Linking In Situ Measurements, Remote Sensing, and Models to Validate MODIS Products Related to the Terrestrial Carbon Cycle

Warren B. Cohen
Pacific Northwest Research Station, USDA Forest Service

David P. Turner
Forest Science Department, Oregon State University

Stith Tom Gower
Department of Forest Ecology and Management, University of Wisconsin

Steven W. Running
School of Forestry, University of Montana

 
Funded by NASA's Terrestrial Ecology Program,
Dr. Diane E. Wickland, Manager  

---

Synopsis

The overall goal of BigFoot is to provide validation of MODLand (MODIS Land Science Team) science products, including land cover, leaf area index (LAI), fraction absorbed photosynthetic active radiation (f_APAR) , and net primary production (NPP). To do so, we use ground measurements, remote sensing data, and ecosystem process models at sites representing different biomes. BigFoot sites are 5 x 5 km in size and surround the relatively small footprint (1 km²) of CO₂ flux towers. At each site we make multi-year in situ measurements of ecosystem structure and functional characteristics that are related to the terrestrial carbon cycle. Our sampling design allows us to explicitly examine scales of fine-grained spatial pattern in these properties, and provides for a field-based ecological characterization of the flux tower footprint. Multi-year measurements ensure that inter-annual validity of MODLand products can be assessed.

For each measurement year we derive land cover, f_APAR, and LAI surfaces by linking our in situ measurements to Landsat ETM+ data. These BigFoot surfaces are developed using logic that preserves functionally important fine-grained information. Errors in these surfaces are quantified and the surfaces summarized to provide a characterization of vegetation patterns in the greater flux tower footprint. Using these land cover and LAI surfaces and derived climate surfaces, we model NPP over the 5 x 5 km BigFoot footprint. Two independent ecosystem process models are used: Biome-BGC and IBIS. The ability of the models to capture environmental and ecological controls on water and carbon cycles is assessed with the following comparisons: modeled NPP against in situ measurements of NPP, modeled GPP to tower-based calculations of GPP, and modeled daily water vapor and CO₂ fluxes to tower estimates. We validate MODLand land cover, LAI, f_APAR, and NPP surfaces by comparing them to BigFoot surfaces derived using field measurement data. A series of exercises that isolate important scaling factors is conducted, so that their effects on NPP model estimates can be better understood. This involves rerunning the models after converting site-specific land cover classes into broad, globally applicable classes, successive coarsening of land cover and LAI surface grain size, and generalizing the light use efficiency factor () to coincide with the more generalized land cover classes.

There are nine BigFoot study sites that span eight major biomes, from desert to tundra, to tropical forest. At these sites, in addition to validation of MODIS products, we quantify carbon content and NPP, examine how these variables vary spatially and temporally, and how NPP is related to climatic variables. Collectively, the standardized NPP data from the contrasting biomes elucidates biophysical controls on NPP, and their sensitivitiy to changing climate and land use. Our standardized data also allow for direct testing of whether light use efficiency (LUE) differs among plant functional types, or seasonally for a given type.

A global terrestrial observation system is needed to assist in the validation of global products such as land cover and NPP from MODIS and other sensor and modeling programs. A key component of such a system is the eddy flux tower network, FLUXNET; however, flux sensors measure net ecosystem productivity (NEP), not NPP. BigFoot is learning how NEP and NPP are related, and through modeling, how to integrate a wide range of carbon cycle observations. Another key component of an observing system is the use of remote sensing and models to scale tower fluxes and field measurements. Although this may be relatively common at a given site, no other project is doing so with standardized methods across so many biomes. As such, BigFoot is a pathfinding activity that will contribute to the development of useful scaling principles. The project can also serve as a nucleus for the global terrestrial observing system that is needed to validate global, generalized products used to monitor the health of the terrestrial biosphere.