This was presented by Peter Clark as a part of a series of contributed talks from the 2022 FEMC Annual Conference. To learn more about the conference, visit: https://www.uvm.edu/femc/cooperative/conference/2022. Field-based forest inventories, like the United States Department of Agriculture's Forest Inventory and Analysis (FIA) program, can provide unbiased stock change estimates of carbon fluxes but are limited by the density and temporal frequency of the sample. However, when field samples are combined with remote-sensing data, a spatially explicit representation of forest inventory information can be produced in the form of a map-based stock change assessment to estimate carbon fluxes at finer spatial and temporal resolutions than possible with field inventory information alone. Time series pixel predictions provide the flexibility to aggregate these individual predictions to units relevant to local forest management, opening the door to the promotion and incentivization of forest-based climate solutions. Landsat time series imagery and the accompanying open-source toolkit, including the Google Earth Engine cloud computing platform and the LandTrendr temporal segmentation algorithm, make this fine-resolution map-based stock change approach feasible.
Over 5,000 FIA plots sampled between 2002 and 2019 across New York State were coupled with Landsat spectral indices, disturbance metrics derived from LandTrendr, and topographic and climatic geodata to develop machine learning models for aboveground biomass (AGB) prediction. With these models we generated annual maps (1990-2019), at a 30m resolution, to characterize historical AGB stocks, changes, and spatial patterns across New York State. We then compared our map-derived estimates to a common set of FIA plots, and FIA aggregate estimates across time and across a range of scales. We present this approach and the resulting map products to meet continued demand for time series biomass prediction and mapping, and applications in carbon stock change estimation and ecosystem stewardship.