The Vegetation Monitoring and Remote Sensing team develops new techniques for using remote sensing and other geospatial technologies within large-scale, multiobjective resource inventory systems.
The primary research goal is the development and application of inventory and monitoring techniques that largely depend on remotely acquired information, which must be integrated with field-based information in innovative and complex statistical sampling designs.
Resource Monitoring and Assessment researchers also investigate how emerging geospatial/remote sensing technologies and analytical techniques can best be applied to assess the extent, condition, use, and trends of forested ecosystems and natural resources.
The southwest Oregon project involves mapping and estimating forest change in southwest Oregon with integrated repeat plot and satellite measurements and comparisons of projections with single-time LiDAR-based forest structure estimates developed using an independent ground plot dataset.
Field measurements of carbon pools from forest inventories are crucial for national carbon accounting, but the temporal dispersion of measurements creates difficulties for estimating carbon flux and the impact of land use, management, and disturbance for discrete points in time. Our current inability to quantify the causes of changes in inventoried forest carbon is a major stumbling block in national greenhouse gas accounting. In addition, older inventories are incomplete, missing data from some forest lands and most non-tree carbon pools. The rich dataset of repeated field measurements at thousands of plots can be used to develop accurate models using historical trends in satellite spectra, and form the basis for making accurate plot-level predictions of carbon flux and pools.
LiDAR data have been successfully used to estimate above ground biomass/carbon and tree cover in numerous recent studies. With the increasing availability of LiDAR data, its usefulness in multi-level sampling approaches that incorporate Forest Inventory and Analysis (FIA) plot and satellite data has not been fully explored. A unique combination of datasets is available for southwest Oregon, including a rich library of satellite imagery, re-measurements of FIA plots, and independent ground plots measured solely for LiDAR model development and validation. These datasets provide the opportunity to answer a variety of questions that have not been rigorously evaluated over a geographic area of this magnitude.
Simulated tree stands with precisely known horizontal and vertical structure, and canopy cover in various topographic conditions are constructed in a three-dimensional voxel space. Ray tracing connecting viewing locations, similar to those of an airborne camera, and selected within-stand ground locations are used to determine canopy cover in an oblique setting.
The study quantifies perceived canopy cover as a function of topographic and vegetation conditions. Results suggest that view angles can dramatically affect perceived canopy cover, especially n the presence of steep slopes and medium canopy cover. In such conditions, assessment of canopy cover via interpretation of short focal length aerial photographs would lead to biased estimates.
Past attempts to improve FIA plot georeferencing by using LiDAR data have not been particularly successful. A revised methodology and algorithmic enhancements have improved success rates, even for inventory plots in the clustered FIA design. Most plots in stands with heterogeneous canopies and numerous small openings can now be georeferenced with sub-meter precision using exclusively high-density LiDAR data and information recorded during FIA field visits. In many cases, precise plot coordinates can be retrieved with search radii larger than the customary maximum plot location fuzzing distance employed by FIA to protect location confidentiality.
Standard approaches that use LiDAR data to assess canopy cover are imprecise and often biased. Using detailed field measurements of canopy cover and terrestrial LiDAR data sets, the bias is quantified. A novel, unbiased method that dynamically adjusts to variability in local laser data density is introduced. The new method is computationally efficient and independent of vegetation and topographic conditions.
LiDAR return clouds are processed to generate object representations in voxel space and ray tracing algorithms are then used to determine whether the trajectory that originates from the sun terminates onto an object, thereby classifying all scene objects or object parts as sunlit or in shadow. Because the azimuth and elevation angle of the sun for any time and location on Earth are known, it is now possible to assess the lighting regime across the landscape either for a chosen moment or dynamically across time. This type of research has potentially numerous applications, including forest inventory operations, precise estimation of apparent canopy cover from off-nadir viewing locations, habitat suitability analyses, tree regeneration potential, and location-specific stream shading and temperature regimes. It can also be used in support of management and decision making at both tactical and strategic levels.
PNW-RMA researchers are comparing plot-level and LiDAR data to assess and correct tree density estimates from remotely sensed forest characteristics.
Working with the Remote Sensing Applications Center (RSAC) to generate remote sensing-based forest stand information (height/cover class by primary/secondary/tertiary species class, disturbance class) over about 300 FIA plots on the western Kenai Peninsula, Alaska. Evaluate utility of photo plots as a component of inventory design in remote regions.
In this analysis, researchers are collecting photo re-measurements over photo and field plots established in the AIRIS inventory of the Tanana Valley (early 1980s). They are also collecting field re-measurement over a limited selection of field plots in the vicinity of Fairbanks, and comparing field and plot re-measurements to changes observed through analysis in Landsat time series via LandSync.
This project involves developing methods to estimate aboveground carbon/biomass in boreal forests using L-band satellite airborne LiDAR to inform the analysis of L-band PolSAR.
Researchers are analyzing 5-year re-measurement of airborne LiDAR strip data acquired on the western lowlands of the Kenai Peninsula. They are comparing estimates of change observed in LiDAR to statistical estimates obtained from FIA plot data, especially in burned areas.
This project involves evaluating the utility of airborne LiDAR data as a monitoring tool in Amazonian tropical forests. We are quantifying aboveground forest carbon and estimating quantity removed by logging operations in 1 year using multi-temporal LiDAR. We are also identifying impacted areas (tree removals, skid trails, landing, roads, etc.) using 3-D LiDAR forest structure remeasurements.