Science AI Journal
Earth and Planetary Sciences · Research topic

Open research questions in Meteorological Phenomena and Simulations

35 unresolved questions extracted from the limitations and future-work sections of 491 Meteorological Phenomena and Simulations papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • Although the precise mechanism remains uncertain, the consistent presence of this pattern across multiple independent prediction systems indicates that it is an operational artifact of the current EOP evaluation chain rather than a genuine geophysical signal.

    Celestial pole offsets prediction with multiple data sources: a data-driven approach · 2026 · DOI

  • Glasgow1 3 345 visiting us in the hospital, say, or buying us a drink when we get laid off from work— but if they soon move on, then the tie was weaker than the truest bonds of love, and if their love was real in one sense, it also was lacking in another.

    The Virtues of the All-Weather Fan · 2026 · DOI

  • Evapotranspiration (ET) underpins water, energy, and carbon cycling, yet remains among the least observed hydrologic fluxes, creating a persistent paradox: hydrology is advancing rapidly through artificial intelligence and data-driven methods while its observational foundation remains sparse and fragmented.

    Toward Fit-for-Purpose Evapotranspiration Observations · 2026 · DOI

  • Particle number size distribution (PNSD) is fundamental for characterizing aerosols and quantifying aerosol–cloud interactions, while simulations remain uncertain due to strong spatiotemporal variability during air-parcel transport.

    Reconstructing particle number size distributions via aerosol history–informed multiple deep learning models · 2026 · DOI

  • Abstract The boundary-layer jet (BLJ) over the northern South China Sea exerts a strong influence on heavy rainfall along the South China coast, yet its predictability across different modeling frameworks remains insufficiently understood.

    Complementary error structures of AI and numerical models in forecasting boundary-layer jets over the South China Sea · 2026 · DOI

  • In this work, we make the case for the use of EBMs in combination with physics-informed feature engineering to yield interpretable ML algorithms for certain meteorological applications. This approach has several advantages, including, but not limited to (1) the ability to fully understand the strategies used by the ML algorithm when making identifications, exposing potential failure modes; (2) the opportunity to adjust its strategies to more closely match the strategies expected based on domain knowledge; and (3) the ability to develop a generalizable model from just a few data samples, or, if data for a similar task is available, utilizing those instead in a way analogous to transfer learning. We have illustrated how these advantageous aspects of the EBM framework can aid in the approach of detecting OT locations from satellite imagery. We emphasize, however, that this application of EBMs was only possible due to feature engineering that first simplified the task at hand. Nevertheless, we believe that this method has the potential to be used in a wide variety of meteorological applications. At first sight, the identification and tuning of EBM model strategies, which is the part of the EBM development process illustrated in Section 4, may appear to be a lot of extra work, especially when compared to the hands-off training procedure of a comparable neural network model. One should keep in mind, however, that for a neural network model, the identification of strategies should come as a separate step after its training is completed, e.g., using XAI methods, but that step is often neglected, since it is nearly impossible to detect most of its strategies anyway. EBMs should thus not be dismissed for enabling and, in fact, requiring this important step during their development process. In other words, this step is simply the price to pay to obtain an interpretable model. For the application of identifying OTs, the next step in this research should be the creation of a large hand-labeled data set that identifies OTs in GOES visible imagery. Creating such a labeled data set is a larger effort, but it is needed to fully evaluate how well the EBM model matches human labeling. More generally, much work remains to further explore the use of EBMs in the field of meteorology in terms of both identifying the most suitable applications and developing a larger range of engineered features—endeavors we hope will serve to further improve the performance of these models. 38 Acknowledgments. This material is based upon work supported by the National Science Foun- dation under AI Institute Grant No. 2019758 and CAIG grant No. 2425923; and by the Machine Learning Strategic Initiative at the Cooperative Institute for Research in the Atmosphere at Colorado State University. Data availability statement. The dataset and python code used to train, validate, and test the EBM model will be made publicly available before publication.

    Knowledge-Guided Machine Learning: Illustrating the use of Explainable Boosting Machines to Identify Overshooting Tops in Satellite Imagery · 2026 · DOI

  • As noted in Sect. 2.1.2, there is some ambiguity regarding the averaging time of ground-truth measurement wind speeds for the present case study; while averaging time is generally recommended to be ten minutes, we cannot be certain that this standard is applied for all ground-truth data. As different averaging times may be desired for different remote measurement techniques, the applicability of the derived case-study model might reasonably be questioned. However, on the scale of a measurement, the averaging time would only affect the dispersion in the derived region-averaged wind speed error model, the bias would likely be negligibly affected. If the averaging time of the modelled data were smaller than required by the measurement technique, then dispersion of the region-averaged wind speed error model would be overestimated and vice versa. Ideally, higher resolution ground-truth wind speed data could enable investigation into the significance of this limitation, but these data are not available to our knowledge. One drawback of the combined models is that, relative to simpler models that do not consider spatiotemporal autocorrelation of wind speed errors, computation time and memory usage for random sampling within a Monte Carlo framework (see Sect. 2.3) may be prohibitively large. Random sampling of the multivariate normal distribution requires a decomposition of the n × n covariance matrix (e.g., Gentle, 2009); the fastest method to do this using the NumPy software package (Harris et al., 2020), the Cholesky decomposition, is of order n3 (e.g., Watkins, 2002). Thus, as n increases, computational time might become prohibitively large. Moreover, even when the Cholesky decomposition is performed “in-place”, memory usage scales quadratically, which may cause out-ofmemory conditions on desktop computers. This occurred as part of the analysis in Sect. 3.3 (see rightmost data point in Fig. 5a), where the simulated survey with 162 600 measurements caused memory overflow. Practically, however, memory overflow issues like this can be solved by leveraging modern high-performance scientific computing resources. To model spatiotemporal autocorrelation with the Gaussian copula through the semivariogram, we assumed secondorder stationarity of the transformed data, z (refer to Sect. 2.3). In general, we do not expect wind speed errors for any one weather station and time period to be stationary since, for example, microscale variations in topog- Atmos. Meas. Tech., 19, 3761–3780, 2026 https://doi.org/10.5194/amt-19-3761-2026 B. M. Conrad and M. R. Johnson: Accounting for spatiotemporally correlated errors in wind speed 3775 Figure 7. Comparisons of optimized models for two regions (British Columbia (BC) and North Dakota (ND)) and NWP models (NAM12 and HRRR): (a) bias and standard deviation of the region-averaged wind speed error model, (b) spatial correlogram, and (c) temporal correlogram.

    Accounting for spatiotemporally correlated errors in wind speed for remote surveys of methane emissions · 2026 · DOI

  • It should be noted that the current study is limited by the extreme scarcity of geoeffective events (116 in total), and the insufficient number of positive samples has to some extent affected the statistical stability of evaluation metrics and further improvement of the model.

    Intelligent prediction model for geoeffectiveness of coronal mass ejections · 2026 · DOI

  • The tobac cloud tracking framework has not been extensively applied to ICON model simulations of Southern Scandinavian convection to systematically quantify cloud lifecycle properties (initiation, growth, dissipation rates) and their sensitivity to microphysical parameterizations during different synoptic regimes.

    Convection over Southern Scandinavia: a Modeling Perspective · 2026 · DOI

  • Stratospheric hydration measurements from convective events over Southern Scandinavia are limited to the TPEx Campaign dataset; systematic comparison of hydration and ice microphysics observations across multiple convective seasons and diverse atmospheric conditions over the region is needed to validate model predictions.

    Convection over Southern Scandinavia: a Modeling Perspective · 2026 · DOI

  • The ice formation pathways in convective overshoots over Southern Scandinavia have not been systematically characterized using the two-moment multi-class cloud microphysics scheme, limiting understanding of how different ice nucleation mechanisms (deposition freezing, contact freezing, immersion freezing) operate during convection in this region.

    Convection over Southern Scandinavia: a Modeling Perspective · 2026 · DOI

  • The quantitative relationship between atmospheric stability indices (CAPE, CIN, KI, TTI) and the intensity of downbursts in this region needs to be established through comparative analysis across multiple events.

    Meteorological Analysis of Strong Wind-Producing Clouds: A Case Study of Klaten, 18 November 2024 · 2026 · DOI

  • To efficiently scale to higher resolution or to move towards computational parity with GenCast-Perturbed and similar models, distillation49 and other efficiency techniques should be explored.

    Probabilistic weather forecasting with machine learning · 2024 · DOI

  • Abstract Convective momentum transport (CMT) has mostly been studied for deep convection, whereas little is known about its characteristics and importance in shallow convection.

    Uncertainties related to the representation of momentum transport in shallow convection · 2017 · DOI

  • These limitations stem from overreliance on short-term patterns, which are insufficient to capture chaotic weather dynamics, especially under partial observations.

    TSSM: Triaxial State Space Model for Global Station Weather Forecasting with Temporal-Variable-Historical Modeling · 2026

  • Future research should explore whether the scale- dependent meridional contrast identified here persists globally or reflects unique regional dynamics linked to orographic, jet structure, and gravity wave activities.

    Increasing trend of vertical wind shear revealed by high-resolution radiosonde data over the United States · 2026 · DOI

  • We demonstrate that achieving stable online coupling requires enforcing physical constraints and careful dataset curation, and that strong offline performance alone is insufficient.

    From stable online coupling to decade-long climate simulations: A machine learning parameterization for cloud microphysics in ICON · 2026

  • Yet this relationship remains poorly quantified, especially at instantaneous timescales where observational uncertainty and inherent variability pose challenges.

    Data-driven Modeling of Stratiform and Convective Rain Area · 2026 · DOI

  • Subsequent efforts should focus on enhancing the method's robustness and applicability by expanding observational samples and introducing independent validation data.

    Optimization Study of Low-Altitude Turbulence Intensity Modeling Based On TKE-XGBoost · 2026 · DOI

  • Model a relies heavily on observational data, requiring synchronized, high-quality observations from both wind profile radar and microwave radiometer, which limits its applicability.

    Optimization Study of Low-Altitude Turbulence Intensity Modeling Based On TKE-XGBoost · 2026 · DOI

  • The study relies on radar reflectivity and radial velocity products; validation with surface wind speed measurements and damage assessments would enhance the characterization of strong wind impacts.

    Meteorological Analysis of Strong Wind-Producing Clouds: A Case Study of Klaten, 18 November 2024 · 2026 · DOI

  • We conclude that new model couplings should make use of integration tests as meteorological evaluations by themselves are insufficient, given that errors are difficult to attribute because of the interplay between observational errors and multiple parameterization schemes (e.

    WRF‐TEB: Implementation and Evaluation of the Coupled Weather Research and Forecasting (WRF) and Town Energy Balance (TEB) Model · 2020 · DOI

  • Although our simple models do not perform better than an operational weather model, machine learning warrants further exploration as a weather forecasting tool; in particular, the potential efficiency of CNNs might make them attractive for ensemble forecasting.

    Can Machines Learn to Predict Weather? Using Deep Learning to Predict Gridded 500‐hPa Geopotential Height From Historical Weather Data · 2019 · DOI

Most-cited papers in Meteorological Phenomena and Simulations

Most recent work

Find a gap in your own Meteorological Phenomena and Simulations sub-topic

This page shows what the Meteorological Phenomena and Simulations literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Earth and Planetary Sciences

Atmospheric chemistry and aerosols
135 open questions
Archaeology and ancient environmental studies
96 open questions
Marine and coastal ecosystems
83 open questions
Oceanographic and Atmospheric Processes
69 open questions
earthquake and tectonic studies
53 open questions
Marine and coastal plant biology
52 open questions
Paleontology and Stratigraphy of Fossils
51 open questions
Atmospheric Ozone and Climate
48 open questions
35 open questions have been extracted from the limitations and future-work passages of 491 Meteorological Phenomena and Simulations papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.