Precision Meteorological Prediction Employing A Data-Driven, Adaptive, Real-Time (DART) Approach

Author

Sujit Sinha, University of KentuckyFollow

Date Available

12-20-2024

Year of Publication

2023

Document Type

Doctoral Dissertation

Degree Name

Doctor of Philosophy (PhD)

College

Engineering

Department/School/Program

Mechanical Engineering

Advisor

Dr. Alexandre Martin

Abstract

The objective of this research is to improve precision meteorological prediction, benefiting both uncrewed aerial vehicle (UAV) operations by avoiding hazardous flight conditions and tracking contaminant or toxic clouds caused by incidents, accidents, or fires. The operational concept is to collect meteorological observations from instruments mounted on a swarm of UAVs, telemeter the data to a weather simulation model, revise the predicted solution by adapting the flow field to the observations, and then transmit the updated forecast to appropriate personnel as well as to the guidance system of UAVs for repositioning to the next set of most beneficial measurements. While a data-driven, adaptive, real-time (DART) approach is the ultimate goal, the scope of this research is to specifically assess the viability of the Weather Research and Forecast (WRF) model augmented with a retrospective cost adaptation (RCA) algorithm. Thus, this effort is focused on the data-driven and adaptive aspects of the approach, while being mindful of the real-time requirement. The WRF model was customized to feed measured u and v horizontal wind speed components at specified times and locations into the RCA algorithm, which then generated independent forcing terms, F_X and F_Y, that were respectively applied to the x-direction and y-direction horizontal momentum equations, thereby adapting the simulation to the measured u and v values. As a result, the flow field of the entire geographic region of interest was affected. To collect u and v model validation data, three properly instrumented UAVs were test-flown on 10 November 2022, within a limited geographic region, at the University of Kentucky flight field. Unfortunately, data from the flight field's weather tower was not available for the date of the flight test, and consequently, data gathered by one of the UAVs had to be used as control measurements for model validation rather than as input to the RCA algorithm. Three different validation approaches were examined, which included a current two-locations approach, a batch upload approach, and a batch upload approach with inverse distance weighting. In all cases, when compared to the baseline WRF model, the WRF model with the RCA algorithm improved simulation accuracy based on reductions in the u and v mean absolute errors relative to observation. The level of reduction of the u and v mean absolute errors ranged between 14.0% and 66.1% depending on the validation approach. Independent verification and validation using flight test data covering the entire simulated geographic region of interest ought to be performed to confirm and better refine the amount of accuracy improvement.

Digital Object Identifier (DOI)

https://doi.org/10.13023/etd.2023.492

Funding Information

This study was supported by the National Science Foundation, Computer and Network Systems award (No.: CNS-1932105) in 2019.

Recommended Citation

Sinha, Sujit, "Precision Meteorological Prediction Employing A Data-Driven, Adaptive, Real-Time (DART) Approach" (2023). Theses and Dissertations--Mechanical Engineering. 218.
https://uknowledge.uky.edu/me_etds/218