Big Data and Big Models: Using NEON Data to Inform the Community Land Model

Abstract:

A grand challenge in environmental science is predicting the future trajectory of the terrestrial carbon cycle as simulated in coupled Earth System Models. However, simulations remain highly uncertain despite ever increasing availability of observations. Therefore, finding new ways to use data to evaluate, benchmark and constrain models, and improve forecasts through data assimilation, is critical to making progress in reducing these uncertainties.

One of the largest sources of “big data” for use in biogeosciences is becoming available from the NSF-funded National Ecological Observatory Network (NEON). This represents enormous potential for enhancing carbon cycle modeling, but also a new challenge in how such large amounts of information can be utilized most effectively. NEON data will be streaming from approximately 15,000 sensors, of roughly 200 distinct types with measurement frequencies up to 40 hertz. Observations will be made at 2000 plots at over 60 sites, and NEON will acquire airborne hyperspectral and LiDAR data over 5500 km² at sub 1-meter resolution annually. In all, NEON will be producing about a petabyte of data per year.

In this presentation we first provide a update on the status of NEON. Second, we illustrate how many NEON observations correspond with key pools and fluxes of carbon and water in land surface models at a variety of spatial and temporal scales. We then highlight some of the challenges associated with using site level data such as these with gridded, global land surface models due to mismatches in scales, and how these might be addressed. Finally, we demonstrate how infrastructure we have developed coupling the Community Land Model and the Data Assimilation Testbed allows us to assimilate NEON-type data, and conduct an observing system simulation experiment that investigated the impact of network design on modeled carbon pools and fluxes across North America.