Authors: Francesco Serafin*, Colorado State University, Olaf David, Colorado State University, Charles Ehlschlaeger, CERL US Army Engineer Research and Development Center, Andre Dozier, Colorado State University, Jack Carlson, Colorado State University
Topics: Spatial Analysis & Modeling
Keywords: Machine Learning,NEAT,Surrogate-Model,FICUS,CSIP/OMS
Session Type: Paper
Start / End Time: 9:55 AM / 11:35 AM
Room: Washington 6, Marriott, Exhibition Level
Presentation File: No File Uploaded
Applications of spatio-temporal models originating from research should be ubiquitous to use in both research and field environments. However, due to their complexity they are rarely suited “out-of the box” for field applications. Results from these models are considered accurate but operating an entire system requires dedicated knowledge, extensive set up, and significant computational time. Questions from field applications conversely require quick and "accurate enough" answers. As a result, attempts to use research models in the field has caused problems for IT deployment, model usability for field user, performance expectations, etc. This contribution proposes a machine learning (ML)-based surrogate model approach aiming to capture the intrinsic knowledge of a spatio-temporal model into an ensemble system of artificial neural networks and make it available for providing simplified answers to on the field problem-specific questions. A surrogate modeling approach was developed to help transitioning from research to the field by enabling a modeling framework to interact with ML libraries to emerge model surrogates for a(ny) modelling solution. The Cloud Services Integration Platform CSIP/OMS was extended and utilized to harvest data and derive the surrogate model at the modeling framework level. NeuroEvolution of Augmenting Topology (NEAT) in an ensemble application, combined with ANN uncertainty quantification are the main methodologies used. Two examples applications have been prototyped and will be presented, a a sheet and rill erosion model and a model for synthesizing plausible geo-referenced households from actual census. The latter is part of Framework Integrating the Complexity of Uncertain System (FICUS) project.