Near real-time forecasting of terrestrial carbon and water pools and fluxes
Abstract
Near-term iterative ecological forecasting has the potential to improving and accelerating our science, answering fundamental questions about predictability, and responding to societal needs for improved environmental decision making in an era of rapid environmental change. Here we describe our efforts to develop an automated, near-real time forecast system for terrestrial pools and fluxes of carbon and water. This system builds on PEcAn (Predictive Ecosystem Analyzer) to automate the daily execution of an ensemble of terrestrial ecosystem models across a network of Ameriflux eddy-covariance tower sites able to provide near real-time data. Forecasts are driven by 16-day ensemble weather forecasts and available through a daily email and an interactive web visualization portal, which explore both future predictions and validate how well each of the previous 16 daily forecasts did at predicting the most recent observations. Here we report a retrospective analysis of how predictive error varied across sites, models, and environmental conditions, and which uncertainties dominated forecast error, so as to better understand the forecast horizons for carbon and water fluxes and the prediction of plant water stress. The forecast system then employs PEcAn's state data assimilation (SDA) tools to iteratively update the next forecast based on current observations. The SDA currently assimilates observed soil moisture, eddy-covariance carbon and water fluxes, and MODIS LAI. Since a key objectives of our forecast is the prediction and detection of plant stress, both in agricultural systems and native vegetation, we aim to add ECOSTRESS thermal data, SMAP soil moisture, and multispectral imagery (e.g. MODIS, LANDSAT, Sentinel, GOES) to our assimilation soon. Our longer-term goal is to validate our ability to predict fluxes out-of-sample and extend our assimilation system to larger spatial regions rather than just tower sites.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.B13A..04D
- Keywords:
-
- 0410 Biodiversity;
- BIOGEOSCIENCES;
- 0439 Ecosystems;
- structure and dynamics;
- BIOGEOSCIENCES;
- 0480 Remote sensing;
- BIOGEOSCIENCES;
- 1910 Data assimilation;
- integration and fusion;
- INFORMATICS