Olympic Habitat Development Study
US FOREST SERVICE

 

Station 5 – USING MODELS TO PREDICT THE FUTURE

Forest managers often want to determine the long-term effects of alternative scenarios on tree growth and mortality.  This allows them to determine which alternative treatment (including no action) would best meet their objectives.  They do this with models which use current conditions,(numbers of trees, tree sizes and species) information on site productivity and treatment specifics to predict how trees will grow and how many will die.  Available models which predict tree growth for the Pacific coast were developed primarily based on data from unmanaged stands or stands which were treated in a uniform way (e.g., thinned to the same density in all areas. The two main publically available models used in Washington and Oregon are FVS(Forest Vegetation Simulator) and ORGANON. (Links to these models are under Resources).

To use existing models for stands which have been treated in a non-uniform manner (such as variable density thinning), there are 2 main choices: (it would also be possible to use more complicated spatial models but few people have the data to use those types of models)

  1. Use an average of the conditions in the whole area, or
  2. Predict growth in each uniform area (e.g., skips separately from the thinned matrix) and then summarize the results based on an average weighted by the area each component occupies.

We compared growth rates based on our measured trees to predicted values from both modeling approaches using both models. We did this for large trees (>8 cm or >3.0 inches) and also small trees. (<8 cm or <3.0 inches)

We tagged and measured trees and determined their locations (graph).Location of trees in Fresca research area. Location of trees in Rail research area.
These graphs show tree locations at Fresca and Rail and the boundaries of the gaps (small squares) and skips (larger rectangles). Small trees are shown as black dots and larger trees as grey circles (circle size is proportional to tree diameter).
When we compared 10-year growth of the larger trees that we measured at Fresca to growth that the models predicted, the predictions from both models were reasonably close to the measured values (graph).Comparison of actual 10-year diameter growth and growth predicted by 2 growth models for trees > 8 cm in diameter.   Both models predicted smaller differences between growth in the skip areas and growth in the thinned matrix than we measured.  ORGANON tended to underestimate growth in this thinned stand, but since the ORGANON model did not include Sitka spruce when it was developed, it is understandable that it might not be calibrated for this type of stand.

When we looked at growth of the small trees (graph)Comparison of actual 10-year diameter growth and growth predicted by 2 growth models for trees < 8 cm in diameter. we had a much smaller data set than we had for the larger trees so we combined data from Fresca with data from Rail, another OHDS site nearby. We can see that:

  1. Both models under-predicted growth of small trees in skips and the thinned matrix and the under-prediction in FVS is particularly striking.
  2. Both models over-predicted growth in and surrounding the gaps.  This makes sense as the models would assume a very low density group of trees since most trees in the gaps were cut;(except for trees <15cm or <6 inches, hardwoods, or species we wanted to favor) however, the trees in the small gaps are actually quite influenced by the large trees surrounding the gaps (shading them and through below-ground competition.

Stand-level growth models(models which predict growth for groups of trees by tree size and species) may do a reasonably good job of predicting average growth for large trees.  However, they are not well suited for use in stands with variable treatments (such as variable-density thinning) because:

  1. They under-predict the variability in growth (and thus the range in future tree size).
  2. They under-predict growth of small trees (thus, it would be almost impossible for them to predict development of multi-layered stands as the small trees would not grow out of the small size category for many years).

 

Models allow us to determine which is the most appropriate management for a stand.
It is useful to predict growth and mortality in alternative scenarios. Models allow us to determine which is the most appropriate management for a stand. Non-spatial models will not predict tree growth near a gap.
Models allow us to determine which is the most appropriate management for a stand.