The timber harvest capability of LANDIS allows flexible simulation of the broad spectrum of silvicultural activities that are commonly implemented on managed forests (Gustafson et al. 2000) . These capabilities are simulated across two distinct hierarchies of disturbance intensity and the spatial configuration. The intensity of management activity ranges from thinning through single-tree selective harvest to clearcutting. The specific details of how these activities affect species and cohort structure are controlled by the user, allowing an almost infinite range of management activity to be simulated. The spatial configuration of management activity is controlled by the designation of Management Areas (MA) in which distinct management activities and intensities can be simulated on the stands within that MA (fig. 7).
LANDIS implements timber harvest within a specific hierarchical management structure. The overall landscape is divided into MAs, each to be treated with specific harvest regimes at specific intensities. The MAs need not be contiguous (i.e., multiple areas having the same MA designation may be delineated). Furthermore, some management units may be specified to have no harvest at all. Harvesting on land in other ownerships can be simulated by representing those ownerships as a distinct MA.
Within MAs that are to be harvested, LANDIS expects to find the land base delineated into stands. These are represented by a map layer in which stand polygons have been gridded so that each site (where site is the equivalent of a cell or a pixel in the raster) has the value of the stand ID number. Often, land in other ownerships will be interspersed among these stands. Any lands that will never have harvests allocated on them can be represented by zeros in the stand map. Based on the map layers of MAs and stands, each site becomes associated with an MA and a stand ID.
Harvests are implemented by removing specific cohorts of specific species on sites selected for harvest. The sites selected for harvest are determined by one of eight “harvest regimes” from which the user may choose. These harvest regimes vary in the number of entries required to complete the silvicultural treatment, and in whether they are applied to 1) an entire stand (stand-filling), 2) a portion of a stand or to multiple stands (stand-spreading), and 3) multiple patches within a single stand (i.e., group selection). The regimes currently available are 1) one-entry, stand-filling, 2) periodic-entry, stand-resampling * , stand-filling, 3) two-entry, stand-filling, 4) one-entry, stand-spreading, 5) two-entry, stand-spreading, 6) periodic-entry, group selection, 7) periodic-entry-fixed stand, two-entry, stand-filling, and 8) periodic-entry-stand-resampling, two-entry, stand-filling. Stand resampling means that in each entry year, the stands within the management area will be ranked again using the initial ranking algorithm. Thus, the specific stands treated in each entry may vary, but the treatment applied will not.
These regimes allow simulation of multiple-entry silvicultural treatments such as shelterwood, seed-tree, and group selection. The periodic entry options allow for automatic harvesting of stands at some specified interval (e.g., rotation length). Stand-filling regimes are applied to every site in a single stand, while stand-spreading regimes begin at a focal site in a stand; harvest terminates when a certain harvest size is reached. This size may be reached before the stand is completely harvested, or the harvest may spill over into an adjacent stand. All eligible sites within a stand must be harvested before harvest can spill over into another stand, but the process may continue until the size has been reached. This feature allows timber management to be used to change the patch size distribution of the landscape, and to allow patterns to emerge that are less constrained by the underlying stand map than are the stand-filling harvest regimes.
Stands for the application of these regimes are selected using a ranking algorithm chosen by the user. Stands within each MA are prioritized for harvest (ranked) according to rules that reflect criteria that might be used by forest planners or researchers. The ranking algorithms currently include Stand Age (oldest first), Economic Importance (most valuable stands first), Regulate Age Class Distribution (attempt to produce an even distribution), and Random (choose stands randomly). Because stands are ranked independently for each MA, different ranking algorithms can be applied to different MAs. When an adjacent stand needs to be harvested by a stand-spreading regime, the highest ranked of all the adjacent stands is selected.
Variation in the intensity of harvest activity is controlled by rules governing the removal of age cohorts of species found on the site being harvested. The rules for each harvest regime are specified by the user in the form of removal masks that for each species specify which age cohorts (if present on the site) are to be removed. For example, the prescribed burning mask might specify that the youngest cohort of all species be removed. A shelterwood mask might specify that all cohorts be removed for all species except one or two older cohorts of one or two species during the first entry, and all older cohorts during the second entry. Because the removal masks are generated by the user, the user has unlimited flexibility in tailoring the masks to the treatments to be simulated and the characteristics of the species found on the landscape.
Figure 4.- Flow chart of the LANDIS harvest module show harvest actions with one LANDIS iteration.
The extent of harvesting is controlled by the user. For each MA the user must specify the regimes to be applied, the total area to be harvested under each regime, and the ranking algorithm to be used for each regime. If a stand-spreading regime is chosen, the user must specify the size distribution for the harvests. Simulations are controlled by a parameter input file. Once this file is produced, it can be modified to introduce changes to the scenario simulated, allowing alternatives and replicates to be readily generated.
The Harvest module completes two major loops to simulate one time period of timber harvest: 1) It sequentially visits each MA in which timber harvest is prescribed, and 2) within each MA, it sequentially simulates each harvest regime specified for that MA, beginning with scheduled re-entries. Different harvesting rules can be specified for each MA. For each harvest regime, the module allocates harvests to stands until the target number of sites has been allocated. The order in which stands are chosen is determined by the ranking procedure. At the completion of the model run, LANDIS produces a log file allowing comparison of the harvest targets with the allocations actually made. LANDIS also creates output files that report how many sites containing each cohort (by species) were cut (useful to estimate volume and value generated by harvest activity), and how many sites containing cohorts (by species) remain on each stand (useful to estimate stand vertical structure).