatially coincident G-LiHT, and incorporates both additive and multiplicative mapping between the canopy height metrics generated from both datasets. A Bayesian implementation provides probabilistic uncertainty quantification in both parameter and geolocation error estimates. Results show a systematic geolocation error of 9.62 m in the southwest direction. In addition, estimated geolocation errors within GEDI footprints were highly variable, with results showing a ~0.45 probability the true footprint center is within 20 m. Estimating and correcting geolocation error via the model outlined here can help inform subsequent efforts to integrate spaceborne LiDAR data, like GEDI, with other georeferenced data.

A key challenge in spatial statistics is the analysis for massive spatially-referenced data sets. Such analyses often proceed from Gaussian process specifications that can produce rich and robust inference, but involve dense covariance matrices that lack computationally exploitable structures. The matrix computations required for fitting such models involve floating point operations in cubic order of the number of spatial locations and dynamic memory storage in quadratic order. Recent developments in spatial statistics offer a variety of massively scalable approaches. Bayesian inference and hierarchical models, in particular, have gained popularity due to their richness and flexibility in accommodating spatial processes. Our current contribution is to provide computationally efficient exact algorithms for spatial interpolation of massive data sets using scalable spatial processes. We combine low-rank Gaussian processes with efficient sparse approximations. Following recent work by [1], we model the low-rank process using a Gaussian predictive process (GPP) and the residual process as a sparsity-inducing nearest-neighbor Gaussian process (NNGP). A key contribution here is to implement these models using exact conjugate Bayesian modeling to avoid expensive iterative algorithms. Through the simulation studies, we evaluate performance of the proposed approach and the robustness of our models, especially for long range prediction. We implement our approaches for remotely sensed light detection and ranging (LiDAR) data collected over the US Forest Service Tanana Inventory Unit (TIU) in a remote portion of Interior Alaska.

Gathering information about forest variables is an expensive and arduous activity. As such, directly collecting the data required to produce high-resolution maps over large spatial domains is infeasible. Next generation collection initiatives of remotely sensed Light Detection and Ranging (LiDAR) data are specifically aimed at producing complete-coverage maps over large spatial domains. Given that LiDAR data and forest characteristics are often strongly correlated, it is possible to make use of the former to model, predict, and map forest variables over regions of interest. This entails dealing with the high-dimensional ($\sim$$10^2$) spatially dependent LiDAR outcomes over a large number of locations (~10^5-10^6). With this in mind, we develop the Spatial Factor Nearest Neighbor Gaussian Process (SF-NNGP) model, and embed it in a two-stage approach that connects the spatial structure found in LiDAR signals with forest variables. We provide a simulation experiment that demonstrates inferential and predictive performance of the SF-NNGP, and use the two-stage modeling strategy to generate complete-coverage maps of forest variables with associated uncertainty over a large region of boreal forests in interior Alaska.

The goal of this research was to develop and examine the performance of a geostatistical coregionalization modeling approach for combining field inventory measurements, strip samples of airborne lidar and Landsat-based remote sensing data products to predict aboveground biomass (AGB) in interior Alaska's Tanana Valley. The proposed modeling strategy facilitates pixel-level mapping of AGB density predictions across the entire spatial domain. Additionally, the coregionalization framework allows for statistically sound estimation of total AGB for arbitrary areal units within the study area---a key advance to support diverse management objectives in interior Alaska. This research focuses on appropriate characterization of prediction uncertainty in the form of posterior predictive coverage intervals and standard deviations. Using the framework detailed here, it is possible to quantify estimation uncertainty for any spatial extent, ranging from pixel-level predictions of AGB density to estimates of AGB stocks for the full domain. The lidar-informed coregionalization models consistently outperformed their counterpart lidar-free models in terms of point-level predictive performance and total AGB precision. Additionally, the inclusion of Landsat-derived forest cover as a covariate further improved estimation precision in regions with lower lidar sampling intensity. Our findings also demonstrate that model-based approaches that do not explicitly account for residual spatial dependence can grossly underestimate uncertainty, resulting in falsely precise estimates of AGB. On the other hand, in a geostatistical setting, residual spatial structure can be modeled within a Bayesian hierarchical framework to obtain statistically defensible assessments of uncertainty for AGB estimates.

Recent advancements in remote sensing technology, specifically Light Detection and Ranging (LiDAR) sensors, provide the data needed to quantify forest characteristics at a fine spatial resolution over large geographic domains. From an inferential standpoint, there is interest in prediction and interpolation of the often sparsely sampled and spatially misaligned LiDAR signals and forest variables. We propose a fully process-based Bayesian hierarchical model for above ground biomass (AGB) and LiDAR signals. The process-based framework offers richness in inferential capabilities, e.g., inference on the entire underlying processes instead of estimates only at pre-specified points. Key challenges we obviate include misalignment between the AGB observations and LiDAR signals and the high-dimensionality in the model emerging from LiDAR signals in conjunction with the large number of spatial locations. We offer simulation experiments to evaluate our proposed models and also apply them to a challenging dataset comprising LiDAR and spatially coinciding forest inventory variables collected on the Penobscot Experimental Forest (PEF), Maine. Our key substantive contributions include AGB data products with associated measures of uncertainty for the PEF and, more broadly, a methodology that should find use in a variety of current and upcoming forest variable mapping efforts using sparsely sampled remotely sensed high-dimensional data.

Many forest management planning decisions are based on information about the number of trees by species and diameter per unit area. This information is commonly summarized in a stand table, where a stand is defined as a group of forest trees of sufficiently uniform species composition, age, condition, or productivity to be considered a homogeneous unit for planning purposes. Typically information used to construct stand tables is gleaned from observed subsets of the forest selected using a probability-based sampling design. Such sampling campaigns are expensive and hence only a small number of sample units are typically observed. This data paucity means that stand tables can only be estimated for relatively large areal units. Contemporary forest management planning and spatially explicit ecosystem models require stand table input at higher spatial resolution than can be affordably provided using traditional approaches. We propose a dynamic multivariate Poisson spatial regression model that accommodates both spatial correlation between observed diameter distributions and also correlation between tree counts across diameter classes within each location. To improve fit and prediction at unobserved locations, diameter specific intensities can be estimated using auxiliary data such as management history or remotely sensed information. The proposed model is used to analyze a diverse forest inventory dataset collected on the United States Forest Service Penobscot Experimental Forest in Bradley, Maine. Results demonstrate that explicitly modeling the residual spatial structure via a multivariate Gaussian process and incorporating information about forest structure from LiDAR covariates improve model fit and can provide high spatial resolution stand table maps with associated estimates of uncertainty.

The vast extent and inaccessibility of boreal forest ecosystems are barriers to routine monitoring of forest structure and composition. In this research, we bridge the scale gap between intensive but sparse plot measurements and extensive remote sensing studies by collecting forest inventory variables at the plot scale using an unmanned aerial vehicle (UAV) and a structure from motion (SfM) approach. At 20 Forest Inventory and Analysis (FIA) subplots in interior Alaska, we acquired overlapping imagery and generated dense, 3D, RGB (red, green, blue) point clouds. We used these data to model forest type at the individual crown scale as well as subplot-scale tree density (TD), basal area (BA), and aboveground biomass (AGB). We achieved 85% cross-validation accuracy for five species at the crown level. Classification accuracy was maximized using three variables representing crown height, form, and color. Consistent with previous UAV-based studies, SfM point cloud data generated robust models of TD (r(sup 2) = 0.91), BA (r(sup 2) = 0.79), and AGB (r(sup 2) = 0.92), using a mix of plot- and crown-scale information. Precise estimation of TD required either segment counts or species information to differentiate black spruce from mixed white spruce plots. The accuracy of species-specific estimates of TD, BA, and AGB at the plot scale was somewhat variable, ranging from accurate estimates of black spruce TD (+/−1%) and aspen BA (−2%) to misallocation of aspen AGB (+118%) and white spruce AGB (−50%). These results convey the potential utility of SfM data for forest type discrimination in FIA plots and the remaining challenges to develop classification approaches for species-specific estimates at the plot scale that are more robust to segmentation error.

The first European Space Agency (ESA) and NASA collaboration in an airborne campaign to support ESA's FLuorescence EXplorer (FLEX) mission was conducted in North Carolina, USA during September-October 2013 (FLEX-US 2013) at the Parker Tract Loblolly Pine (LP) Plantation (Plymouth, NC, USA). This campaign combined two unique airborne instrument packages to obtain simultaneous observations of solar-induced fluorescence (SIF), LiDAR-based canopy structural information, visible through shortwave infrared (VSWIR) reflectance spectra, and surface temperature, to advance vegetation studies of carbon cycle dynamics and ecosystem health. We obtained statistically significant results for fluorescence, canopy temperature, and tower fluxes from data collected at four times of day over two consecutive autumn days across an age class chronosequence. Both the red fluorescence (F685) and far-red fluorescence (F740) radiances had highest values at mid-day, but their fluorescence yields exhibited different diurnal responses across LP age classes. The diurnal trends for F685 varied with forest canopy temperature difference (canopy minus air), having a stronger daily amplitude change for young vs. old canopies. The Photochemical Reflectance Index (PRI) was positively correlated with this temperature variable over the diurnal cycle. Tower measurements from mature loblolly stand showed the red/far-red fluorescence ratio was linearly related to canopy light use efficiency (LUE) over the diurnal cycle, but performed even better for the combined morning/afternoon (without midday) observations. This study demonstrates the importance of diurnal observations for interpretation of fluorescence dynamics, the need for red fluorescence to understand canopy physiological processes, and the benefits of combining fluorescence, reflectance, and structure information to clarify canopy function versus structure characteristics for a coniferous forest.

Fire in the boreal region is the dominant agent of forest disturbance with direct impacts on ecosystem structure, carbon cycling, and global climate. Global and biome-scale impacts are mediated by burn severity, measured as loss of forest canopy and consumption of the soil organic layer. To date, knowledge of the spatial variability in burn severity has been limited by sparse field sampling and moderate resolution satellite data. Here, we used pre- and post-fire airborne lidar data to directly estimate changes in canopy vertical structure and surface elevation for a 2005 boreal forest fire on Alaskas Kenai Peninsula. We found that both canopy and surface losses were strongly linked to pre-fire species composition and exhibited important fine-scale spatial variability at sub-30m resolution. The fractional reduction in canopy volume ranged from 0.61 in lowland black spruce stands to 0.27 in mixed white spruce and broad leaf forest. Residual structure largely reflects standing dead trees, highlighting the influence of pre-fire forest structure on delayed carbon losses from above ground biomass, post-fire albedo, and variability in understory light environments. Median loss of surface elevation was highest in lowland black spruce stands (0.18 m) but much lower in mixed stands (0.02 m), consistent with differences in pre-fire organic layer accumulation. Spatially continuous depth-of-burn estimates from repeat lidar measurements provide novel information to constrain carbon emissions from the surface organic layer and may inform related research on post-fire successional trajectories. Spectral measures of burn severity from Landsat were correlated with canopy (r = 0.76) and surface (r = -0.71) removal in black spruce stands but captured less of the spatial variability in fire effects for mixed stands (canopy r = 0.56, surface r = -0.26), underscoring the difficulty in capturing fire effects in heterogeneous boreal forest landscapes using proxy measures of burn severity from Landsat.