21 Nov 2022
21 Nov 2022
Status: this preprint is open for discussion.

Data Driven Regional Weather Forecasting

Randall Clark1, Luke Fairbanks1, Ramon Sanchez1, Pacharadech Wacharanan1, and Henry Abarbanel2 Randall Clark et al.
  • 1Department of Physics, University of California San Diego, La Jolla, CA 92093, USA
  • 2Department of Physics and Marine Physical Laboratory (Scripps Institution of Oceanography), University of California San Diego, La Jolla, CA 92093,USA

Abstract. Using data alone, without knowledge of underlying physical models, nonlinear discrete time regional forecasting dynamical rules are constructed employing well tested methods from applied mathematics and nonlinear dynamics. Observations of environmental variables such as wind velocity, temperature, pressure, etc allow the development of forecasting rules that predict the future of these variables only. A regional set of observations with appropriate sensors allows one to forgo standard considerations of spatial resolution and uncertainties in the properties of detailed physical models. Present global or regional models require specification of details of physical processes globally or regionally, and the ensuing, often heavy, computational requirements provide information of the time variation of many quantities not of interest locally. In this paper we formulate the construction of data driven forecasting (DDF) models of geophysical processes and demonstrate how this works within the familiar example of a ‘global’ model of shallow water flow on a mid-latitude β plane. A sub-region, where observations are made, of the global flow is selected. A discrete time dynamical forecasting system is constructed from these observations. DDF forecasting accurately predicts the future of observed variables.

Randall Clark et al.

Status: open (until 16 Jan 2023)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse

Randall Clark et al.

Randall Clark et al.


Total article views: 167 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
127 30 10 167 2 1
  • HTML: 127
  • PDF: 30
  • XML: 10
  • Total: 167
  • BibTeX: 2
  • EndNote: 1
Views and downloads (calculated since 21 Nov 2022)
Cumulative views and downloads (calculated since 21 Nov 2022)

Viewed (geographical distribution)

Total article views: 167 (including HTML, PDF, and XML) Thereof 167 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 06 Dec 2022
Short summary
We've trained a machine learning system, based on data assimilation and data driven forecasting fundamentals, to predict 'shallow water' fluid dynamics, such as atmosphere, using limited amounts of data from regional measurements within a twin experiment. The model performs well and is robust to noise, leveraging intuition of physical dynamics at play in the shallow water system to help the machine learning system contextualize the data.