PARENT SESSION
Contributed Oral Session 64: GIS / Remote Sensing and Landscape Ecology
Tuesday, August 9, 1:30 PM - 5:00 PM, Meeting Room 519 B, Level 5, Palais des congrès de Montréal

Improved estimates of net primary productivity from MODIS satellite data at regional and local scales.

Pan, Yude^*,1, Hom, John¹, Birdsey, Richard¹, McCullough, Kevin¹, Clark, Kenneth ¹, ¹ USDA Forest Service, Newtown Square, PA 19073, USA

ABSTRACT- Estimates of net primary productivity (NPP) from the MODIS satellite have demonstrated utility at global to continental scales. We evaluated the accuracy of MODIS NPP estimates for forest types of the Mid-Atlantic region of the U.S. by inter-comparison and cross-validation among annual NPP estimates from different methodologies: processed MODIS satellite data, a forest ecosystem process model (PnET-CN), and the forest inventory and analysis (FIA) data. This unique comparison provided an extensive evaluation of the MODIS NPP product for different forest types at a regional scale, and indicated possible corrections to the MODIS NPP estimates. The MODIS algorithm provides accurate NPP estimates for the dominant oak-hickory forests in the region, underestimates NPP for less-dominant northern hardwood forests, and overestimates NPP for coniferous forests. Two causes of inaccurate estimates of NPP by MODIS were identified: 1) an aggregated classification and parameterization of diverse deciduous forests in different climatic environments into a single class that averages different radiation conversion efficiencies; and 2) lack of soil water constraints on NPP for forests or areas that occur on thin or sandy, coarse-grained soil. We developed the "available soil water index" (SW_index) for adjusting the MODIS NPP estimates, which significantly improved NPP estimates for coniferous forests. We identified a need for refined classification and parameterization to improve the NPP estimates for northern hardwood forests. At the regional scale, underestimates and overestimates by MODIS NPP offset so that the regional mean is very close to the mean estimates from PnET-CN and FIA. Estimates from the ecosystem process model are closer to the field-based estimates from FIA because of more specific forest classes and parameters derived from ecological studies in the region. The MODIS NPP estimates have many advantages such as globally continuous monitoring and remarkable accuracy for large scales. However, at regional and local scales, our study indicates that it is necessary to adjust estimates to regional vegetation types and soil water conditions.

Key words: MODIS, net primary production (NPP), ecosystem modeling, forest inventory data