16 Jun 2022
16 Jun 2022
Status: this preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).

DeepPrecip: A deep neural network for precipitation retrievals

Fraser King1, George Duffy2,3, Lisa Milani4,5, Christopher G. Fletcher1, Claire Pettersen6, and Kerstin Ebell7 Fraser King et al.
  • 1Dept. of Geography & Environmental Management, University of Waterloo, 200 University Ave W, Waterloo, Ontario, Canada
  • 2NASA, Jet Propulsion Laboratory, 4800 Oak Grove Dr, Pasadena, 91109, California, USA
  • 3Earth and Environmental Sciences, University of Syracuse, 900 South Crouse Ave, Syracuse, New York, USA
  • 4NASA, Goddard Space Flight Center, 8800 Greenbelt Rd, Greenbelt, Maryland, USA
  • 5Earth System Science Interdisciplinary Center, University of Maryland, 5825 University Research Ct suite 4001, College Park, Maryland, USA
  • 6Climate and Space Sciences and Engineering, University of Michigan, Space Research Building, Climate &, 2455 Hayward St, Ann Arbor, Michigan, USA
  • 7Institute for Geophysics and Meteorology, University of Cologne, Albertus-Magnus-Platz, Cologne, Germany

Abstract. Remotely-sensed precipitation retrievals are critical for advancing our understanding of global energy and hydrologic cycles in remote regions. Radar reflectivity profiles of the lower atmosphere are commonly linked to precipitation through empirical power laws, but these relationships are tightly coupled to particle microphysical assumptions that do not generalize well to different regional climates. Here, we develop a robust, highly generalized precipitation retrieval from a deep convolutional neural network (DeepPrecip) to estimate 20-minute average surface precipitation accumulation using near-surface radar data inputs. DeepPrecip displays high retrieval skill and can accurately model total precipitation accumulation, with a mean square error (MSE) 99 % lower, on average, than current methods. DeepPrecip also outperforms a less complex machine learning retrieval algorithm, demonstrating the value of deep learning when applied to precipitation retrievals. Predictor importance analyses suggest that a combination of both near-surface (below 1 km) and higher-altitude (1.5 – 2 km) radar measurements are the primary features contributing to retrieval accuracy. Further, DeepPrecip closely captures total precipitation accumulation magnitudes and variability across nine distinct locations without requiring any explicit descriptions of particle microphysics or geospatial covariates. This research reveals the important role for deep learning in extracting relevant information about precipitation from atmospheric radar retrievals.

Fraser King et al.

Status: open (until 22 Jul 2022)

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

Fraser King et al.

Fraser King et al.


Total article views: 119 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
86 29 4 119 2 2
  • HTML: 86
  • PDF: 29
  • XML: 4
  • Total: 119
  • BibTeX: 2
  • EndNote: 2
Views and downloads (calculated since 16 Jun 2022)
Cumulative views and downloads (calculated since 16 Jun 2022)

Viewed (geographical distribution)

Total article views: 91 (including HTML, PDF, and XML) Thereof 91 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 02 Jul 2022
Short summary
Under warmer global temperatures, precipitation patterns are expected to substantially shift, with critical impacts to the global water-energy budget. In this work, we develop a machine learning model for predicting snow and rain accumulation based on surface radar observations of the lower atmosphere. Our model demonstrates improved skill over traditional methods and provides new insights into the regions of the atmosphere that provide the most significant contributions to high model accuracy.