A novel approach to mapping and quantifying age classes of forest habitat to support avian habitat management in the upper Great Lakes

Background/Question/Methods

The implementation of management plans to halt and then reverse declining trends of forest-associated bird species dependent on early successional habitat (ESH) requires information about the amount, configuration, quality, and location of habitat. ESH-dependent bird species rely on certain forest age classes, but current habitat datasets do not allow managers to address questions about the amount and configuration of these age classes across a wide range of spatial scales pertinent to management efforts. These deficiencies impede assessments of bias for monitoring programs and the modeling and mapping of species distribution patterns. Landsat time series stacks (LTSS) contain spectral-temporal information revealing the timing and severity of forest canopy disturbance events. We used LTSS and a Vegetation Change Tracker (VCTw) algorithm to produce maps of 5-year forest age classes across the upper Midwest states of Michigan, Minnesota, and Wisconsin. Using Forest Inventory and Analysis (FIA) data from the U.S. Forest Service, we carried out whole area assessments by comparing FIA and VCTw age class area estimates for the study region. We also conducted site-specific assessments by comparing age classes between VCTw pixels and FIA inventory plots. These assessments will enable informed use of the VCTw age class layer in habitat assessments.       

Results/Conclusions

Across the study region, we estimated total forest land to be 23.4 million ha based on VCTw; this estimate was greater than the 21.7 million ha (±181,000) estimate based on FIA data. In contrast, VCTw area estimates for 5-year age classes younger than 20 years old were less than estimates produced with FIA data. Comparisons of VCTw pixels to FIA plots indicated that VCTw correctly classified 90.6% of forested FIA plots and 92.3% of non-forested plots. We assigned 5-year age classes younger than 20 years old to plots based on in-field assessments by FIA personnel. We found that the percentage of plots correctly classified by VCTw ranged from 13.7% to 30.2% for the 15-to-20-year and 0-to-5-year age classes, respectively. Classifications results were not appreciably affected by assigning age classes to FIA plots based on canopy disturbance or by using 10-year rather than 5-year age classes. For our study region, our assessment results are similar to assessments comparing other widely used geospatial products, such as LANDFIRE, to FIA data. Future evaluations will further build confidence in the age class map by examining the efficacy of age class-related predictors in regional models of wildlife abundance.