Steps Towards Better Forest Structure Predictions
White-tailed deer (Odocoileus virginianus) spotted in a Forest Inventory and Analysis (FIA) plot. Damage from white-tailed deer can disproportionately damage or kill smaller trees.
Forest structure, which is forest vegetation’s arrangement, size, and composition, influences the products and services forests provide, including timber production, wildlife habitat, nutrient cycling, and carbon sequestration. Tree size and abundance are critical components of forest structure. The relationships between tree size and abundance, and how these relationships change following disturbance can influence silvicultural decisions such as stocking targets. Measuring tree size and abundance in the field, however, can be time consuming and costly.
To better understand changes in tree size-abundance relationships following disturbance, and to make it easier to estimate forest structure using remote sensing, Jonathan Knott, a Forest Service research forester, recently collaborated on research led by the University of Maine. The research team leveraged long-term forest monitoring networks in two studies, one to better understand the effects of disturbance on tree size-abundance relationships and one to develop a way to accurately estimate tree size distribution using readily available remotely sensed data. Together, these studies apply ecological theory to practical use, and their findings take steps towards more accurate predictions of forest structure across time and space.
Disturbance
Natural disturbances, such as fires, animal damage, and storms, can change relationships between tree size, specifically the diameter distribution of trees, and abundance in a forest. While disturbances themselves can be unexpected, this team found that changes in tree size-abundance relationships following several types of common disturbances were predictable.
Using 160,000 long-term forest monitoring plots from the Nationwide Forest Inventory, which is run by the Forest Service Forest Inventory and Analysis program, the research team measured tree size-abundance relationships(e.g., the distribution of trees in different size classes) using the most recent inventory data from each plot. They then grouped these plots by the most recent type of disturbance experienced by the plot using the following categories:
Ground fire damage, as shown here following the Dixie Fire on the Lassen National Forest in California, can disproportionately damage small trees compared to large trees.
- No disturbance
- Insect damage
- Human-caused damage (excluding silviculture and logging)
- Disease damage
- Fire (including both crown and ground fire)
- Ground fire only
- Animal damage or grazing
- Wind
Of the seven types of disturbances, animal damage and ground fires frequently killed more small trees relative to large trees when compared to undisturbed plots. Consistent changes in tree size-abundance relationships following common disturbances suggest that changes in forest structure can be predicted, which can inform management in areas prone to these disturbances.
Remote Sensing
Across large landscapes, it may be impractical to measure tree size and abundance in the field, but it may be more feasible with remote sensing. This research team developed a method to use remotely sensed data, including high resolution aerial imagery, to determine tree size distribution. The method was validated using monitoring plots from two other long-term forest monitoring networks with plots in the United States: National Ecological Observatory Network (NEON) and Forest Global Earth Observatory (ForestGEO).
A National Ecological Observatory Network (NEON) Twin Otter plane flies over a field site to capture aerial imagery.
The first step in determining tree size distribution was to retrieve remotely sensed data. These data can be collected using crewed or uncrewed aerial systems (UAS) or, depending on the location, may be available from an online database or from a third-party data collection service. Next, algorithms were used to identify each tree’s canopy in the imagery and estimate each canopy’s size. The canopy size estimates were then used to estimate the diameter of each tree. These diameter estimates formed the basis for modeling, which incorporated environmental data from the plot and set size limitations to increase reliability of remote sensing-based predictions. The model was then used to predict the tree size distribution for a given area.
The last step in this method was to validate the results by comparing tree size distributions calculated from long-term forest monitoring network data with those produced by the model. Testing the remote sensing method by using data from the NEON network showed that the model explained over 70 percent of variance in tree size distribution, indicating a strong fit. When the method was tested by using data from two ForestGEO sites—Harvard Forest in Massachusetts and the Smithsonian Environmental Research Center (SERC) in Maryland—results showed that estimates of tree size distribution from remote sensing were reasonable, but were a better fit for Harvard Forest than SERC. This may be because SERC has more diverse tree species and different forest types than Harvard Forest.
Overall, this method could be a cost-effective way to estimate tree size distribution, but it may be more accurate in forests that have fewer tree species. The estimates of tree size distributions could also be improved by using scaling relationships specifically tailored to different regional forest types and environmental conditions.
Conclusion
Together, the analyses and methods developed here can contribute to better predictions of forest structure and make the process of measuring and predicting forest structure more cost effective and accessible to forest landowners and managers. The research team is currently using the insights gained from this research to improve forest structure predictions and develop tools to make it easier to measure tree size distributions and predict forest structure.
This research was partially funded by the Northeastern States Research Cooperative. The University of Maine additionally received funding from the National Air and Space Administration (Award 80NSSC23K0421) and U.S. Department of Agriculture National Institute of Food and Agriculture (MEO 022425).