Developing aboveground biomass yield curves for dominant boreal tree species from time series remote sensing data

e-journal

14213

200211721

Forest aboveground biomass (AGB) is an important attribute informing on carbon storage, forest function, and habitat condition. Accurate knowledge of current AGB and its dynamics is essential for sustainable forest management and carbon monitoring. Common methods for estimating AGB, such as permanent sample plots, yield curves, or simulations, often fail to adequately capture the spatial distribution and structural complexity of forest attributes. To address these limitations, we present an integrated model-driven, data-informed approach for developing AGB yield curves exclusively from remotely sensed data, including an annual time series data of Landsat informed annual AGB values, tree species composition, and age. We applied this approach to a 76.5 million-hectare study area, encompassing diverse forest conditions, species, and ages, partitioned into 34 150 × 150-km analysis tiles to account for local variation. The 37-year AGB time series (1984–2021) were filtered to create a representative and noise-reduced sample set for developing remote sensing-derived AGB yield curves (RSYC). Using a nonlinear mixed-effects modeling framework, we generated 127 RSYC models for eight tree species across the study area. Developed yield curves offered insights into AGB dynamics across different forest types and conditions. The performance of RSYC models was evaluated using three independent datasets: permanent sample plots, existing yield curves, and an established growth and yield simulator. Assessment of the RSYC models showed the influence of geographic position and tree species representation in the reference data. In general, the RSYC models tended to underestimate AGB and AGB increments, with relative RMSE ranging between 22.66% and 70.30% for permanent sample plots. We discuss the challenges associated with model validation, data filtering processes, and the advantages of utilizing wall-to-wall AGB time series data from remote sensing. Our findings confirm the feasibility of developing AGB yield curves exclusively from remotely sensed data, covering a wide range of species and stand structural conditions representing a large spatial extent.Keywords: