Regional Scale Associations between Tree Diversity and Remotely Sensed Estimates of Productivity within the U.S.A.

Ecologists theorize that climate variation accounts for most of the differences recorded in biodiversity across regions. The effects of climatic variation can be best integrated through models of productivity. We tested four models of productivity, derived with remotely sensed data, against estimates of tree species diversity per hectare obtained from over 178 000 Forest Inventory and Analysis survey plots delineated within >2500 1000 km2 cells for the contiguous U.S.A. The productivity models represent increased complexity. The simplest included 16-day MODIS (MODerate resolution Imaging Spectro-radiometer) maximum value normalized difference and enhanced vegetation indices (NDVI and EVI) images. The second level of complexity utilized models based on the radiation use efficiency concept requiring some climatic data and included the 8- day MODIS-GPP and 10-day SPOT (Systeme Pour l`Observation de la Terre) Vegetation, net primary production products. The most complex model, 3-PGS (Physiological Principles Predicting Growth using Satellite data) demands additional climate and soils information. We estimate the maximum gross productivity of U.S. forests to approach 30 tC ha-1 yr-1. MODIS NDVI and EVI saturate for moderately productive forests (~ 10-15 tC ha-1 yr-1), resulting in an exponential relationship with tree diversity. Both the MODIS GPP and SPOT NPP products reveal a similar power function relationship with tree diversity due to underestimation of highly productive (>20 tC ha-1 yr-1) forests located in the northwest and eastern regions of the U.S.A. The 3-PGS model predicted a parabolic relation between tree diversity and the projected full range of forest GPP. However, the model only accounted for ~ 50% of the observed variation in species richness. Relationships remained similar whether productivity was assessed throughout the year or for the growing season. While simple satellite-based models provide reasonable estimates of forest productivity, consideration of soil properties is required to capture the full range of productivity recorded across the U.S.A.