Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Do Echo Top Heights Improve Deep Learning Nowcasts?

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read Adding echo top height to radar nowcasts helps only for the lightest rain.

desk verdict A careful, well-scoped empirical null result on adding echo top height to a deterministic nowcaster; the negative conclusion is broadly supported, but the filtered single-year test set and missing uncertainty estimates keep it from being the last word. read the letter →

arxiv 2507.00845 v1 pith:Y3X43SRS submitted 2025-07-01 cs.CV cs.LG

classification cs.CVcs.LG
keywords precipitationnowcastingechotopheightradarreflectivitydeeplearning3DU-NetfractionsskillscoreNetherlands
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks whether adding echo top height (ETH) — the highest altitude at which radar detects precipitation — as an extra input channel improves deep-learning nowcasts of rainfall. It trains a 3D U-Net on Dutch radar data with and without ETH, comparing eight model instances per configuration. The answer is mostly no: ETH improves skill at the 0.1 mm/h threshold, but at 1 mm/h and above the scores are equal or worse, and the ETH models systematically underestimate rainfall. The paper concludes that ETH carries physically relevant information but, in this proof-of-concept setup, does not deliver convincing value for nowcasting, and it warns that auxiliary variables can add bias and variance rather than skill.

What carries the argument

The central object is echo top height (ETH), a two-dimensional field giving, at each grid point, the maximum altitude at which the radar detects reflectivity above a 7 dBZ threshold; it is a compact stand-in for the vertical extent of precipitation. The architecture that carries the argument is a single-pass 3D U-Net (an encoder–decoder convolutional network with skip connections) that treats time as the third dimension and takes radar reflectivity and ETH as separate input channels, predicting 18 future frames in one forward pass to avoid recursive bias accumulation. The evaluation machinery is a set of verification scores — MSE, MAE and mean error, precision/recall/ETS at four thresholds, and Fractions Skill Score at five thresholds and three spatial scales — chosen so that any genuine reduction in nowcast uncertainty from ETH should show up as less blurring and better spatial skill.

What would settle it

Evaluate the same with- and without-ETH comparison on a multi-year test set that includes widespread moderate-rain events, or retrain on a sample not truncated to the top 1000 reflectivity frames per year; if ETH models then beat no-ETH models on Equitable Threat Score at 1 mm/h or 2.5 mm/h without a more negative mean error, the paper's negative verdict would be overturned.

Watch

Extended reading notes

Core claim

The central claim is that, on this dataset and architecture, echo top height does not provide convincing proof-of-concept value for precipitation nowcasting. Models given ETH as a separate input channel achieve higher Fractions Skill Score and Equitable Threat Score only at the lowest rain-rate threshold (0.1 mm/h) and at short lead times; at thresholds of 1 mm/h and above, the two configurations score similarly, with the no-ETH models slightly ahead on ETS. The ETH models also show larger run-to-run variance and a more negative mean error, meaning they tend to predict less rain than observed. Three case studies show the effect is event-dependent: ETH helps when the storm has high echo tops, but hurts or confuses the model when rainfall is intense yet echo tops are low, and the added input can amplify error variance even when average error improves.

Load-bearing premise

The result assumes that the curated sample — roughly the top 1000 events per year by summed reflectivity, restricted to the post-2016 radar period, cropped to where 7 km echo tops are detectable, with 2022 as the only test year — fairly represents the range of rainfall intensities where ETH could matter.

Editorial extensions

If this is right

  • Adding ETH as an extra channel is technically feasible in a 3D U-Net and slightly improves detection of very light rain (0.1 mm/h), especially at short lead times.
  • At rain rates of 1 mm/h and above, ETH does not improve Fractions Skill Score or Equitable Threat Score; at higher thresholds the no-ETH models are equal or better.
  • ETH inputs systematically increase the negative bias (mean error), so flood- and warning-oriented applications would see more underestimated intensities.
  • The effect of ETH is case-dependent: it can preserve intensity and reduce blurring when echo tops are high, but can degrade forecasts when intense rain is accompanied by low echo tops.
  • A deterministic U-Net with a pointwise loss is a workable diagnostic benchmark for testing auxiliary variables, but the paper argues that generative or probabilistic approaches and better ETH preprocessing may be needed to unlock ETH's potential.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural next test is to stratify the test set by observed echo top height; the case studies suggest ETH helps in high-ETH convection, so a conditioned evaluation might show positive skill that the aggregate metrics hide.
  • The negative bias may be a side effect of the pointwise MSE loss combined with an extra input that correlates with but does not uniquely determine surface rain; an architecture that decouples intensity from occurrence, or a loss that penalizes underestimation asymmetrically, might change the balance.
  • Because the radar system changed in 2016 and the test year is a single year, the result is a snapshot of one country's radar regime; transfer to other climates or radar networks could plausibly reverse the verdict.
  • ETH artifacts from the radar sweep geometry (circular bands) are visible even in the cropped domain; cleaning these artifacts by masking unreliable pixels is a cheap preprocessing fix that could alter the comparison.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper investigates whether adding echo top height (ETH) as an auxiliary input channel to a deterministic 3D U-Net improves radar-based precipitation nowcasting in the Netherlands. The authors compare eight models trained with ETH and eight without, using pixel-wise metrics (MSE, MAE, ME), threshold-based metrics (precision, recall, ETS), the Fractions Skill Score (FSS), and three qualitative case studies. The central finding is that ETH improves skill only at the lowest rain-rate threshold (0.1 mm/h), does not improve high-intensity nowcasts, and introduces an additional negative bias, leading the authors to conclude that ETH does not provide convincing proof-of-concept evidence for nowcasting. The paper includes code links, animations, and a candid discussion of limitations.

Significance. If the result holds, it is a useful and reproducible negative result for the nowcasting community: it suggests that naively adding ETH as an input channel is insufficient, and it provides a benchmark for future work on auxiliary radar-derived variables. The paper deserves credit for training multiple models under different seeds and data splits, for reporting multi-scale FSS along with threshold metrics, and for making code and visualizations publicly available. The negative-bias finding and the case-specific behavior are also of operational interest. The main limitations, as discussed in the report, concern the dataset construction and the lack of uncertainty quantification, which weaken the strength of the null conclusion.

major comments (3)
  1. [§4.1] The dataset description is internally inconsistent about the training sample size: selecting the top-1000 observations per year for 2008–2022 gives 15,000 starting points, but the text then restricts training to data from October 2016 onward, which would leave roughly 6,000 starting points. Please clarify whether the 15,000 figure includes pre-2016 data and how the restriction was applied. This matters because the composition of the training set and the representativeness of the 2022-only test set are load-bearing for the paper's null claim.
  2. [§5.1 and §5.2] The central null claim rests on comparisons of MSE, MAE, ME, and ETS between the two model groups, but no statistical significance tests or confidence intervals are reported. Given that the paper concludes that ETH 'introduced an additional negative bias' and is 'detrimental' at higher intensities, the authors should provide paired bootstrap or permutation tests across the eight model pairs (or across events) for the key metrics. Without such tests, the reported mean differences—many of which are small relative to the standard deviations in Figures 5 and 7—cannot be distinguished from sampling noise.
  3. [§5.3] The FSS matrices in Figure 8 show essentially zero scores at thresholds of 5 and 10 mm/h for both configurations at all lead times. The statement that ETH does not improve spatial predictive skill for heavier rainfall is therefore based on a near-floor comparison in which the observed fields contain very few positive pixels; this does not robustly support the conclusion. Please either restrict the high-threshold FSS analysis to events with meaningful rain coverage, supplement it with a metric that remains interpretable at low base rates, or explicitly quantify the uncertainty on the zero and near-zero differences.
minor comments (5)
  1. [§5] In the first paragraph of Section 5, 'Ta gain additional insight' should read 'To gain additional insight'.
  2. [§3.1] In the final paragraph of Section 3.1, 'it it is possible' should read 'it is possible'.
  3. [§4] In the first paragraph of Section 4, 'for it's simplicity' should read 'for its simplicity'.
  4. [§5.4] In the paragraph introducing the case studies, 'with a without ETH' should read 'with and without ETH', and 'ploted' should be 'plotted'.
  5. [Figures 5–7] The green/red color scheme used for the two model groups is difficult to distinguish for color-blind readers; consider using more separable colors or line styles.

Circularity Check

0 steps flagged · score 0.0 of 10

Empirical ablation study with no derivation chain; all skill differences are measured outcomes, so no circularity.

full rationale

This paper is an empirical comparison, not a derivation: the central question is whether adding echo top height (ETH) as an extra input channel to a 3D U-Net improves precipitation nowcasts. The claimed results—modest gains at 0.1 mm/h, similar MSE, more negative ME, and lower ETS at higher thresholds—are directly measured on a held-out 2022 test set, not derived from fitted constants or from the definition of ETH. The model is trained with and without ETH under identical splits and seeds, so the comparison is self-contained and apples-to-apples. Section 4.1's dataset construction (top-1000 reflectivity-ranked events, post-2016 data, cropped domain) is a sampling choice that could affect external validity or statistical power, but it does not make the ETH comparison circular; both model groups see the same samples. Section 3's scatter plots describing the ETH-reflectivity relationship are descriptive, not a derived prediction. The only self-citations are reference [16] (earlier volumetric nowcasting work) and reference [15] (a Delft thesis supervised by a co-author, used for clutter removal); both are motivational or methodological and neither carries the paper's central claim. There is no fitted parameter renamed as a prediction, no uniqueness theorem imported from the authors, and no ansatz smuggled in via citation. Consequently, there is no equation-level or definition-level circularity, and the paper's negative conclusion stands as a measured empirical outcome rather than a construction of its inputs.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The main load-bearing choices are dataset selection, the fixed Z-R relation, and the architecture family; these are documented in Section 4 and condition the scope of the conclusion.

free parameters (5)
  • Annual top-N event selection = 1000 observations per year
    Sequences start from the top 1000 observations per year by summed squared reflectivity, about 1.5% of all data; this skews the training and test distribution toward intense events.
  • Spatial crop extent = 336 x 272 km
    Central subset chosen so echo tops of at least 7 km are detectable everywhere; the crop shapes the evaluation domain.
  • ETH reflectivity threshold = 7 dBZ
    Echo top height is defined by the KNMI product as the highest altitude with reflectivity above 7 dBZ; the authors did not vary this threshold.
  • Rain-rate thresholds for metrics = 0.1, 1, 2.5, 5, 10 mm/h
    The conclusions about low versus high intensity skill depend on these hand-chosen thresholds.
  • Case-study model selection = best FSS at 2.5 mm/h, 16 km, 30 min
    One model per group was selected for qualitative figures based on this criterion, which affects the case study narrative.
assumptions (4)
  • domain assumption Post-October 2016 ETH data are internally consistent after the KNMI radar replacement.
    Section 4.1 excludes earlier data because of a distributional shift; the comparison assumes the later ETH product is comparable across years.
  • domain assumption Marshall-Palmer Z-R relation converts reflectivity to rainfall rate.
    Section 4.1 applies Z=200R^1.6; all intensity targets and error metrics inherit the biases of this fixed relation.
  • ad hoc to paper The top-1000-per-year subset represents the rain events a nowcaster must handle.
    Section 4.1 selects intense events to counter clear-sky bias; representativeness for operational conditions is assumed, not demonstrated.
  • domain assumption A deterministic 3D U-Net with MSE loss is an adequate testbed for detecting ETH value.
    The conclusion of limited ETH value is conditional on this architecture family, as the authors note in Section 6 when discussing generative or probabilistic alternatives.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Do Echo Top Heights Improve Deep Learning Nowcasts?." pith.science (2026). https://pith.science/paper/Y3X43SRS

@misc{pith2026250700845,
  author       = {Pith},
  title        = {Pith review of: Do Echo Top Heights Improve Deep Learning Nowcasts?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y3X43SRS}},
  note         = {Machine review of arXiv:2507.00845}
}
read the original abstract

Precipitation nowcasting -- the short-term prediction of rainfall using recent radar observations -- is critical for weather-sensitive sectors such as transportation, agriculture, and disaster mitigation. While recent deep learning models have shown promise in improving nowcasting skill, most approaches rely solely on 2D radar reflectivity fields, discarding valuable vertical information available in the full 3D radar volume. In this work, we explore the use of Echo Top Height (ETH), a 2D projection indicating the maximum altitude of radar reflectivity above a given threshold, as an auxiliary input variable for deep learning-based nowcasting. We examine the relationship between ETH and radar reflectivity, confirming its relevance for predicting rainfall intensity. We implement a single-pass 3D U-Net that processes both the radar reflectivity and ETH as separate input channels. While our models are able to leverage ETH to improve skill at low rain-rate thresholds, results are inconsistent at higher intensities and the models with ETH systematically underestimate precipitation intensity. Three case studies are used to illustrate how ETH can help in some cases, but also confuse the models and increase the error variance. Nonetheless, the study serves as a foundation for critically assessing the potential contribution of additional variables to nowcasting performance.

Figures

Figures reproduced from arXiv: 2507.00845 by the authors.

Figure 1
Figure 1. This Example shows a single vertical slice of the full 3-dimensional radar ob￾servation. The radar station is located the the bottom left corner. Each ray represents data captured by a different radar sweep. However, working with a full 3D volume can be very computationally ex￾pensive. Also, if we consider the time steps as a separate data dimension, we effectively arrive at 4D data volumes that the model needs to p… view at source ↗
Figure 2
Figure 2. Side-by-side comparison of a single observation from the dataset (February 20th, 2022), showing the spatial distribution of echo top height (left) and corresponding rainfall rate (right). This illustrates the relationship between storm structure aloft and precipitation intensity at 1500 meters above ground. In this case, a reflectivity threshold of 7 dBZ was used to calculate echo top height. To illustrate the relat… view at source ↗
Figure 3
Figure 3. The distribution of radar reflectivity with their corresponding echo top height observations for three different days with strong precipitation. The plots suggest a roughly proportional relationship between the two variables. The echo top data also clearly contains artifacts corresponding to the maximum observable echo top height of various radar sweeps, visible as density cutoffs at various heights. 3.1 Related Use… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Maximum observed echo top height per pixel across all training and test set samples (October 2016 to December 2022). The highlighted area indicates the spatial extent selected for model training, ensuring vertical coverage of at least 7 km. For the model training and e…
Figure 5
Figure 5. Figure 5: The means and standard deviations of models with and without echo top height input for three metrics – mean squared error (MSE), mean absolute error (MAE), and mean error (ME) or bias. Each group consists of 8 separate models trained of with different test-validation s…
Figure 6
Figure 6. Figure 6: Mean difference of models with and without echo top height input for three metrics – mean squared error (MSE), mean absolute error (MAE), and mean error (ME) or bias – for each test sample based on their maximum and spatially averaged precipitation rate. The MSE and MA…
Figure 7
Figure 7. Figure 7: Means and standard deviations of models with and without echo top height for three metrics – precision, recall, and equitable threat score (ETS). Evaluations are performed at four rainfall intensity thresholds: 0.1, 1, 2.5, and 5 mm/h. Each group consists of 8 separate…
Figure 8
Figure 8. Figure 8: The matrices show the mean Fraction Skill Score (FSS) for models trained with and without echo top height (ETH) input. Evaluations are performed at five rainfall intensity thresholds (0.1, 1, 2.5, 5, and 10 mm/h) and three spatial tolerance scales (1, 4, and 16 km), ac…
Figure 9
Figure 9. Figure 9: Series of nowcasts generated by the models with and without ETH for a precip￾itation event on February 20th, 2022, at 20:00. For reference, the corresponding ETH observation and ground truth rainfall rate observation are also shown. Each panel presents the predicted or…
Figure 10
Figure 10. Figure 10: Series of nowcasts generated by the models with and without ETH for a precipitation event on May 19th, 2022, at 03:00. For reference, the corresponding ETH observation and ground truth rainfall rate observation are also shown. Each panel presents the predicted or obse…
Figure 11
Figure 11. Figure 11: Series of nowcasts generated by the models with and without ETH for a precipitation event on August 17th, 2022, at 15:00. For reference, the corresponding ETH observation and ground truth rainfall rate observation are also shown. Each panel presents the predicted or o…

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. York's Cavity Formalism and Quantum Modified Thermodynamics of (2+1)D Black Holes

    gr-qc 2025-06 reject novelty 4.0 of 10

    The paper claims Barrow entropy corrections reshape BTZ black hole thermodynamics in a cavity, but the free energy analysis rests on a wrong extrinsic curvature term and an incorrect zero-crossing interpretation.

Reference graph

Works this paper leans on

28 extracted references · 26 canonical work pages · cited by 1 Pith paper

  1. [1]

    0: a convolutional neural net- work for radar-based precipitation nowcasting

    Ayzel, G., Scheffer, T., Heistermann, M.: Rainnet v1. 0: a convolutional neural net- work for radar-based precipitation nowcasting. Geoscientific Model Development 13(6), 2631–2644 (2020)

  2. [2]

    Quarterly Journal of the Royal Meteorological Society: A journal of the atmo- spheric sciences, applied meteorology and physical oceanography132(620), 2127– 2155 (2006)

    Bowler, N.E., Pierce, C.E., Seed, A.W.: Steps: A probabilistic precipitation fore- casting scheme which merges an extrapolation nowcast with downscaled nwp. Quarterly Journal of the Royal Meteorological Society: A journal of the atmo- spheric sciences, applied meteorology and physical oceanography132(620), 2127– 2155 (2006)

  3. [3]

    https://www.tudelft.nl/dhpc/ark:/44463/DelftBluePhase2 (2024)

    Delft High Performance Computing Centre (DHPC): DelftBlue Supercomputer (Phase 2). https://www.tudelft.nl/dhpc/ark:/44463/DelftBluePhase2 (2024)

  4. [4]

    Journal of atmospheric and oceanic tech- nology 10(6), 785–797 (1993)

    Dixon, M., Wiener, G.: Titan: Thunderstorm identification, tracking, analysis, and nowcasting—a radar-based methodology. Journal of atmospheric and oceanic tech- nology 10(6), 785–797 (1993)

  5. [5]

    Nature communications13(1), 1–10 (2022)

    Espeholt,L.,Agrawal,S.,Sønderby,C.,Kumar,M.,Heek,J.,Bromberg,C.,Gazen, C., Carver, R., Andrychowicz, M., Hickey, J., et al.: Deep learning for twelve hour precipitation forecasts. Nature communications13(1), 1–10 (2022)

  6. [6]

    arXiv preprint arXiv:2402.04290 (2024)

    Gong, J., Bai, L., Ye, P., Xu, W., Liu, N., Dai, J., Yang, X., Ouyang, W.: Cascast: Skillful high-resolution precipitation nowcasting via cascaded modelling. arXiv preprint arXiv:2402.04290 (2024)

  7. [7]

    Weather and Forecasting30(5), 1201–1217 (2015)

    Hwang, Y., Clark, A.J., Lakshmanan, V., Koch, S.E.: Improved nowcasts by blend- ing extrapolation and model forecasts. Weather and Forecasting30(5), 1201–1217 (2015)

  8. [8]

    Weather and Forecasting 24(5), 1297 – 1308 (2009)

    Keil, C., Craig, G.C.: A displacement and amplitude score employing an op- tical flow technique. Weather and Forecasting 24(5), 1297 – 1308 (2009). https://doi.org/10.1175/2009WAF2222247.1, https://journals.ametsoc.org/view/ journals/wefo/24/5/2009waf2222247_1.xml

Show all 28 references
  1. [9]

    https://dataplatform

    Koninklijk Nederlands Meteorologisch Instituut (KNMI): Precipitation - radar 5 minute echo top height composites over the netherlands. https://dataplatform. knmi.nl/dataset/radar-echotopheight-5min-1-0 (2024). https://doi.org/urn:xkdc: ds:nl.knmi::radar_echotopheight_5min/1.0/...

  2. [10]

    knmi.nl/dataset/radar-tar-refl-composites-1-0 (2024)

    Koninklijk Nederlands Meteorologisch Instituut (KNMI): Precipitation - radar 5 minutereflectivitycompositesoverthenetherlands-archive.https://dataplatform. knmi.nl/dataset/radar-tar-refl-composites-1-0 (2024). https://doi.org/urn:xkdc: ds:nl.knmi::radar_tar_refl_composites/1.0...

  3. [11]

    Journal of applied meteorology and climatology 34(6), 1286–1300 (1995) Do Echo Top Heights Improve Deep Learning Nowcasts? 25

    Li, L., Schmid, W., Joss, J.: Nowcasting of motion and growth of precipitation with radar over a complex orography. Journal of applied meteorology and climatology 34(6), 1286–1300 (1995) Do Echo Top Heights Improve Deep Learning Nowcasts? 25

  4. [12]

    Hydrology and Earth System Sciences17(8), 3095–3110 (2013)

    Liu, J., Bray, M., Han, D.: A study on wrf radar data assimilation for hydrological rainfall prediction. Hydrology and Earth System Sciences17(8), 3095–3110 (2013)

  5. [13]

    Jour- nal of Atmospheric Sciences 5(4), 165 – 166 (1948)

    Marshall, J.S., Palmer, W.M.K.: The distribution of raindrops with size. Jour- nal of Atmospheric Sciences 5(4), 165 – 166 (1948). https://doi.org/10.1175/ 1520-0469(1948)005<0165:TDORWS>2.0.CO;2

  6. [14]

    Organization, W.M.: Guidelines for Nowcasting Techniques. No. 978-92-63-11198- 2, World Meteorological Organization, 7 bis, avenue de la Paix, P.O. Box 2300, CH-1211 Geneva 2, Switzerland (2017)

  7. [15]

    van Os, S.: Precipitation Nowcasting using a Generative Adversarial Network. Master’s thesis, Delft University of Technology, Faculty of Civil Engineering & Geosciences (2024), http://resolver.tudelft.nl/uuid: 35a11963-4524-442f-97a1-016694b3813f, mentor: M.A. Schleiss

  8. [16]

    In: CD- CEO@ IJCAI

    Pavlík, P., Rozinajová, V., Ezzeddine, A.B.: Radar-based volumetric precipitation nowcasting: A 3d convolutional neural network with u-net architecture. In: CD- CEO@ IJCAI. pp. 65–72 (2022)

  9. [17]

    Geoscientific Model Development12(10), 4185– 4219 (2019)

    Pulkkinen, S., Nerini, D., Pérez Hortal, A.A., Velasco-Forero, C., Seed, A., Ger- mann, U., Foresti, L.: Pysteps: an open-source python library for probabilistic precipitation nowcasting (v1.0). Geoscientific Model Development12(10), 4185– 4219 (2019). https://doi.org/10.5194/...

  10. [18]

    Nature597(7878), 672–677 (2021)

    Ravuri, S., Lenc, K., Willson, M., Kangin, D., Lam, R., Mirowski, P., Fitzsimons, M., Athanassiadou, M., Kashem, S., Madge, S., et al.: Skilful precipitation now- casting using deep generative models of radar. Nature597(7878), 672–677 (2021)

  11. [19]

    Monthly Weather Review 136(1), 78–97 (2008)

    Roberts, N.M., Lean, H.W.: Scale-selective verification of rainfall accumulations from high-resolution forecasts of convective events. Monthly Weather Review 136(1), 78–97 (2008)

  12. [20]

    https: //dataplatform.knmi.nl/ (2025), https://dataplatform.knmi.nl/, accessed: 2025- 04-22

    Royal Netherlands Meteorological Institute (KNMI): KNMI Data Platform. https: //dataplatform.knmi.nl/ (2025), https://dataplatform.knmi.nl/, accessed: 2025- 04-22

  13. [21]

    Advances in neural information processing systems28 (2015)

    Shi, X., Chen, Z., Wang, H., Yeung, D.Y., Wong, W.K., Woo, W.c.: Convolutional lstm network: A machine learning approach for precipitation nowcasting. Advances in neural information processing systems28 (2015)

  14. [22]

    Advances in neural information processing systems30 (2017)

    Shi, X., Gao, Z., Lausen, L., Wang, H., Yeung, D.Y., Wong, W.k., Woo, W.c.: Deep learning for precipitation nowcasting: A benchmark and a new model. Advances in neural information processing systems30 (2017)

  15. [23]

    Bulletin of the American Meteorological Society 95(3), 409–426 (2014)

    Sun, J., Xue, M., Wilson, J.W., Zawadzki, I., Ballard, S.P., Onvlee-Hooimeyer, J., Joe, P., Barker, D.M., Li, P.W., Golding, B., et al.: Use of nwp for nowcasting convective precipitation: Recent progress and challenges. Bulletin of the American Meteorological Society 95(3), 4...

  16. [24]

    Advances in Meteorology 2018(1), 8202031 (2018)

    Wu, W., Zou, H., Shan, J., Wu, S.: A dynamical z-r relationship for precipitation estimation based on radar echo-top height classification. Advances in Meteorology 2018(1), 8202031 (2018)

  17. [25]

    Weather and Forecasting 25(4), 1235–1248 (2010)

    Yang, Y.H., King, P.: Investigating the potential of using radar echo reflectivity to nowcast cloud-to-ground lightning initiation over southern ontario. Weather and Forecasting 25(4), 1235–1248 (2010)

  18. [26]

    Science385(6707), 427–432 (2024)

    Zhang, W., Zhou, T., Wu, P.: Anthropogenic amplification of precipitation variabil- ity over the past century. Science385(6707), 427–432 (2024). https://doi.org/10. 1126/science.adp0212, https://www.science.org/doi/abs/10.1126/science.adp0212

  19. [27]

    Nature pp

    Zhang, Y., Long, M., Chen, K., Xing, L., Jin, R., Jordan, M.I., Wang, J.: Skilful nowcasting of extreme precipitation with nowcastnet. Nature pp. 1–7 (2023) 26 P. Pavlík, M. Schleiss et al

  20. [28]

    Advances in Atmospheric Sciences 40(6), 1043–1057 (2023)

    Zou,H.,Wu,S.,Tian,M.:Radarquantitativeprecipitationestimationbasedonthe gated recurrent unit neural network and echo-top data. Advances in Atmospheric Sciences 40(6), 1043–1057 (2023)

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.