In the past quarter, the DMS Lab has updated the RFID system to incorporate a database within the remote operating system. In conjunction with the portable ID scanner, a PocketPC Windows CE device is interfaced to the RFID system to provide a means to catalog the ID numbers at the point of acquisition. The ActiveX data object (ADO) database provides a means to track on a tree-by-tree basis and produce detailed distribution reports using Microsoft Access or similar programs.

-Sean Hoyt

Forest Structure Analysis from airborne LIDAR data

Research objectives and status of project: The management of the forest resource for multiple objectives, including timber, wildlife habitat, watershed protection, and recreation, requires detailed information relating to forest structure characteristics. Field-based techniques for collection of forest structure information are often expensive and time-intensive. There is potential for significant cost reduction in the collection of forest structure information through the use of emerging active remote sensing technologies, including LIDAR (Light Detection And Ranging). An airborne LIDAR system emits several thousand laser pulses per second, allowing measurement of forest canopy, understory, and the underlying terrain surface.

An ongoing research project funded by the Department of Defense, USDA Forest Service, and the University of Washington Precision Forestry Cooperative is investigating the use of LIDAR for extraction of forest structure information across areas within Fort Lewis Military Reservation, WA and Capitol State Forest, WA. LIDAR data was acquired over 48 km² within Fort Lewis and 5.2 km² within Capitol Forest. Algorithms have been developed to extract information relating to forest canopy coverage, as well as the vertical distribution (i.e. structure) of forest vegetation, at a specified spatial resolution, over areas characterized by varying forest type and structure, including oak prairie, mature Douglas-fir, and mature alder stands. The output of these algorithms is compared to extensive field data collected within both Fort Lewis and Capitol Forest.

There is also potential for the use of high-density LIDAR data for spatially explicit forest structure analysis. Helicopter-borne LIDAR sensors can literally "paint" the forest, collecting up to 4 laser measurements per square meter, which allows for extraction of detailed three-dimensional geometric information relating to the dimensions of individual tree crowns. Algorithms have been developed to carry out inference on the dimensions of individual trees, given the distribution of LIDAR measurements collected over the forest. The output of these algorithms is compared to extensive, spatially-referenced field measurements collected at Capitol Forest.

- Hans-Erik Andersen

In the early part of the fall, time was spent looking at some of the major products in landscape visualization and 3-D modeling. These products included: Envision from the USDA Forest Service, Terragen from Planetside, Bryce from Corel, and World Construction Set from 3D Nature. Time was also spent reading on Urban Landscaping issues in preparation for work on the Tiger Mountain State Park visualization.

Specific to the Tiger Mountain Visualization project, all digital data sets, including Ortho photography, have been obtained, as well as hard copies of the planned units in the project area, and the Habitat Conservation plan. For this project, EnVision will be used.

- Matthew Walsh

The forest harvest scheduling problem often involves green-up requirements. As a result, the adjacency problem, which requires integer programming (IP) must be solved. A local optimization method, employing 1-opt and 2-opt moves was developed to schedule harvest units over a variable time horizon. The method allows for a variable number of stand types and time periods to be considered in the same schedule. The methodology applied is that of a hierarchical model, so that the adjacency problem is treated as part of a tactical problem. Volume or economic targets are specified by a strategic problem and passed to the adjacency solver. First the adjacency solver approximates the solution by solving the relaxed problem. This solution is used to construct a starting point for the discrete solution. From this starting point, using the one opt two opt method, the adjacency solver minimizes the deviation between the targets set by the strategic problem and that attainable in the tactical problem with adjacency constraints active. It does this for all stand types and time periods simultaneously.

The Pack forest landscape, which is composed of 183 stands over roughly 4500 acres, was used to test the solver. Seven ten year time periods were considered, with 6 different stand types delimited by species. Only 158 of the 183 stands were considered for timber production. These stands were simulated using the Forest Vegetation Simulator, giving yield data for the life of the planning horizon. A mock strategic plan was simulated by solving the relaxed LP associated with the adjacency problem, with volume flow being constrained to not deviate more than 5% from the volume harvested in the first period and with the objective of maximizing NPV with an interest rate of 5%. The solver converged to a feasible solution in 26.5 s which was within .8% of the upper bound, specified by the relaxed problem.

- Sam Pittman