We have introduced the idea of iterative near-term ecological forecasting, which we believe is a win-win for accelerating environmental research and making it more relevant to society (Dietze et al., 2018). Forecasts embody the scientific method, requiring that (i) all predictions be specific and quantitative, and (ii) all validations be out-of- sample against future events. Forecasts encourage robust, reproducible science, and provide natural protection against overfitting. Frequent, iterative forecasts provide constant feedback, which will accelerate learning, improve model interpretation, and help prioritize future research.
The terrestrial carbon cycle is an area of intensive research for the ecological community because of the considerable (and policy-relevant) uncertainty in long-term future projections. As a consequence, it is an area rich with manipulative experiments, field and tower observations, and remote sensing. Increasingly, these data are becoming available close to real time. There are also a wide range of process-based models available, each representing different hypotheses about ecosystem processes, ranging in complexity from simple models with a few dozen lines of code to the land surface components of Earth System models. The combination of these factors makes the terrestrial carbon cycle an ideal focal area for near-term iterative forecasting.
Results/Conclusions
Many interactive ecological processes are represented in Earth System Models (ESMs) that influence fluxes into and out of the biosphere. As our understanding of these processes is imperfect so land surface models have errors and biases when compared to each other and to observations. Here, we implement an Ensemble Adjustment Kalman Filter (EAKF), a sequential state data assimilation technique to reduce these errors and biases. We implement the EAKF using the Data Assimilation Research Testbed (DART), which is coupled with the Community Land Model version 5 (CLM5). By constraining the current state of the biosphere with multiple types of in-situ and remote sensing observations we can produce an accurate analysis. This can then be used as a starting point for improved forecasts over a variety of time periods.
From a societal perspective, outputs of these forecasts using a model of the whole Earth System include predictions of forest and agricultural productivity, water balance, carbon sequestration, and wildfire risk. Output from ESMs include data products (e.g. maps of biomass and vegetation structure and communities) that are directly usable by a wide range of other researchers, including as input for making other forecasts (e.g. of biodiversity).