Data-model integration to develop a knowledge landscape map for improved prediction
Abstract
Earth system scientists are increasingly being asked to predict the occurrence of surprising events in order to minimize their negative impacts and promote system recovery. However, predictions are often based on knowledge of a select subset of system components or by extrapolating data from other locations or previous time periods. These extrapolations are further compromised when nonlinear and cross-scale interactions result in unanticipated emergent behavior. Our goal was to develop a trans-disciplinary approach that facilitates integration of large and diverse types of data and knowledge to fill data/knowledge gaps for underrepresented locations across spatially heterogeneous landscapes, and to ascertain the extent to which knowledge of the past can, or cannot, inform the future.
This poster presents our data-model integration approach, and illustrates its utility to integrate diverse, long-term environmental data with detailed process-based data and knowledge spanning multiple levels of organization obtained from disparate locations to create a fully integrated "knowledge landscape map". This map integrates multiple lines of evidence from specific study locations and time periods in a process-based approach by accounting for spatial heterogeneity in patterns and temporal nonlinearities in processes at multiple interacting scales. We use this map to estimate perennial grass primary production through time (1989-present) across a spatially heterogeneous landscape. The approach differs from simple extrapolation because it accounts for thresholds in grass response through time that distinguishes extreme events (drought, wet periods) that can create surprising dynamics if not accounted for when using relationships based on long-term data.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.H23O2159P
- Keywords:
-
- 1829 Groundwater hydrology;
- HYDROLOGYDE: 1847 Modeling;
- HYDROLOGYDE: 1916 Data and information discovery;
- INFORMATICSDE: 1986 Statistical methods: Inferential;
- INFORMATICS