The Essential Terrestrial Variables (ETV’s) in Support of a National Framework for Numerical Watershed Prediction (Invited)

C. Duffy
L. N. Leonard
S. Ahalt
R. Idaszak
David G. Tarboton, Utah State University
R. P. Hooper
L. E. Band

Abstract H31N-08 presented at 2012 Fall Meeting, AGU


There is a clear national need to provide geoscience researchers with seamless and fast access to essential geo-spatial/geo-temporal data to support physics-based numerical models necessary to understand, predict and manage the nations surface and groundwater resources. Fundamental advances in science such as the evaluation of ecosystem and watershed services, the detection and attribution of the impact of climatic change, represent examples that will require high resolution spatially explicit assessments. In this paper we propose the concept of Essential Terrestrial Variables (ETV's), which we define as those variables that are nominally required to support watershed/catchment numerical prediction anywhere in the continental US and ultimately at the global scale. ETV's would represent a fundamental community resource necessary to build the products/parameters/forcings commonly used in distributed, fully-coupled watershed and river basin models. We argue that there are at last 3 fundamental issues that must be resolved before implementation of ETV's in support of a national water model: 1) data access and accessibility, 2) data scale and scalability, 3) community provenance and data sustainability. At the present time, there is no unified data infrastructure for supporting watershed models, and the data resource itself (weather/climate reanalysis products, stream flow, groundwater, soils, land cover, satellite data products, etc.) resides on many federal servers with limited or poorly organized access, with many data formats and without common geo-referencing. Beyond the problem of access to national data, the scale and scalability of computation for both data processing and model computational represents a major hurdle. This predicament is especially true since a full-scale national strategy for numerical watershed prediction will require data resources to reside very close to numerical model computation. Finally model/data provenance should be sufficient to allow reproducible results and scientific workflows that support continuous data and model tracking, geo-referencing, and general support for model/data reproducibility, data analytics and visualization. Furthermore, these workflows should be readily publishable and discoverable by the broader community with provisions for community comment on workflow utility, effectiveness, and improvement. US policy on access to basic data is generally enlightened in that most national geospatial data can be acquired. However, it is clear that fast and efficient access to all the data and models is not yet available to the scientific community for systems that support integrated watershed modeling. Adoption of ETV's and the concomitant supporting cyber infrastructure would be an important step towards a national framework for numerical watershed prediction.