skip to main content

Title: Dual-Polarization Radar Retrievals of Coastal Pacific Northwest Raindrop Size Distribution Parameters Using Random Forest Regression

Radar retrievals of drop size distribution (DSD) parameters are developed and evaluated over the mountainous Olympic Peninsula of Washington State. The observations used to develop retrievals were collected during the 2015/16 Olympic Mountain Experiment (OLYMPEX) and included the NASA S-band dual-polarimetric (NPOL) radar and a collection of second-generation Particle Size and Velocity (PARSIVEL2) disdrometers over the windward slopes of the barrier. Nonlinear and random forest regressions are applied to the PARSIVEL2 data to develop retrievals for median volume diameter, liquid water content, and rain rate. Improvement in DSD retrieval accuracy, defined by the mean error of the retrieval relative to PARSIVEL2 observations, was achieved when using the random forest model when compared with nonlinear regression. Evaluation of disdrometer observations and the retrievals from NPOL indicate that the radar retrievals can accurately reproduce observed DSDs in this region, including the common wintertime regime of small but numerous raindrops that is important there. NPOL retrievals during the OLYMPEX period are further evaluated using two-dimensional video disdrometers (2DVD) and vertically pointing Micro Rain Radars. Results indicate that radar retrievals using random forests may be skillful in capturing DSD characteristics in the lowest portions of the atmosphere.

more » « less
Award ID(s):
Author(s) / Creator(s):
 ;  ;  
Publisher / Repository:
American Meteorological Society
Date Published:
Journal Name:
Journal of Atmospheric and Oceanic Technology
Page Range / eLocation ID:
p. 229-242
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Raindrop size distributions (DSD) and rain rate have been estimated from polarimetric radar data using different approaches with the accuracy depending on the errors both in the radar measurements and the estimation methods. Herein, a deep neural network (DNN) technique was utilized to improve the estimation of the DSD and rain rate by mitigating these errors. The performance of this approach was evaluated using measurements from a two-dimensional video disdrometer (2DVD) at the Kessler Atmospheric and Ecological Field Station in Oklahoma as ground truth with the results compared against conventional estimation methods for the period 2006–17. Physical parameters (mass-/volume-weighted diameter and liquid water content), rain rate, and polarimetric radar variables (including radar reflectivity and differential reflectivity) were obtained from the DSD data. Three methods—physics-based inversion, empirical formula, and DNN—were applied to two different temporal domains (instantaneous and rain-event average) with three diverse error assumptions (fitting, measurement, and model errors). The DSD retrievals and rain estimates from 18 cases were evaluated by calculating the bias and root-mean-squared error (RMSE). DNN produced the best performance for most cases, with up to a 5% reduction in RMSE when model errors existed. DSD and rain estimated from a nearby polarimetric radar using the empirical and DNN methods were well correlated with the disdrometer observations; the rain-rate estimate bias of the DNN was significantly reduced (3.3% in DNN vs 50.1% in empirical). These results suggest that DNN has advantages over the physics-based and empirical methods in retrieving rain microphysics from radar observations.

    more » « less
  2. Lupo, Anthony (Ed.)
    We examine several different features of DSDs based on data and observations from two mid-latitude coastal locations: (a) the Delmarva peninsula, USA, and (b) Incheon, South Korea. In each case, the full DSD spectra were obtained from two collocated disdrometers. Two events from location (a) and one event from location (b) are presented. For (a), observations and retrievals from NASA’s S-band polarimetric radar are included in the analyses as well as retrieved DSD parameters from the dual-wavelength precipitation radar onboard the Global Precipitation Measurement satellite. For (b), the disdrometer-based DSD data are compared with measurements from another sensor. Our main aim is to examine the underlying shape of the DSDs and their representation by the generalized gamma model. 
    more » « less
  3. Abstract

    Current bulk microphysical parameterization schemes underpredict precipitation intensities and drop size distributions (DSDs) during warm rain periods, particularly upwind of coastal terrain. To help address this deficiency, this study introduces a set of modifications, called RCON, to the liquid-phase (warm rain) parameterization currently used in the Thompson–Eidhammer microphysical parameterization scheme. RCON introduces several model modifications, motivated by evaluating simulations from a bin scheme, which together result in more accurate precipitation simulations during periods of warm rain. Among the most significant changes are 1) the use of a wider cloud water DSD of lognormal shape instead of the gamma DSD used by the Thompson–Eidhammer parameterization and 2) enhancement of the cloud-to-rain autoconversion parameterization. Evaluation of RCON is performed for two warm rain events and an extended period during the Olympic Mountains Experiment (OLYMPEX) field campaign of winter 2015/16. We show that RCON modifications produce more realistic precipitation distributions and rain DSDs than the default Thompson–Eidhammer configuration. For the multimonth OLYMPEX period, we show that rain rates, rainwater mixing ratios, and raindrop number concentrations were increased relative to the Thompson–Eidhammer microphysical parameterization, while concurrently decreasing raindrop diameters in liquid-phase clouds. These changes are consistent with an increase in simulated warm rain. Finally, real-time evaluation of the scheme from August 2021 to August 2022 demonstrated improved precipitation prediction over coastal areas of the Pacific Northwest.

    Significance Statement

    Although the accurate simulation of warm rain is critical to forecasting the hydrology of coastal areas and windward slopes, many warm rain parameterizations underpredict precipitation in these locations. This study introduces and evaluates modifications to the Thompson–Eidhammer microphysics parameterization scheme that significantly improve the accuracy of rainfall prediction in those regions.

    more » « less
  4. The OLYMPEX field campaign, which took place around the Olympic Mountains of Washington State during winter 2015/16, provided data for evaluating the simulated microphysics and precipitation over and near that barrier. Using OLYMPEX observations, this paper assesses precipitation and associated microphysics in the WRF-ARW model over the U.S. Pacific Northwest. Model precipitation from the University of Washington real-time WRF forecast system during the OLYMPEX field program (November 2015–February 2016) and an extended period (2008–18) showed persistent underprediction of precipitation, reaching 100 mm yr−1over the windward side of the coastal terrain. Increasing horizontal resolution does not substantially reduce this underprediction. Evaluating surface disdrometer observations during the 2015/16 OLYMPEX winter, it was found that the operational University of Washington WRF modeling system using Thompson microphysics poorly simulated the rain drop size distribution over a windward coastal valley. Although liquid water content was represented realistically, drop diameters were overpredicted, and, consequently, the rain drop distribution intercept parameter was underpredicted. During two heavy precipitation periods, WRF realistically simulated environmental conditions, including wind speed, thermodynamic structures, integrated moisture transport, and melting levels. Several microphysical parameterization schemes were tested in addition to the Thompson scheme, with each exhibiting similar biases for these two events. We show that the parameterization of aerosols over the coastal Northwest offered only minor improvement.

    more » « less
  5. This study evaluates moist physics in the Weather Research and Forecasting (WRF) Model using observations collected during the Olympic Mountains Experiment (OLYMPEX) field campaign by the Global Precipitation Measurement (GPM) satellite, including data from the GPM Microwave Imager (GMI) and Dual-Frequency Precipitation Radar (DPR) instruments. Even though WRF using Thompson et al. microphysics was able to realistically simulate water vapor concentrations approaching the barrier, there was underprediction of cloud water content and rain rates offshore and over western slopes of terrain. We showed that underprediction of rain rate occurred when cloud water was underpredicted, establishing a connection between cloud water and rain-rate deficits. Evaluations of vertical hydrometeor mixing ratio profiles indicated that WRF produced too little cloud water and rainwater content, particularly below 2.5 km, with excessive snow above this altitude. Simulated mixing ratio profiles were less influenced by coastal proximity or midlatitude storm sector than were GMI profiles. Evaluations of different synoptic storm sectors suggested that postfrontal storm sectors were simulated most realistically, while warm sectors had the largest errors. DPR observations confirm the underprediction of rain rates noted by GMI, with no dependence on whether rain occurs over land or water. Finally, WRF underpredicted radar reflectivity below 2 km and overpredicted above 2 km, consistent with GMI vertical mixing ratio profiles.

    more » « less