Eric Gustafson and Luke Rasmussen
HARVEST was designed as a strategic research and planning tool, allowing assessment of the spatial pattern consequences of broad timber management strategies. The model is well suited to evaluate alternative strategies, providing comparable predictions about how the alternatives affect the age (or successional stage) distribution and forest type composition of the forest, the spatial distribution of forest interior and edge habitats, and the patch structure of the resulting forest landscape. With HARVEST, the object is not to find a scheduling solution (i.e., determining the order in which individual stands should be harvested), but to assess the spatial pattern consequences of general management strategies.
HARVEST 6.1 was designed with the user in mind and features a graphical user interface that makes it easier for the user to interact with the program. HARVEST 6.1 provides greater control of the effects of cutting then earlier versions. For example, the user can control the age of cells after they have been cut, and also has the ability to convert the forest type of the cell. HARVEST 6.1 also has analysis capabilities not found in earlier versions.
New Features in Version 6.1: HARVEST parameter menu
- The user can specify any combination of two treatment effects when a harvest is implemented, 1) the age to which the treated cells are set, including reducing the current cell age by a constant and 2) whether the forest type will be converted upon cutting (e.g., planting, different type of the residual stand, deterministic succession)
- Display the Age, Forest Type and the Management Area maps at any time
- The user can require harvest units to completely fill stand boundaries, even when a target cutblock size is specified
- The user can specify a maximum age, above which stands cannot be cut
- Save the Forest Type Map, since it now can change
- The menus of version 6.1 have been re-designed
- The user can conduct analyses of the spatial pattern of the landscape both before and after simulated harvest using the spatial pattern analysis software, APACK
- Version 6.1 adds calculation of the fragmentation index GISfrag (Ripple et al. 1991)
- Execution speed has been increased
About HARVEST Lite
Harvest Lite is an educational version of HARVEST, a timber harvest simulation model designed to allow comparison of alternative forest management strategies. Harvest Lite allows control of the three major model parameters that most affect the spatial pattern of managed forested landscapes. Harvest Lite simulates 8 decades of forest management activity during each model run, and allows users to quantify the spatial pattern of the resulting landscape.
FORTRAN Source code
Because of the requirements to compile the code, source code can only be offered by special requests. Please email the author for more information, or to make a source code request.
Documentation and Downloads for HARVEST and HARVEST Lite
Please fill out the optional registration below. Registered users will receive bug fix notifications and news of new releases.
If you prefer not to register, you may continue to the download section.
|HARVEST 6.1 Software (2.2 mb, self-extracting zip file)||exe|
|Two Sample Map Sets for HARVEST 6.1||exe|
|HARVEST 6.1 User's Guide|
|HARVEST 6.1 Script Format|
|Pleasant Run sample data set from Hoosier NF for HARVEST 6.1||zip|
|HARVEST Lite executable software (1 mb)||exe|
|HARVEST Lite Documentation|
|Some of these documents are in PDF format. You can obtain a free PDF reader from Adobe.|
Publications using HARVEST
Bug Fixes to HARVEST v6.1
Problem: Stands were not being re-typed after converting cells to a new forest type when the revisit interval = zero (i.e., when the ‘Reenter stands' check box was not checked). This was a problem when more than half the stand was converted. In such a case the stand should be classified as the new forest type. When the revisit interval was not equal to zero, it did re-type the stands correctly.
Fix: Stands are now always re-typed after conversion.
Problem: The random number generator. The distribution of uniform deviates was biased about 2% low (mean approximately equal to 0.48) because a typographical error in a denominator prevented the largest values (0.99-1.0) from being generated. This bias affected HARVEST runs in several ways.
- The very largest and very smallest harvest sizes in the normal size distribution were slightly less likely to occur. Because of the way random normal deviates are calculated, and the way harvest size is calculated, this bias is deemed to be very minor. When a maximum or minimum harvest size is specified that is within 3 standard deviations of the mean harvest size, this bias has no effect.
- In HARVEST, uniform random deviates are used to select stands or cells from a stack (list of stands or cells). These stacks often contain all stands eligible for harvest (i.e., the correct MA, forest type, and old enough), or all cells eligible for cutting within a stand. The random deviates are used to select the next cell or stand for cutting. When the number of cells or stands in the stack was less than about 50, there was no bias. As the number exceeded 50, the bias prevented about 2% of the items at the end of the stack from being selected. The bias affects the selection of items when the stack is large, but as cells or stands are cut, the bias disappears when the number of uncut items drops below about 50. Because cells and stands typically enter these stacks in an order determined by their spatial location on the grid (from left to right and from top to bottom), cells and stands toward the lower right tended to be the ones that were not selected when stack size >50. This bias mostly affected the order in which items were cut (items near the end of the stack were cut disproportionately later in the sequence) since it did not prevent items from being harvested.
We have compared some model runs with the biased and an unbiased random number generator, and have found little difference in landscape pattern.
Fix: The denominator in the random number generator was corrected and there is now no bias in random deviates.
Last Modified: 03/30/2015