atmosphere Article
Observed Response of the Raindrop Size Distribution to Changes in the Melting Layer Patrick N. Gatlin 1, * 1 2
*
ID
, Walter A. Petersen 1
ID
, Kevin R. Knupp 2 and Lawrence D. Carey 2
NASA Marshall Space Flight Center, Huntsville, AL 35805, USA;
[email protected] Department of Atmospheric Science, University of Alabama in Huntsville, Huntsville, AL 35805, USA;
[email protected] (K.R.K.);
[email protected] (L.D.C.) Correspondence:
[email protected]; Tel.: +1-256-961-7910
Received: 15 July 2018; Accepted: 10 August 2018; Published: 18 August 2018
Abstract: Vertical variability in the raindrop size distribution (RSD) can disrupt the basic assumption of a constant rain profile that is customarily parameterized in radar-based quantitative precipitation estimation (QPE) techniques. This study investigates the utility of melting layer (ML) characteristics to help prescribe the RSD, in particular the mass-weighted mean diameter (Dm ), of stratiform rainfall. We utilize ground-based polarimetric radar to map the ML and compare it with Dm observations from the ground upwards to the bottom of the ML. The results show definitive proof that a thickening, and to a lesser extent a lowering, of the ML causes an increase in raindrop diameter below the ML that extends to the surface. The connection between rainfall at the ground and the overlying microphysics in the column provide a means for improving radar QPE at far distances from a ground-based radar or close to the ground where satellite-based radar rainfall retrievals can be ill-defined. Keywords: microphysics; radar; precipitation
1. Introduction Radar-based measurements of rainfall are vital to understanding the distribution of precipitation over large watersheds. Even though dual-polarimetric radar technology and the associated parameters measured enable improved rainfall estimation compared to traditional (i.e., single-polarization) radar [1–3], dual-polarimetric radars still suffer from many of the same sampling uncertainties associated with accurately estimating rainfall at distant ranges from the radar. For example, regardless of polarization characteristics, beam broadening causes the spatial resolution of radar measurements to degrade with increasing range [4], enabling non-uniform beam filling to impact the accuracy of radar-based estimates of rainfall [5,6]. Ground-based weather radars obtain measurements at increasingly higher heights as the radar beam travels away from the radar due to Earth curvature and gradients in the refractive index driven by vertical gradients of atmospheric temperature and humidity. Thus, radar derived rainfall maps may not be representative of the actual rainfall totals observed at the ground, especially at great distances from the radar (e.g., during winter a typical 1◦ wide radar beam scanned at an elevation angle of 0.5◦ may encounter melting precipitation around 100 km from the radar) and in regions of complex terrain that can inhibit use of measurements obtained at the lowest elevation angles. It is customary in radar-based quantitative precipitation estimation (QPE) to assume a constant rainfall profile below the lowest altitude of a given radar measurement or to apply a correction based on some model of the vertical profile of reflectivity (VPR), for example, References [7,8], but vertical variability in the raindrop size distribution (RSD) can disrupt these methods by undermining the basic assumption of a constant rain profile. Satellite-based retrieval techniques can also be greatly impacted by RSD variability since due to main beam clutter they must use measurements that are typically no Atmosphere 2018, 9, 319; doi:10.3390/atmos9080319
www.mdpi.com/journal/atmosphere
Atmosphere 2018, 9, x FOR PEER REVIEW Atmosphere 2018, 9, 319
2 of 17 2 of 18
impacted by RSD variability since due to main beam clutter they must use measurements that are typically no lower than 500–1000 m above the ground at nadir points [9]. Moreover, satellite-based radar of rainfall at higher microwave can be greatly affected by lower estimates than 500–1000 m abovemade the ground at nadir points [9].frequencies Moreover, satellite-based radar estimates attenuation [10,11].atAlthough numerous frequencies other factorscan contribute to the sampling uncertainties[10,11]. (e.g., of rainfall made higher microwave be greatly affected by attenuation instrument calibration, sampling etc.), some of which are rather complex [1,5], all calibration, else being Although numerous other factorsstrategy, contribute to the sampling uncertainties (e.g., instrument equal it ultimately our ability describe the vertical column and its sampling strategy,isetc.), some to offully which are rather complex [1,5], of allprecipitation else being equal it variability ultimately using measurements (ground, airbornecolumn or space-based) that determines the accuracy of radar is ourradar ability to fully describe the vertical of precipitation and its variability using radar QPE. measurements (ground, airborne or space-based) that determines the accuracy of radar QPE. This This study study addresses addresses radar-based radar-basedQPE, QPE,and andmore more specifically, specifically, the the range-dependent range-dependent errors errors by by examining the physical origin of the RSD as well as the factors that influence its size evolution. examining the physical origin of the RSD as well as the factors that influence its size evolution. Raindrops Raindropsoften oftendevelop developfrom fromthe themelting meltingof ofprecipitation-sized precipitation-sizedice icehydrometeors hydrometeors[12]. [12].For Forexample, example, melting snowflakesareare a common of rainwater in precipitation. stratiform precipitation. Thus, melting snowflakes a common sourcesource of rainwater in stratiform Thus, characteristics characteristics of the melting layer (ML) prescribe the initial state of the RSD, and as such, of the melting layer (ML) prescribe the initial state of the RSD, and as such, measurements of the measurements of the ML may provide additional beneficial to radar Accordingly, ML may provide additional information beneficialinformation to radar QPE. Accordingly, weQPE. hypothesize that a we hypothesize that(and/or a relatively thick ML (and/or low ML)precipitation produced byresults stratiform precipitation relatively thick ML low ML) produced by stratiform in larger raindrops. results larger have raindrops. studies have were foundobserved that larger raindrops observed at the Severalinstudies found Several that larger raindrops at the groundwere beneath more intense ground beneath more reflectivity bright bands [13–16]. An example ofusing this existing radar reflectivity brightintense bands radar [13–16]. An example of this existing conceptual model vertical conceptual model using vertical profiles of reflectivity is shown in of Figure 1. Thicker deeper) profiles of reflectivity is shown in Figure 1. Thicker (or deeper) portions the bright band in(or equivalent portions of the bright in equivalent radarvia reflectivity factor are seemingly connectedzone via radar reflectivity factorband are seemingly connected precipitation fallstreaks to a temperature precipitation fallstreaks to a temperature favorable for dendritic growth of[12,17], which is favorable for dendritic growth [12,17], whichzone is seemingly enhanced in the presence weak convective seemingly enhanced in thicker the presence of of weak motions. Below these thicker of the motions. Below these regions the convective bright band, enhanced reflectivity (Z) isregions often present bright band, enhanced reflectivity (Z)ofis larger often present at low-levels (0.95) difference between the drop size characteristics (i.e., Dm and Dmax ) of the RSDs observed while the ML is thicker than 315 m compared to those observed when the ML is thinner than 315 m. However, disdrometer observations collected when ML bottom was lower or higher than the median height of 2310 m AGL is8 very Atmosphere 2018, 9, xthe FOR PEER REVIEW of 17 similar between the two sets of Dm samples. In contrast, the Dmax sample observed during low MLs low MLs (i.e., arethan larger thesampled Dmax sampled high MLs. We attribute to (i.e.,
