Abstract
BACKGROUND: As Earth observation satellite data proliferate, so too do maps derived from them. Even when two co-located maps are produced with low overall error, the spatial distribution of error may not be the same. Increasingly, methods will be needed to understand differences among purportedly similar products. For this study, we have used the four aboveground biomass (AGB) maps for conterminous US generated under NASA's Carbon Monitoring System. We have developed systematic approach to (1) assess both the absolute accuracy of individual maps and assess the spatial patterns of agreement among maps, and (2) investigate potential causes of the spatial structure of agreement among maps to gain insight into reliability of methodological choices in map making. RESULTS: The comparison of the four biomass maps with FIA based total biomass estimates at national scale suggest that all the maps have higher biomass estimate compared to FIA. When the four maps were compared among each other, the result shows that the maps S and K have more similar spatial structure whereas the maps K and W have more similar absolute values. Although the maps K and W were generated using completely different methodological workflow, they agree remarkably. All the maps did well in the dominant forest type with maximum agreement between them. The comparison of difference between maps S and K with regional maps suggests that these maps were able to capture the disturbance and not so much regrowth pattern. CONCLUSIONS: The study provides a comprehensive systematic approach to compare and evaluate different real data products using examples of four AGB maps. Although ostensibly the four maps map the same variable, they have different spatial distribution at different scale. Except the 2003 map, one can use other maps at the coarser spatial resolution. Finally, the disparate information available through different maps indicates a need for a temporal framework for consistent monitoring of carbon stock at national scale.