About

Prof. Nesbitt is a Professor and Head of the Department of Climate, Meteorology, and Atmospheric Sciences (CliMAS) at the University of Illinois Urbana-Champaign. He also holds a joint appointment in the Environmental Sciences Division at Argonne National Laboratory. His research and teaching interests reside in radar meteorology, satellite meteorology, tropical meteorology, mesoscale meteorology, and data science. His work has contributed to the understanding of processes involving deep convective systems and high-impact weather, measurements of and processes within global precipitation systems using satellites, processes in winter storms and snowfall, and understanding precipitation systems in complex orography. He has participated in over 20 field experiments, notably serving as the lead Principal Investigator of the NSF/NOAA/NASA RELAMPAGO field campaign and as a Co-Lead Investigator in the DOE CACTI campaign. He is a co-author of the 2018 textbook Radar Meteorology, A First Course, and has received several awards including the NASA Earth and Space Science Fellowship, NASA New Investigator Program award, NASA Robert H. Goddard Award for Scientific Achievement, and was named a Fellow of the American Meteorological Society in 2025.

Research topics

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

Selected publications

• Teaching Machine Learning Techniques Using Agentic Coding

  Zenodo (CERN European Organization for Nuclear Research) · 2026-01-20

  articleOpen access

  Presentation from ESIP 2026 January meetig. Agentic coding using large language models has the potential to simplify the task of developing data analysis and machine learning training scripts for Earth Science. The models are powerful, however proper context is key to generating useful scripts. Teaching responsible use of these tools requires students to understand the data provenance, machine learning techniques, data quality, and model validation. We will discuss the design of course modules that meet these needs and demonstrate how community curated context files can improve the quality of the generated code. Plenty of time will be allowed for participants to discuss the feelings and issues this type of teaching raises.

  PublisherDOI

• Teaching Machine Learning Techniques Using Agentic Coding

  Open MIND · 2026-01-20

  article

  Presentation from ESIP 2026 January meetig. Agentic coding using large language models has the potential to simplify the task of developing data analysis and machine learning training scripts for Earth Science. The models are powerful, however proper context is key to generating useful scripts. Teaching responsible use of these tools requires students to understand the data provenance, machine learning techniques, data quality, and model validation. We will discuss the design of course modules that meet these needs and demonstrate how community curated context files can improve the quality of the generated code. Plenty of time will be allowed for participants to discuss the feelings and issues this type of teaching raises.

  DOI

• The Spatial Area and Other Attributes of GOES-16 Overshooting Tops as Indicators of Potential Hail

  Monthly Weather Review · 2025-08-18 · 1 citations

  article

  Abstract Recent studies using idealized simulations suggest that storms that generate large hail should exhibit deep and wide overshooting tops (OTs). Our work herein extends these and related studies to explore possible relationships between observed OT characteristics and hail size observed at the ground. All hail reports from 2018 through 2022 across the contiguous United States were organized using a grid-hour approach. An OT detection algorithm applied to GOES-16 data was used to find the nearest OT to each selected report. OT area (OTA), OT depth (OTD), and OT volume (OTV) were quantified and statistically related to hail size. OTA tended to exhibit a statistically significant difference across the three hail size groups (nonsevere, severe, and significant severe), with a decrease with increasing hail size. OTD also tended to exhibit a statistically significant difference across the hail size groups, with an increase with increasing hail size. A maximum expected hail size (MESH)-based report proxy was used to explore possible dependencies of these results on hail reports; the general tendencies of increased hail size for decreased OTA and increased OTD were also found using this proxy. Such tendencies were additionally found using a dataset limited to hail associated with supercells. Radar-derived OTA for this limited dataset was also explored and tended to increase with hail size. Finally, possible relationships between hail size and area of a proximal, midtropospheric radar reflectivity core were evaluated and found to be positive and statistically significant. Applications of these findings for risk assessment and operational forecasting are possible but will require further analyses.

  PublisherDOI

• Artificial Intelligence–Enabled Digital Twin for U.S. Cities

  Bulletin of the American Meteorological Society · 2025-09-18 · 1 citations

  article
  PublisherDOI

• Improved Understanding of How Kinematic and Thermodynamic Environmental Changes Impact Modeled Overshooting Top Characteristics

  Geophysical Research Letters · 2025-10-31

  articleOpen access

  Abstract Overshooting tops (OTs) are domed protrusions of deep convective updrafts that extend past the anvil of a cumulonimbus. Recent work has shown that OT depth and area may be sensitive to the thermodynamic environment in the upper troposphere/lower stratosphere (UTLS). What remains unknown is the extent to which the UTLS influence on OT characteristics competes with the influence of the tropospheric updraft. This study uses numerical simulations of supercell thunderstorms to test the relative influences of kinematic and thermodynamic environmental changes on updraft characteristics and ultimately, OT area and depth. Results show static stability in the UTLS is important for modulating depth, but not area. Tropospheric vertical wind shear, which is important for controlling the size of a supercell updraft, is shown to be important for the area of the OT but not its depth.

  PublisherOA PDFDOI

• Evaluating the Performance of Microphysics Schemes against Observations during High-Impact Lake-Effect Snow Events

  Weather and Forecasting · 2025-08-04

  article

  Abstract In the Great Lakes region (GLR), lake-effect snow (LeS) events are a common occurrence, in which narrow, intense bands of convection cause snowfall downwind of the lakes. The shallow convection associated with LeS is dynamically different from deeper, synoptically driven snow, and the particle size distributions (PSDs) of the precipitation have different shapes, as well. This work considers whether or not the Thompson–Eidhammer microphysics scheme, which includes single-moment prediction of snow, is accurate in estimating the PSDs of LeS convection. The High-Resolution Rapid Refresh (HRRR) configuration of the Weather Research and Forecasting (WRF) Model is used to simulate three different LeS events in the GLR using two different microphysics schemes: the Thompson–Eidhammer “aerosol-aware” scheme and the Morrison double-moment scheme. Model-estimated PSDs are calculated and compared to observed PSDs at three locations in the region: Marquette, Michigan; Gaylord, Michigan; and Buffalo, New York. Model-predicted liquid water equivalent snowfall and snow density are also compared to observed products. It is found that parameterization performance varies depending on location, with Thompson struggling to create the correct PSD shape for Marquette. Both microphysics schemes do not perform well in predicting particles greater than 6 mm in diameter except in Buffalo, where both simulated and observed PSDs contain snow particles greater than 10 mm in diameter. Significance Statement In the Great Lakes region, lake-effect snow events can cause heavy snowfall with large accumulations over a narrow region downwind of the lakes. Current numerical weather prediction models struggle to exactly capture the large spatial variability of lake-effect snow and rely on parameterizations to model the characteristics of the snowfall. This study is the first to use a database of microphysical observations to evaluate how well a current operational forecast model is able to replicate the snow particle sizes in lake-effect snow events around the Great Lakes.

  PublisherDOI

• Improved Understanding of how Kinematic and Thermodynamic Environmental Changes Impact Modeled Overshooting Top Characteristics

  2025-07-15

  preprintOpen access

  Overshooting tops (OT) are domed protrusions of deep convective updrafts that extend past the anvil of a cumulonimbus. Recent work has shown that certain characteristics of an OT such as its depth may be sensitive to the thermodynamic environment in the upper troposphere/lower stratosphere (UTLS). What remains unknown is the extent to which the characteristics of the tropospheric updraft, such as its size and intensity, influence an OT compared to the UTLS thermodynamic environment. This study uses numerical simulations of supercell thunderstorms to test the relative influences of kinematic and thermodynamic environmental changes on updraft characteristics and ultimately, OT area and depth. Results show static stability in the UTLS is important to modulating depth, but not area. Tropospheric vertical wind shear, which is important for controlling the size of a supercell updraft, is shown to be important for the area of the OT but not its depth.

  PublisherOA PDFDOI

• Properties of Cold Pools from PERiLS 2022–23

  Monthly Weather Review · 2025-07-25

  article

  Abstract Cold pools play a range of important roles in quasi-linear convective systems (QLCSs), including maintenance via the development of new convective cells as well as baroclinic generation of horizontal vorticity. Although a number of QLCS cold pools have been characterized in the literature using one or a few sensors, their variability (both internally and across a range of environments) has still not been widely studied. This gap in knowledge extends particularly to high-shear low-CAPE (HSLC) convective environments common to the cool season in the southeastern United States, where the Propagation, Evolution, and Rotation in Linear Storms (PERiLS) field campaign was focused. PERiLS specifically targeted environmental and storm-scale processes in QLCSs, including their cold pools. Our analysis focuses on the heterogeneity and temporal variability of cold pools across short time and spatial scales using numerous surface and sounding observations across five PERiLS QLCSs. The PERiLS cold pools are generally weaker than those previously studied in warm-season, midlatitude QLCSs, likely due to the lower CAPE and higher relative humidity values common to HSLC environments during PERiLS. Nevertheless, the distributions of most PERiLS cold pool variables at least partially overlap with those of previously studied QLCSs. The median PERiLS measurement reveals a cold pool that is ≈2.5 km deep, having a surface temperature decrease of ≈−6°C, and a peak outflow wind gust of ≈13 m s −1 . In the spirit of a “cold pool audit,” we present the internal and case-to-case variability of these particularly well-observed QLCSs. Significance Statement Evaporatively cooled air masses (“cold pools”) are created by quasi-linear convective systems (“QLCSs,” also called “squall lines”), and they in turn play important roles in the maintenance and structures of QLCSs. There have been relatively few direct measurements of cold pool variability, especially for the frequently severe QLCSs occurring during the cool season in the southeastern United States. Numerous surface and upper-air measurements from the recent Propagation, Evolution, and Rotation in Linear Storms (“PERiLS”) field experiment are used to document Southeastern QLCS cold pools. The PERiLS cold pools were surprisingly similar to, albeit somewhat weaker than, those found in prior studies of warm-season QLCSs in other regions.

  PublisherDOI

• Characteristics of Tornadic and Nontornadic QLCS Mesovortices Observed Using Radar and Pod Data from PERiLS

  Weather and Forecasting · 2025-07-29

  article

  Abstract The challenges associated with nowcasting quasi-linear convective system (QLCS) tornadoes are well documented. One key challenge is that QLCS tornadoes typically develop within mesovortices (MVs), but not all MVs are tornadic. This study used radar and in situ Pod data collected during the Propagation, Evolution, and Rotation in Linear Storms (PERiLS) field campaign to examine the characteristics that differentiate tornadic (TOR), wind-damaging (WD), and nondamaging (ND) MVs at various stages in their lifetimes and to investigate the low-level structure of QLCS MVs. Thirty-one QLCS MVs were manually identified and cataloged using the lowest elevation scans of the nearest WSR-88D and C-band on Wheels (COW) radars during the two years of PERiLS. TOR MVs, over their entire lifetimes, had stronger rotational velocities (Vrots), smaller diameters, and slightly longer lifetimes compared to WD and ND MVs. When MVs were analyzed during their pretornadic, predamaging, and prewarning phases (prephases), TOR and WD MVs had similar Vrots; however, TOR MVs typically had smaller diameters and contracted leading up to tornadogenesis, which could benefit nowcasters. In five cases, MVs were observed at the lowest WSR-88D elevation scans but were not visible in the COW data; the MV structure at different elevation angles for one case is presented. Eight Pods showed evidence of MV intercepts, demonstrated most notably by decreases in pressure. COW data, along with relatively weak wind speeds measured by Pods that collected data on MVs, suggest that vertical variations in low-level MV structure and strength can exist, which may not be adequately captured by the WSR-88D network.

  PublisherDOI

• Comparing multi-source urban flood indicators: satellite, simulation, and citizen-reported data

  Environmental Research Water · 2025-09-01 · 2 citations

  articleOpen accessCorresponding

  Urban flooding arises from complex mechanisms, making it challenging to capture accurately with a single detection method. This study evaluates three complementary approaches to detect flooding across three Chicago neighborhoods: (i) Sentinel-1 synthetic aperture radar (SAR), offering weather-independent, high-resolution (10 m) imagery of surface inundation; (ii) the storm water management model (SWMM), simulating combined sewer overflow and drainage performance; and (iii) citizen-generated 311 service requests, capturing observed flooding impacts. By analyzing six storms ranging from severe to mild, we examine how each source uniquely contributes to identifying urban flood events. SAR imagery effectively identifies standing water but can miss brief flooding due to satellite revisit constraints. SWMM provides detailed insights into system-wide drainage behavior yet may underestimate localized street-level flooding. Meanwhile, 311 calls reflect real-world flooding impacts but are vulnerable to underreporting. Statistical overlap analysis highlights chronic flood hotspots repeatedly identified across multiple detection methods, indicating persistent infrastructure and topographic vulnerabilities. Temporal analysis further reveals that while SWMM flooding aligns closely with rainfall peaks, 311 calls typically precede or persist beyond these peaks. Our findings emphasize the value of using satellite observations, hydrological modeling, and resident-reported data in a complementary manner to better interpret patterns in flood timing, severity, and spatial distribution—providing insights that can inform targeted infrastructure improvements and contribute to urban flood resilience planning.

  PublisherDOI

Recent grants

• Collaborative Research: Using RELAMPAGO Observations to Understand the Thermodynamic, Kinematic, and Dynamic Processes Leading to Heavy Precipitation

  NSF · $687k · 2017–2022

Frequent coauthors

• Edward J. Zipser

  University of Oxford

  54 shared

• Greg M. McFarquhar

  52 shared

• Daniel J. Cecil

  Marshall Space Flight Center

  48 shared

• Paola Salio

  44 shared

• Timothy J. Lang

  Marshall Space Flight Center

  26 shared

• Robert J. Trapp

  University of Illinois Urbana-Champaign

  26 shared

• E. Richard Toracinta

  The Ohio State University

  25 shared

• Robert M. Rauber

  21 shared

Labs

• Steve Nesbitt's LabPI

  Not provided

Awards & honors

• NASA Earth and Space Science Fellowship
• NASA New Investigator Program award
• NASA Robert H. Goddard Award for Scientific Achievement
• Distinguished Visitor of the Universidad Nacional de Córdoba…
• Fellow of the American Meteorological Society (2025)