Tue, Aug 03, 2021:On Demand
Background/Question/Methods
Forest management is one of the most extensive and continual drivers of ecosystem dynamics, yet little is known about how management decisions influence forest ecology at regional to continental scales. Our research uses a macrosystems approach to expand our understanding of forest function under different management schemes in the face of a changing climate, and ways to incorporate them into Earth System Models. To map forest management, we analyzed MODIS time series of enhanced vegetation index with the Breaks for Additive Season and Trend (BFAST) algorithm. We used extracted BFAST parameters, ancillary covariates and expert classified training samples in a Random Forest (RF) classifier to infer forest management practices. The Ecosystem Demography Model (ED2) was then used to capture regional impacts of different management approaches and natural disturbances on forest dynamics. Management scenarios used to drive ED2 were derived from interviews with US forest experts, which allowed us to assess regional drivers of management over time, and how climate change impacts and policies will affect management. ED2 results were used to test if management is a stronger predictor of forest structure and function than climate and edaphic conditions, and if this effect persists as aerial extent increases from stand to continent (portfolio effect).
Results/Conclusions We identified, mapped, and modeled four forest management categories: production, passive, ecological, and preservation. Thus far, RF-derived management classes were mapped in two study regions, the US Pacific Northwest and Southeast, with overall accuracies of 89% and 91%, respectively. The most important covariates for classification were: proportion conifer, proportion riparian, and largest decrease in pre-break mean. We completed >100 interviews with public and private forest experts. Interviews inform ED2 model scenarios for each management approach to reflect variation by region and ownership. For example, in California, most family-owned forest is passively managed while industrial ownership uses production forestry equally split between clearcuts and heavy thinnings. Management types are now being incorporated into the ED2 Model by region. Preliminary ED2 results using test sites in the midwestern US and two climate change scenarios show that forest management can mitigate climate change risk. Recognizing that policy, socioeconomics and sustaining renewable resources drive management, we are developing sociologically reasonable scenarios for ED2 simulation of forest dynamics across the US, and testing how these results scale from stands to the continent.
Results/Conclusions We identified, mapped, and modeled four forest management categories: production, passive, ecological, and preservation. Thus far, RF-derived management classes were mapped in two study regions, the US Pacific Northwest and Southeast, with overall accuracies of 89% and 91%, respectively. The most important covariates for classification were: proportion conifer, proportion riparian, and largest decrease in pre-break mean. We completed >100 interviews with public and private forest experts. Interviews inform ED2 model scenarios for each management approach to reflect variation by region and ownership. For example, in California, most family-owned forest is passively managed while industrial ownership uses production forestry equally split between clearcuts and heavy thinnings. Management types are now being incorporated into the ED2 Model by region. Preliminary ED2 results using test sites in the midwestern US and two climate change scenarios show that forest management can mitigate climate change risk. Recognizing that policy, socioeconomics and sustaining renewable resources drive management, we are developing sociologically reasonable scenarios for ED2 simulation of forest dynamics across the US, and testing how these results scale from stands to the continent.