IPC 15 full program
Session 1
Precipitation Measurement Missions Science
Conveners: Sarah Ringerud (NASA), George Huffman (NASA)
June 15, 2026
Considerations for the Future of Global Precipitation Datasets Authors: George J. Huffman (NASA Goddard Space Flight Center); Sarah Ringerud (NASA Goddard Space Flight Center); Jackson Tan (University of Maryland Baltimore County, NASA Goddard Space Flight Center) Corresponding author email: george.j.huffman@nasa.gov Abstract: The Global Precipitation Measurement Mission (GPM) has passed a decade in operation, building a legacy of unprecedented advances in understanding the global distribution and characteristics of rain and snow. This is demonstrated through thousands of attributions in the scientific literature, public outreach and education, and real-time use by forecasters and decision makers. Much of this applications benefit derived from GPM is due to the use of the GPM Core Observatory (CO), with coincident radar and passive microwave (PMW) radiometer, as a high-quality calibrator for a larger constellation of PMW radiometers with a range of channel frequencies, various orbits, and different viewing geometries. When the constellation collectively provides “frequent” (say, ≤3 hr) sampling, it can enable high-quality global time-interpolated half-hourly coverage. Reprocessing of observations from the Tropical Rainfall Measurement Mission (TRMM) using GPM algorithms combines to create a precipitation record that is completing its third decade. This presentation summarizes discussions and conclusions from a community Future of Precipitation Datasets Workshop, held in March 2026. There are multiple challenges in the near future. For example, changes to the sensors flying on the international constellation of precipitation-relevant satellites need to be evaluated for channel spectral response, window hyperspectral sampling, and sensor and calibration stability that are necessary to support accurate retrievals. As well, most smallsat designs are only designed to last ~3 years, which requires a rapid intercalibration process to make them useful for an appreciable fraction of their lifetimes. The GPM CO has served as an intercalibration standard for the constellation (GMI for Tbs and CORRA for constellation retrievals); its conclusion will require alternative intercalibration approaches. The advent of machine learning schemes for precipitation retrievals will require a testbed capability as these algorithms are advanced and validated. Determining the institutional context for all of these factors will be key to advancing precipitation science and applications with long-term, homogeneously processed data. The presentation will end with a brief summary of the lessons learned during GPM, such as the need for error estimates and for the processing capacity to compute both versions in parallel when there is a version upgrade, as well as the advantages to creating a merged observation-model data field.
OpenSSP Portal Grand Reopening: A Milestone Towards NASA PaSS Authors: Kwo-Sen Kuo (ESSIC, University of Maryland, College Park, Maryland, USA); Bruce Altner (NASA Goddard Space Flight Center); That-Dai-Hai Ton; Robert S Schrom (ESSIC, University of Maryland, College Park, Maryland, USA); Ian S Adams (NASA Goddard Space Flight Center); George J Huffman; Scott A Braun Corresponding author email: kkuo@umd.edu Abstract: OpenSSP, standing for Open Single-Scattering Properties, is first a database of numerically grown or constructed, realistic solid hydrometeors and their (scalar) single-scattering properties (SSPs), and secondly a web interface (portal) to the database for interested researchers to obtain particle structure(s) and their corresponding SSPs. We started the OpenSSP web interface around 2016. It was programmed in JavaScript (JS) and hosted by the Precipitation Processing System (PPS) of NASA. However, the original developer left in 2018, and the JS-based web interface began to fall out of date. By 2023, some of the most useful functions of the portal became unreliable, for example, getting SSPs for an ensemble of particles specified by a particle size distribution (PSD) and/or a mass-dimensional (m-D) relation. NASA's Global Precipitation Mission (GPM) and the Atmospheric Observing System (AOS) projects provided in 2023 support for the renewal of OpenSSP as the first step toward a much richer NASA Particle and Single-Scattering Database (NASA PaSS DB), which will feature an augmented non-liquid hydrometeor collection, including melting hydrometeors and additional solid hydrometeors, with polarimetric SSPs for multiple particle orientations. We also envision a mechanism for NASA PaSS DB to accept community contributions and to include other non-spherical particle species, such as aerosol, dust, or salt particles. OpenSSP is now back in operation at a different URL, https://ParticleScattering.org. (The original URL, https://storm.pps.eosdis.nasa.gov/storm/OpenSSP.jsp, is now defunct.) The following are some notable changes. OpenSSP used to use an HDF file as a convenient substitute for a database management system (DBMS); it now employs a bona fide relational DBMS, PostgreSQL, to deliver better performance in anticipation of the NASA PaSS DB's vastly increased data volume. We have also tweaked the graphical interface to make OpenSSP more intuitive and easier to use.
Using quantile regression in rain retrieval from satellite measurement. Authors: Nicolas Viltard (LATMOS-CNRS); Vibolroth Sambath; Matthieu Meignin (LATMOS-UVSQ); Victor Enescu (LATMOS-UVSQ); Cécile Mallet (LATMOS-UVSQ) Corresponding author email: nicolas.viltard@latmos.ipsl.fr Abstract: DRAIN is a Deep-Learning-based algorithm meant to retrieve rain from GMI brightness temperatures. Taking advantage of co-located data between GMI and DPR, a U-net was trained with four brightness temperature channels as an input (37 GHz H and V and 89 GHz H and V) and DPR Estimated Surface Rain as a target. No other auxiliary data are used as input, assuming the network will be able to capture most of the bidimensionnal transfert function between rain and brightness temperatures. The chosen cost function for training is the quantile loss which provides in each pixel a cumulative distribution function of possible rain rates corresponding to each quantile. In our case, the percentile regression model predicts conditional rain rate quantiles RRQP for P =1,…,99, where RRQP represents the threshold exceeded with probability 1-P % given the input features. Retrieving a quantile distribution for each rainy pixel opens new possibilities for exploiting the retrieved rain field. First, various estimators can be used as “best” estimator of the rain. The most natural one is the theoretical median RRQ50, or eventually the mean of the quantiles RRmean. However, another quantile can be shown to provide an estimator with a better bias: RRQ70. Furthermore, using a histogram matching-like approach, a RRAdj can be computed from the quantiles. We will present the pros and cons of these various estimators and how the quantiles can be used as a confidence interval for the rain estimators. We will also discuss the consequences of using Deep-Learning techniques to retrieve rain in a changing environment.
Evaluation of Precipitation Products under NASA’s Global Precipitation Measurement Mision Ground Validation Program Authors: Ali Tokay (UMBC - NASA/GSFC); David B. Wolff (NASA/GSFC/WFF); Charles N. Helms (UMD - NASA/GSFC); Stephen D. Nicholls (STC - NASA/GSFC); Jason L. Pippitt (Adnet - NASA/GSFC); Alexey V. Chibisov and Mick L. Boulanger Corresponding author email: tokay@umbc.edu Abstract: As part of the NASA’s Global Precipitation Measurement (GPM) Ground Validation program, Platforms for In-Situ Estimation Rainfall Systems (PIERS) has been fabricated at NASA Wallops Flight Facility, Wallops Island Virginia. PIERS consists of dual tipping bucket gauges, a solar panel, and a cell modem. PIERS+ is an upgraded version with the addition of a laser-optical PARSIVEL disdrometer. Both versions have been distributed at granted institutions under NASA Headquarters program, Increasing Participation of Minority Serving Institutions (https://wallops-prf/gsfc.nasa.gov), across the United States. Between GPM and IPMSI there are 30 PIERS sites, 12 of which are PIERS+. This study uses PIERS+ data for evaluating eight (8) publicly available precipitation products, NOAA’s a) Multi-Radar Multi-Sensor (MRMS), b) Stage IV, c) High Resolution Rapid Refresh (HRRR), d) North American Mesoscale (NAM), e) NASA’s Integrated Multi-satelliE Retrievals for GPM (IMERG), f) Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA2), JAXA’s g) Global Satellite Mapping of Precipitation (GSMaP), and h) European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5). All precipitation products considered here are operational gauge adjusted, but they use different gauge-based products. Several products are based on weather radars (MRMS, Stage IV) or ingest radar data (HRRR). Two satellite-based products (IMERG, GSMaP) and two reanalysis (ERA5, MERRA2) are global, while the remaining four products are constructed over contiguous US and beyond. Majority of the PIERS+ sites are in the Mid-Atlantic region with additional sites in Colorado and Texas. The sites have been operating since late 2023 and early 2024. This study focuses on the rain only events that occurred in 2024. This study presents direct comparisons between the PIERS+ and precipitation product event rainfall. A PARSIVEL disdrometer is used to identify the events and determine the event duration, and maximum rain intensity. A tipping bucket gauges is used to determine the event rain total. The year-long rainfall observations provide more in-depth analysis to evaluate the precipitation products not only for overall performance but also for specific conditions (i.e., short versus long events, light versus heavy rain intensity). The preliminary findings showed that the global products with and without gauge adjustment had drastically different performance, emphasizing the challenge over oceans and gauge sparse regions. This highlights that the performance of products relies on the available of ground observations. PIERS+ is an affordable, low maintenance, easy to operate platforms and can play a key role in supporting precipitation products.
GEMS2-Amethyst: Current Status, Data Products, and Future Directions at Weather Stream Authors: Michael Marques (Weather Stream) Corresponding author email: mmarques@weatherstream.com Abstract: The Global Environment Monitoring System (GEMS) is a planned constellation of passive microwave radiometers deployed on CubeSat platforms to enhance global atmospheric observing capabilities. GEMS is designed to augment the existing microwave observing network, with particular emphasis on improving temporal sampling and monitoring of high-impact weather systems in data-sparse regions. GEMS2-Amethyst (GEMS2A), the first satellite in the second-generation GEMS series, features 24 channels spanning the 118 GHz and the 183 GHz bands. The compact 2U radiometer is scheduled for launch in March 2026. Its channel configuration provides precipitation-sensitive observations that complement existing operational microwave platforms. The follow-on sensor, GEMS2-Beryl (GEMS2B), currently in production and scheduled for launch in 2027, includes improvements in radiometric performance and long-term stability. We present an overview of the GEMS2A mission and its on-orbit status, including calibration activities using observation-minus-background (O–B) characterization and double-difference statistics relative to operational microwave sounders. We then describe the GEMS2A product suite, with emphasis on precipitation-related retrievals. GEMS2A brightness temperatures are processed to Level-2 geophysical products using both the Microwave Integrated Retrieval System (MiRS) physical retrieval framework and a machine learning–based retrieval algorithm (GEMS-MLR). GEMS2A data products through Level-2 will be openly available for non-commercial research and operational evaluation.
Leveraging a Smallsat Constellation and AI for Improved Global Precipitation Understanding Authors: Ethan Nelson (Tomorrow.io); S. Joe Munchak (Tomorrow.io); Teagan King (Tomorrow.io); Forest Cannon (Tomorrow.io); Randy Chase (Tomorrow.io); Arun Chawla; Tomorrow.io Corresponding author email: ethan.nelson@tomorrow.io Abstract: Tomorrow.io is in the midst of building out a low-latency, high-revisit operational weather satellite constellation. To that end, we have launched 2 pathfinder radars (now at end of mission) and 11 microwave sounder smallsats with more sounder launches planned, providing us with an unprecedented revisit rate. The microwave sounder smallsats carry the Tomorrow.io Microwave Sounder (TMS) instrument with heritage in the TROPICS mission. TMS has channels in W-band (1), F-band (7), and G-band (4), and sensors undergo a pre-launch and post-launch calibration process. TMS is equipped with two separate on-board calibration sources–an internal calibration target and a noise diode–alongside a deep space view. Despite its small form factor, internal and independent analyses against GMI, ATMS, and ERA5 have shown TMS to be a well characterized sensor radiometrically and calibration has remained fairly stable over mission lifetimes elapsed to date. In this talk, we will discuss our direct satellite-derived thermodynamic and hydrometeor property products, termed L2A-ATVP and L2A-HRPR respectively. These products use a neural network architecture trained on simulated instrument observations derived from the NASA GEOS high-resolution global circulation model nature run. As expected from a sounding instrument, temperature and water vapor retrieval profiles show good correspondence with multiple external sources (radiosondes, long-duration balloons, GNSS-RO profiles, and global model analyses). Additionally, we will discuss our merged precipitation product, which produces a globally seamless, rapid-refresh analysis and nowcast. This product is generated by fusing Level 1 radiances from all 12 TMS channels with other microwave sensors (via GPM XCAL) and geostationary imagery at 10-minute temporal and ~4 km spatial resolution. Trained on multiple radar networks to ensure robustness across weather regimes, the current operational retrieval delivers substantial gains in both categorical and continuous metrics relative to publicly available satellite precipitation products, including >10% improvements in Heidke Skill Score across light, moderate, and heavy precipitation, and overall supports a globally skillful precipitation nowcast.
A High-Resolution Global Convective-Stratiform Precipitation Type Product from Geostationary Infrared Observations Authors: Daniel Watters (University of Oklahoma); Pierre Kirstetter (University of Oklahoma; NOAA National Severe Storms Laboratory); Jackson Tan (University of Maryland Baltimore County; NASA Goddard Space Flight Center); Veljko Petkovic (University of Maryland); George Huffman (NASA Goddard Space Flight Center) Corresponding author email: daniel.watters@ou.edu Abstract: Classifying precipitation types as convective (non-uniform, intense and localized) or stratiform (homogeneous, weaker and typically widespread) supports extreme event identification, diabatic heating characterization, and understanding of precipitation microphysics. Whilst the Global Precipitation Measurement (GPM) Core Observatory provides instantaneous precipitation type observations from space with the Dual-frequency Precipitation Radar (DPR), sampling limits these observations to localized domains which prevents the DPR’s suitability for global monitoring purposes. To address this gap, we have produced a global convective-stratiform precipitation type product at high spatiotemporal resolution (0.05°, 30-minute) underpinned by geostationary infrared (GEO-IR) brightness temperature observations from 1998. The objective of this Generic Retrievals fOr MEaSUREs Infrared-based precipitation Typology (GROMIT) algorithm is to extend the GPM DPR precipitation type retrievals across the globe for monitoring purposes by fusing its capabilities with the extensive high-resolution global GEO-IR record; this innovative algorithm development work is directly relevant to IPC Session 1.1 ‘NASA Precipitation Measurement Mission Science’. GEO-IR observations, whilst sensitive only to cloud tops, can provide global high-resolution coverage within 60°N-S, thus extending DPR precipitation types beyond its 245-km swath on its non-sun-synchronous orbit. The methodology of GROMIT is to retrieve convective precipitation probabilities across the globe with an AI/ML model trained on DPR precipitation type classifications with coincident GEO-IR brightness temperatures as predictors. Our key results highlight AI/ML advances with GROMIT beyond a classical lookup table, showcase GROMIT’s accuracy relative to input DPR retrievals, and compare the scheme to an independent ground reference over the Contiguous U.S. (CONUS), the Ground Validation Multi-Radar Multi-Sensor (GV-MRMS) product, which is principally underpinned by ground-based radars for determining precipitation types. We also demonstrate the potential for the GROMIT IR-based typology product to contribute to our wider NASA MEaSUREs initiative to produce a global high-resolution precipitation type product which leverages and complements the Integrated Multi-satellitE Retrieval for GPM (IMERG) legacy in merging precipitation retrievals from IR and passive microwave sources.
Passive Microwave Missions at EUMETSAT: Contributions to Precipitation Monitoring Authors: Vasileios Barlakas (EUMETSAT); Christophe Accadia (EUMETSAT); Vinia Mattioli (EUMETSAT); Jörg Ackermann (EUMETSAT); Tim Hewison (EUMETSAT); Francesco De Angelis (EUMETSAT); Sabatino Di Michele (EUMETSAT); Imke Krizek (EUMETSAT); Felix Bosco (EUMETSAT); Robin Ekelund (EUMETSAT); Junshen Lu (CS Group Germany); Davide Ori (CS Group Germany); Farouk Lemmouchi (CS Group Germany) Corresponding author email: vasileios.barlakas@eumetsat.int Abstract: The EUMETSAT Polar System is advancing into its next operational phase with the deployment of its Second Generation (EPS‑SG) satellites. These satellites introduce three new passive microwave radiometers: the Microwave Sounder (MWS) on Metop‑SG‑A, and the Microwave Imager (MWI) together with the Ice Cloud Imager (ICI) on Metop‑SG‑B. In parallel, preparatory activities for the prospective EPS‑Sterna constellation are ongoing. MWS is a cross‑track scanning radiometer equipped with 24 channels spanning 23.8–229 GHz. While its primary role is to deliver temperature and humidity sounding information, several channels are also sensitive to cloud ice and precipitation. In contrast, MWI is designed with a strong focus on cloud and precipitation, offering 18 channels from 18 to 183 GHz, including innovative additions near 50–60 GHz and 118 GHz that improve the detection of light precipitation and snowfall. ICI extends this capability into the millimetre and sub‑millimetre region (183–664 GHz), enabling the first operational observations of cloud‑ice properties and ice‑hydrometeor microphysics, and supplying complementary information for interpreting snowfall signals. EPS Sterna is being considered as a complementary small satellite constellation that will augment the temporal sampling and spatial coverage of the EUMETSAT Polar System. The Sterna platforms will carry the Microwave Radiometer (MWR), a cross track scanning instrument with 19 channels between 50 and 325 GHz. This frequency range supports the retrieval of both liquid and frozen precipitation, in addition to its main objective of providing atmospheric temperature and humidity profile information. The presentation will offer an overview of the EPS passive microwave missions, their scientific objectives, and their relevance for precipitation monitoring. Third party missions—most notably the Copernicus Imaging Microwave Radiometer (CIMR)—will also be included due to their planned contribution to Level 2 precipitation products.
IMERG V08: Key Changes and Early Results Authors: Jackson Tan (University of Maryland Baltimore County & NASA Goddard Space Flight Center); George J. Huffman (NASA Goddard Space Flight Center); Robert Joyce (Science and Technology Corp & NASA Goddard Space Flight Center); Eric J. Nelkin (Science and Technology Corp & NASA Goddard Space Flight Center); David T. Bolvin (Science and Technology Corp & NASA Goddard Space Flight Center) Corresponding author email: jackson.tan@nasa.gov Abstract: The Integrated Multi-satellitE Retrievals for GPM (IMERG) product from the NASA–JAXA Global Precipitation Measurement (GPM) mission provides global satellite precipitation estimates for diverse scientific research and societal applications. An international constellation of low-Earth orbit and geosynchronous orbit satellites enables IMERG to achieve a high resolution of 0.1° every half-hour globally, with three Runs at different latencies to support varied research and application needs. With IMERG V08 expected in the summer 2026, this presentation describes key algorithm changes and provides a preview of early results. Major changes to the IMERG V08 algorithm include: (i) the use of the Goddard Profiling (GPROF) V08 product, which adopts a machine learning approach and demonstrates substantial improvement over the previous version; (ii) the adoption of a new IR precipitation algorithm; (iii) better representation of cold season precipitation to address prior underestimation; (iv) a refined calibration procedure, especially across discontinuities in the constellation record such as the GPM orbit boost; and (v) enhanced handling of different sensors to flexibly accommodate their anticipated levels of quality. Collectively, these advances are expected to improve the skill of IMERG estimates, address several known deficiencies in V07, ensure greater consistency across the entire data record, and expand IMERG’s capability to use shorter-lived small satellites. As IMERG V08 is currently planned to be the final major upgrade before the end of the GPM mission, these improvements position IMERG to continue delivering high-quality precipitation estimates to the scientific community for research and application endeavors.
Transformer-Based Modeling of Global Microwave Land-Surface Emissivity Using GPM Authors: Sarah Ringerud (NASA GSFC); Fraser King (University of Wisconsin-Madison) Corresponding author email: sarah.e.ringerud@nasa.gov Abstract: Global land-surface microwave emissivity estimates are critical for data assimilation, numerical weather prediction, and quantitative precipitation estimation, yet explicit calculations at multiple frequencies remains problematic. We introduce a machine learning model that converts 10.65 GHz H-polarization emissivity retrieved under clear-sky conditions into emissivity for the 12 higher-frequency channels of the GPM Microwave Imager (GMI). The model is a feature-tokenizer transformer that ingests ancillary surface variables such as skin temperature, leaf area index, soil texture, and soil moisture. Treating each variable as an individual token lets the transformer direct attention between features and learn nonlinear, spatiotemporally consistent relationships across five years of data [2015-2019]. The network reproduces channel-specific emissivities with R 2 > 0.80 and mean absolute percentage error < 1%, and it simulates topof-atmosphere brightness temperatures with R 2 = 0.77 − 0.95 and mean biases
Machine Learning for Passive Microwave Snowfall Regime Classification: a Global Analysis Authors: Lisa Milani (ESSIC - University of Maryland / NASA Goddard Space Flight Center); Veljko Petkovic (ESSIC - University of Maryland) Corresponding author email: lmilani@umd.edu Abstract: Numerous studies have emphasized the importance of snowfall regime classification for accurate snowfall rate retrieval from Passive Microwave (PMW) observations. Whether precipitation algorithms utilize a-priori information or training datasets, creating comprehensive and representative datasets is crucial for effective snowfall detection and quantification from satellite-based sensors. This research investigates snowfall retrievals in the Goddard PROFiling (GPROF) algorithm, which serves as the PMW precipitation product for the Global Precipitation Measurement (GPM) mission. This study utilizes a combined CloudSat-GPM dataset to generate training data for an eXtreme Gradient Boost (XGB) model. The model establishes relationships between GPM Microwave Imager (GMI) brightness temperatures and Cloud Profiling Radar (CPR)-derived snowfall regimes, classifying observed scenes into four categories: 'not snowing', 'shallow convective', 'deep stratiform', or 'other' snowfall types. The Machine Learning (ML) approach is crucial for identifying and interpreting the complex yet robust connections between atmospheric PMW signals and surface snowfall characteristics. The ML classifier is trained using CloudSat's classification framework, which integrates snow profiles and cloud categorization methodologies, before being implemented in GPROF operations. The presentation will provide an extensive global comparison of snowfall regime classifications, evaluating results derived from CloudSat data against those based exclusively on PMW observations.
High-Latitude and Oceanic Precipitation Development and Assessments in Supporting of IMERG and GPCP Authors: Ali Behrangi (University of Arizona); Yang Song (University of Arizona); Kingsley Kumah (University of Arizona); Omid Zandi (University of Arizona) Corresponding author email: behrangi@arizona.edu Abstract: Accurate estimation of precipitation at high latitudes remains one of the most persistent challenges in satellite-based precipitation retrievals, particularly over regions dominated by frozen surfaces such as seasonally and permanently ice-covered oceans, glaciers, and ice sheets, which are critical for climate, cryosphere, and hydrologic studies yet remain among the least constrained components of the global water cycle. These challenges are compounded by historical limitations in satellite observing systems: although IMERG is being extended back to the late 1998 and GPCP provides records to the early 1980s, high-latitude coverage is constrained by the absence of geostationary infrared observations and the limited availability of passive microwave sensors (primarily SSM/I), motivating the need for alternative, long-term consistent observations. This presentation has three components: (1) an initial assessment of IMERG and GPCP precipitation estimates over frozen surfaces, including sea ice, glaciers, and Antarctica, using a multidisciplinary evaluation framework that incorporates independent satellite and in situ datasets such as ICESat, CryoSat-2, GRACE, CloudSat, surface accumulation measurements, and regional observations, and includes comparisons to the widely used ERA5 reanalysis; , (2) precipitation assessment over the oceans using combination of various observational resources including Buoys, Atolls, and Passive Aquatic Listeners (PALs); , and (3) the development of the University of Arizona High-latitude Infrared-based Precipitation Analysis Version 2 (UA-HIPA), which provides consistent long-term precipitation estimates from AVHRR back to 1981 and improves sampling, accuracy, spatial resolution, and temporal consistency of satellite-based precipitation. This product supplements existing datasets, replacing AIRS and improve temporal homogeneity in GPCP and potential use in extended IMERG. The results highlight persistent deficiencies in current satellite precipitation products over frozen regions while also documenting notable improvements in recent product versions, particularly in their representation of precipitation amount, spatial patterns, and data record consistency. Together, the datasets and methods presented provide valuable baseline benchmarks for evaluating current IMERG and GPCP products and offer guidance for the development of next-generation precipitation retrievals optimized for cold and ice-covered environments.
Session summary
Back to full program.