Shankar's choices

[ISSS2020]

Building a coupled model from process-oriented software components (models or model elements) 2
Building “system of systems” models by assembling sub-models of particular systems (for example, a “lake system” model integrated with a “watershed system”) -2
Operating models/components in multiple different frameworks -1
Operating models and data inputs/outputs efficiently as part of a sequence of tasks (approach: use/encourage file formats that are both standardized and open)-2
Swapping input data sources (for example, comparing behavior of a model with two different satellite-based inputs of land cover, as opposed to having the model hard-wired to one particular source)-3
Controlling parameter values and behavior without recompiling -2
Operating a model on multiple platforms 1
Retrieving information about a model’s current state (including state variables) (implementation question: direct memory exchange vs. file-based exchange vs. web API)-1
Pausing and continuing model execution-1
Adjusting model variables and/or control parameters during a run (for example, to support data assimilation) -2
Computing derivatives where applicable, to facilitate operations such as sensitivity analysis, optimization, and inference (note: different views among participants about whether this should be included in a standard, a “best practice” guideline, or not at all)-2
Metadata and documentation related to interoperability -1
Clarity and precision in definitions of parameters and variables (ontology)-1
Data items to include in metadata: scale (space and time), typical run time, limits (e.g., range of calibration data)-2