About

Matthew R. Kumjian is a Professor of Meteorology and Atmospheric Science at Penn State. His research specialties include cloud physics and radiation, mesoscale meteorology, and severe weather. His specific research interests focus on radar meteorology, especially applications of dual-polarization radar, as well as clouds and precipitation physics, winter storms, severe convective storms, microphysical modeling, and related atmospheric phenomena. Kumjian holds a Ph.D. in Meteorology from The University of Oklahoma, earned in 2012, along with a master's degree in Meteorology from the same institution obtained in 2008, and a B.S. in Meteorology completed in 2006. His academic and research work is dedicated to understanding and modeling complex atmospheric processes, contributing to advancements in weather prediction and severe weather analysis.

Research topics

• Meteorology
• Geology
• Environmental science
• Computer Science
• Climatology
• Geography
• Physics
• Atmospheric sciences
• Mechanics
• Political Science
• Telecommunications
• Sociology
• Mathematics
• Engineering
• Remote sensing
• Geophysics
• Geodesy

Selected publications

• Windsond S1H2 modifications for observations in rain

  Zenodo (CERN European Organization for Nuclear Research) · 2026-01-13

  datasetOpen accessSenior author

  Introduction This repository contains 3D model files used to fabricate and modify Windsond S1 radiosondes. The purpose of these modifications is to limit water ingress into the radiosonde electronics during heavy precipitation. A journal article detailing the design and validation can be found here: Link will be added upon publication Printing and Assembly Radiation Shield Print with 0.2 mm layer thickness, 100% infill, and no supports, oriented such that the shield base is on the printing bed. PLA filament is suitable. Printing speed may need to be reduced to prevent print failures. Finished models must be wrapped in adhesive aluminum foil tape to reflect solar radiation. Recommended thickness is 0.04 mm or less. To connect the new radiation shield to the probe arm use the same approach as the original radiation shield - adhesive foil tabs. Note that the probe arm must run through the middle of the radiation shield to allow ventilation. Sensor Support Arm Print with 0.2 mm layer thickness, 100% infill, and no supports, oriented such that the outer long edge is on the printing bed. PLA filament is suitable. Mount by first passing the sensor through the support arm (without a radiation shield). Adhere the support arm to the outside of the polystyrene cup using hot-melt glue. Tether Hole Cover Print with 0.2 mm layer thickness, 100% infill, and no supports, oriented flat on the printing bed. PLA filament is suitable. Ensure the slot remains open, use a craft knife to open it as needed. Place tether through the cover and carefully pull so no slack remains (be careful not to break the tether). Adhere the cover only to the tether using hot melt glue. Additional Recommendations Seal the power switch using a small piece of waterproof tape. Be careful not to touch the exposed T/RH sensor during assembly. All model files are in millimeters

  PublisherDOI

• Windsond S1H2 modifications for observations in rain

  Zenodo (CERN European Organization for Nuclear Research) · 2026-01-13

  datasetOpen accessSenior author

  Introduction This repository contains 3D model files used to fabricate and modify Windsond S1 radiosondes. The purpose of these modifications is to limit water ingress into the radiosonde electronics during heavy precipitation. A journal article detailing the design and validation can be found here: Link will be added upon publication Printing and Assembly Radiation Shield Print with 0.2 mm layer thickness, 100% infill, and no supports, oriented such that the shield base is on the printing bed. PLA filament is suitable. Printing speed may need to be reduced to prevent print failures. Finished models must be wrapped in adhesive aluminum foil tape to reflect solar radiation. Recommended thickness is 0.04 mm or less. To connect the new radiation shield to the probe arm use the same approach as the original radiation shield - adhesive foil tabs. Note that the probe arm must run through the middle of the radiation shield to allow ventilation. Sensor Support Arm Print with 0.2 mm layer thickness, 100% infill, and no supports, oriented such that the outer long edge is on the printing bed. PLA filament is suitable. Mount by first passing the sensor through the support arm (without a radiation shield). Adhere the support arm to the outside of the polystyrene cup using hot-melt glue. Tether Hole Cover Print with 0.2 mm layer thickness, 100% infill, and no supports, oriented flat on the printing bed. PLA filament is suitable. Ensure the slot remains open, use a craft knife to open it as needed. Place tether through the cover and carefully pull so no slack remains (be careful not to break the tether). Adhere the cover only to the tether using hot melt glue. Additional Recommendations Seal the power switch using a small piece of waterproof tape. Be careful not to touch the exposed T/RH sensor during assembly. All model files are in millimeters

  PublisherDOI

• A comprehensive description of the 1 August 2021 Azzano Decimo hailstorm in northeastern Italy

  Frontiers in Environmental Science · 2026-02-02

  articleOpen access

  On 1 August 2021, a vigorous hailstorm hit Azzano Decimo, in northeastern Italy. The supercell storm produced hailstones up to 10 cm in maximum dimension, which is quite unusual in this area. The storm’s environment registered one of the largest potential instabilities ( <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="m1"> <mml:mrow> <mml:mo>&gt;</mml:mo> <mml:mn>3400</mml:mn> </mml:mrow> </mml:math> J <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="m2"> <mml:mrow> <mml:msup> <mml:mrow> <mml:mi mathvariant="normal">k</mml:mi> <mml:mi mathvariant="normal">g</mml:mi> </mml:mrow> <mml:mrow> <mml:mo>−</mml:mo> <mml:mn>1</mml:mn> </mml:mrow> </mml:msup> </mml:mrow> </mml:math> ) ever observed by the local operational Udine radiosonde site. In this paper, we analyze the mesoscale environment supporting the hailstorm. Observations from the nearby operational Fossalon di Grado dual-polarization radar showed the presence of a pronounced Bounded Weak Echo Region and differential reflectivity column, both proxies for intense updrafts; however, Doppler velocities revealed only weaker winds, with the mesocyclone mostly confined to midlevels. Two independent observers in Azzano Decimo collected nine hailstones, including one with a maximum dimension of 9 cm. The physical structure of these hailstones was analyzed in the National Center for Atmospheric Research cold room, including normal and cross-polarized light photographs of thin sections to identify the different growth layers inside each hailstone. Additionally, ice samples were taken from 1 cm <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="m3"> <mml:mrow> <mml:mo>×</mml:mo> <mml:mn>1</mml:mn> </mml:mrow> </mml:math> cm <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="m4"> <mml:mrow> <mml:mo>×</mml:mo> <mml:mspace width="0.3333em"/> <mml:mn>1</mml:mn> </mml:mrow> </mml:math> mm pieces from these thin slices. The stable isotopic ratio analyses were performed on these specimens using a Picarro cavity ring-down spectrometer. Isotopic content of the hailstone layers revealed significant variability, including some internal layers that showed signs of kinetic fractionation owing to evaporating liquid being incorporated into the growth layer, likely from evaporation of surface liquid during wet growth or collection of recirculated raindrops that experienced evaporation prior to participating in hail growth. Despite such large isotope variability, the Jouzel model analysis suggested that the major growth happened at high altitudes (between 8 and 10 km), which is also in agreement with a reversible-adiabatic parcel model and radar observations from the event.

  PublisherDOI

• Investigation of Physical Processes Contributing to Coastal Convection Initiation on 2 June 2022 during ESCAPE

  Monthly Weather Review · 2026-03-31

  articleSenior author

  Abstract Data obtained during the National Science Foundation (NSF)-funded Experiment of Sea Breeze Convection, Aerosols, Precipitation, and Environment (ESCAPE) field campaign and conventional observations are used to evaluate the physical mechanisms responsible for convection initiation (CI) over coastal Texas during the 2 June 2022 intensive observation period. Failed CI attempts first occurred about 1300 UTC [0800 local time (LT)] along a decaying land breeze located parallel to and offshore of the Texas coastline. Subsequent successful CI, which occurred around 1430 UTC (0930 LT), required the assistance of gravity waves that formed as an upstream MCS cold pool perturbed the stable lower troposphere. Upon the interaction between the leading radar-detected gravity wave and the offshore remnants of the land breeze, convective cells developed along the length of the offshore boundary. Outflow from this line of cells moved inland and initiated new convective cells resulting in widespread convection over the coastal zone that continued past 2000 UTC (1500 LT). Results from this work illustrate the complex chain of mechanisms that can lead to CI over coastal regions and emphasize the challenges encountered by forecasters in predicting coastal CI. Significance Statement Forecasting the timing and location of storm formation is a major challenge, particularly in coastal areas. Understanding the forcing mechanisms responsible for convection initiation is paramount for accurate forecasting. This study provides an in-depth analysis of a single day to determine the causes and characteristics of convection initiation. Our results identify several forcing mechanisms responsible for convection, which helps illustrate the complexities of forecasting in coastal regions.

  PublisherDOI

• Data sharing for JAS article "How Does Low-Level Hodograph Shape Influence Hail Production?"

  Zenodo (CERN European Organization for Nuclear Research) · 2026-01-21

  otherOpen accessSenior author

  Data sharing for JAS article "How Does Low-Level Hodograph Shape Influence Hail Production?"

  PublisherDOI

• Data sharing for JAS article "How Does Low-Level Hodograph Shape Influence Hail Production?"

  Zenodo (CERN European Organization for Nuclear Research) · 2026-01-21

  otherOpen accessSenior author

  Data sharing for JAS article "How Does Low-Level Hodograph Shape Influence Hail Production?"

  PublisherDOI

• Modifying Windsonds to Improve In-Storm Measurements

  2026-03-10

  articleOpen accessSenior author

  Abstract. Obtaining reliable thermodynamic and kinematic profiles from within severe convective storms presents a challenge for radiosondes due to the extreme conditions to which the instrumentation is exposed. The Windsond S1 has emerged as a popular tool for severe storm research; however, exposure to heavy precipitation has been recently documented to cause relative humidity (RH) sensor malfunctions, limiting the reliability of in-storm measurements. We introduce modifications to the Windsond S1 design that limit water ingress and thereby mitigate these issues. The modifications include a redesigned radiation shield that blocks falling raindrops while maintaining adequate ventilation, a stabilizing support arm, and a tether seal. Controlled experiments using irrigation-generated precipitation at approximately 300 mm hr-1 demonstrated that unmodified sondes experienced RH sensor failures, suppressed RH variability, and power failures due to water ingress, while modified sondes remained functional throughout the exposure to the extreme rain rate. Validation profiles comparing modified and unmodified sondes launched under varied atmospheric conditions, including nocturnal flights, showed temperature differences of typically less than 1 °C and RH differences less than 10 %, with no systematic biases introduced by the modifications. These design improvements were applied to an extensive number of Windsonds S1 during the 2025 ICECHIP field campaign for in-storm deployments. Tested modifications to the Windsond S1 require minimal expertise to implement. Further, design files have been made publicly available to support the broader severe storms research community.

  PublisherOA PDFDOI

• Environmental Characteristics Supporting Warm-Season Coastal Convection Initiation near Houston, Texas

  Weather and Forecasting · 2025-05-07

  articleOpen accessSenior author

  Abstract Convection initiation (CI) remains a formidable forecasting challenge, particularly along the coast. Examining Houston, Texas, radar observations of isolated cells’ CI spatiotemporal patterns for Junes from 2017 to 2022 revealed three patterns: cells forming exclusively over land (LAND), over coastal waters (GULF), or domainwide, initiating first over land (DW-L) or the water (DW-G). CI and dissipation times varied by regime. LAND events tended to typify the diurnal cycle, whereas GULF events tended to initiate overnight; both had durations &lt; 10 h. In contrast, DW events began overnight and lasted until evening, with durations often exceeding 10–15 h. Synoptic-scale composites for each regime revealed only minimal forcing for ascent, suggesting the local environment’s importance for CI. Composite vertical profiles for CI locations revealed surface-based CAPE &gt; 1500 J kg −1 and CIN &gt; −40 J kg −1 for each regime. LAND had the hottest and driest lowest 1 km AGL, but was moistest between 1 and 2 km, suggesting LAND parcels originating below 1 km may be susceptible to entrainment and require moister midlevels for successful CI. We also found conditional instability below 1 km AGL for all regimes but a stable layer for GULF and neutral layers for LAND and DW-G between 1 and 2 km. This indicates saturation of air parcels within this layer is insufficient for CI, and mechanical lifting (e.g., sea breeze) would be necessary for CI. Indeed, all regimes featured potential instability throughout the lowest 4 km. However, only the LAND regime had a coastal density gradient conducive to sea-breeze formation; this indicates other lifting mechanisms may be important in the other regimes. Significance Statement Forecasting the timing and location of storm formation is a major challenge, particularly in coastal areas. We endeavor to understand storm formation patterns in the Houston, Texas, area, with the main goal of better understanding how precursor atmospheric conditions may favor or disfavor such storm formation. We find four spatial patterns of storm formation: only over the land, only over coastal gulf waters, or over both land and gulf (but starting over land or the gulf). Average large-scale and local conditions were similar for each regime, with only subtle differences in their low-level temperature and humidity profiles. Results suggest that small-scale features like sea breezes thus are required for initiation, but only the LAND regime has sea-breeze-favorable conditions.

  PublisherOA PDFDOI

• Detection of Multi-Modal Doppler Spectra. Part 1: Establishing Characteristic Signals in Radar Moment Data

  2025-02-25

  preprintOpen access

  Abstract. Vertically pointing millimeter-wavelength radars provide a wealth of information about cloud and precipitation particle properties. Doppler spectral data can inform on how particles of varying vertical velocities contribute to total backscattered power observed. It is more computationally cost effective to process moment data instead of spectra data, but doing so leaves valuable information on the cutting room floor. To confidently identify a multi-modal spectra event, in which two or more modes are present within a layer, Doppler spectral data are essential. This means long-term identification of layers featuring multi-modal spectra can be cost prohibitive. To address this, we explore three multi-modal spectra cases from winter precipitation events to determine characteristic signatures of these layers in the moment data averaged over short time periods (~145 s) and explore how these layers differ from the rest of the vertical profiles. We find that the mean spectrum width and the standard deviation of mean Doppler velocity can be used to determine whether or not a layer is multi-modal. In particular, multi-modal layers in mixed-phase and ice clouds feature larger mean spectrum width (exceeding 0.19 m s-1) and smaller standard deviation of the mean Doppler velocity (below 0.1 m s-1). In Part 1 of this study, the identification criteria and methods are described. In Part 2, we perform a verification of the method for three years of vertically pointing radar data, and explore the meteorological conditions associated with identified multi-modal spectral events.

  PublisherDOI

• Studying Aerosol, Clouds, and Air Quality in the Coastal Urban Environment of Southeastern Texas

  Bulletin of the American Meteorological Society · 2025-08-04 · 3 citations

  article

  Abstract A multi-agency succession of field campaigns was conducted in southeastern Texas during July 2021 through October 2022 to study the complex interactions of aerosols, clouds and air pollution in the coastal urban environment. As part of the Tracking Aerosol Convection interactions Experiment (TRACER), the TRACER- Air Quality (TAQ) campaign the Experiment of Sea Breeze Convection, Aerosols, Precipitation and Environment (ESCAPE) and the Convective Cloud Urban Boundary Layer Experiment (CUBE), a combination of ground-based supersites and mobile laboratories, shipborne measurements and aircraft-based instrumentation were deployed. These diverse platforms collected high-resolution data to characterize the aerosol microphysics and chemistry, cloud and precipitation micro- and macro-physical properties, environmental thermodynamics and air quality-relevant constituents that are being used in follow-on analysis and modeling activities. We present the overall deployment setups, a summary of the campaign conditions and a sampling of early research results related to: (a) aerosol precursors in the urban environment, (b) influences of local meteorology on air pollution, (c) detailed observations of the sea breeze circulation, (d) retrieved supersaturation in convective updrafts, (e) characterizing the convective updraft lifecycle, (f) variability in lightning characteristics of convective storms and (g) urban influences on surface energy fluxes. The work concludes with discussion of future research activities highlighted by the TRACER model-intercomparison project to explore the representation of aerosol-convective interactions in high-resolution simulations.

  PublisherDOI

Recent grants

• PREEVENTS Track 2: Collaborative Research: Improving High-Impact Hail Event Forecasts by Linking Hail Environments and Modeled Hailstorm Processes

  NSF · $305k · 2019–2023

• Collaborative Research: An Integrated Understanding of the Initiation and Subsequent Dynamical and Microphysical Characteristics of Deep Convective Storms during RELAMPAGO

  NSF · $368k · 2017–2022

Frequent coauthors

• Alexander V. Ryzhkov

  Cooperative Institute for Mesoscale Meteorological Studies

  102 shared

• Terry J. Schuur

  NOAA National Severe Storms Laboratory

  34 shared

• Pavlos Kollias

  Stony Brook University

  32 shared

• Robert S. Schrom

  Goddard Space Flight Center

  28 shared

• Marcus van Lier‐Walqui

  Goddard Institute for Space Studies

  27 shared

• O. P. Prat

  26 shared

• Mariko Oue

  Stony Brook University

  26 shared

• Scott M. Ganson

  24 shared