Changing Primary Production and Biomass in Heterogeneous Landscapes: Estimation and Uncertainty Based on Multi-Scale Remote Sensing and GIS Data.

Abstract

Changes in vegetation gross primary production (GPP) and forest aboveground standing biomass C (biomass) were investigated to better understand the impacts of human and natural disturbances on carbon uptake and storage in two northern hemisphere terrestrial ecosystems, which represent primarily forest-dominated and primarily urbanizing landscapes, respectively. Research focused on evaluating changes in (a) GPP in Southeastern Michigan, where the dynamics of vegetation carbon flux are tied closely to urbanization and human development characteristics; and (b) biomass in Eastern Siberia, where biomass changes are strongly affected by disturbance and regrowth.

Both projects exploited remotely sensed data and biophysical or ecosystem models to estimate GPP and biomass. Altering the scale of spatial data and summary units may generate different results. The scaling effects need to be characterized to evaluate scale-related uncertainties associated with carbon estimates. In Michigan, sensitivity of inferences about productivity trends among several development types to levels of aggregation in the Census housing data were examined from the block-group to county scales. In Siberia, impacts of changes in the remote sensing observational scale (i.e., sensor resolution) on the estimated biomass trends were analyzed at resolutions from 60 to 960 meters.

Results showed that GPP increased by 53 g C m-1 in Southeastern Michigan and biomass increased by 3.9 Mg C ha-1 in Eastern Siberia between 1990 and 2000, and that more productive landscapes resulted from tree-cover expansion and forest recovery, respectively. These results corroborate previous findings of increased vegetation activity throughout the northern hemisphere in 1990s. With respect to scaling effects on carbon estimates, in Michigan, relationships between the estimated GPP trends and development types remained consistent across Census scales; and, in Siberia, degradation of remote sensing resolution resulted in the overestimation of changes in biomass by 9-69% at the 960-meter resolution. Results suggested that, for carbon analysis across broader geographic extents (e.g., regional- to national-scale estimation), coarser Census scales up to the county level may be used to evaluate carbon trends by development intensities, while remote sensing data at coarser resolutions may not maintain accuracy of the estimated carbon trends relative to finer resolution data.