No file available
To read the file of this research,
you can request a copy directly from the authors.
TCP-Diffusion: A Multi-modal Diffusion Model for Global Tropical Cyclone Precipitation Forecasting with Change Awareness
Preprints and early-stage research may not have been peer reviewed yet.
Abstract
Precipitation from tropical cyclones (TCs) can cause disasters such as flooding, mudslides, and landslides. Predicting such precipitation in advance is crucial, giving people time to prepare and defend against these precipitation-induced disasters. Developing deep learning (DL) rainfall prediction methods offers a new way to predict potential disasters. However, one problem is that most existing methods suffer from cumulative errors and lack physical consistency. Second, these methods overlook the importance of meteorological factors in TC rainfall and their integration with the numerical weather prediction (NWP) model. Therefore, we propose Tropical Cyclone Precipitation Diffusion (TCP-Diffusion), a multi-modal model for global tropical cyclone precipitation forecasting. It forecasts TC rainfall around the TC center for the next 12 hours at 3 hourly resolution based on past rainfall observations and multi-modal environmental variables. Adjacent residual prediction (ARP) changes the training target from the absolute rainfall value to the rainfall trend and gives our model the ability of rainfall change awareness, reducing cumulative errors and ensuring physical consistency. Considering the influence of TC-related meteorological factors and the useful information from NWP model forecasts, we propose a multi-model framework with specialized encoders to extract richer information from environmental variables and results provided by NWP models. The results of extensive experiments show that our method outperforms other DL methods and the NWP method from the European Centre for Medium-Range Weather Forecasts (ECMWF).
ResearchGate has not been able to resolve any citations for this publication.
In recent years traditional numerical methods for accurate weather prediction have been increasingly challenged by deep learning methods. Numerous historical datasets used for short and medium-range weather forecasts are typically organized into a regular spatial grid structure. This arrangement closely resembles images: each weather variable can be visualized as a map or, when considering the temporal axis, as a video. Several classes of generative models, comprising Generative Adversarial Networks, Variational Autoencoders, or the recent Denoising Diffusion Models have largely proved their applicability to the next-frame prediction problem, and is thus natural to test their performance on the weather prediction benchmarks. Diffusion models are particularly appealing in this context, due to the intrinsically probabilistic nature of weather forecasting: what we are really interested to model is the probability distribution of weather indicators, whose expected value is the most likely prediction. In our study, we focus on a specific subset of the ERA-5 dataset, which includes hourly data pertaining to Central Europe from the years 2016 to 2021. Within this context, we examine the efficacy of diffusion models in handling the task of precipitation nowcasting, with a lead time of 1 to 3 hours. Our work is conducted in comparison to the performance of well-established U-Net models, as documented in the existing literature. An additional comparative analysis has been done with the forecasting capabilities of the CERRA system, part of the Copernicus Climate Change Service. The novelty of our approach, Generative Ensemble Diffusion (GED), lies in its innovative use of a diffusion model to generate a diverse set of possible weather scenarios. These scenarios are then amalgamated into a single prediction in a post-processing phase. This approach mimics the usual weather forecasting technique consisting in running an ensemble of numerical simulations under slightly different initial conditions by exploiting instead the intrinsic stochasticity of the generative model. In comparison to recent deep learning models addressing the same problem, our approach results in approximately a 25% reduction in the mean squared error. Reverse diffusion is a core concept in our GED approach, is particularly relevant to weather forecasting. In the context of diffusion models, reverse diffusion refers to the process of iteratively refining a noisy initial prediction into a coherent and realistic forecast. By leveraging reverse diffusion, our model effectively simulates the complex temporal dynamics of weather systems, mirroring the inherent uncertainty and variability in weather patterns.
The Northwest Pacific (NWP) tropical cyclone (TC) activity exhibits significant decadal variations with alternating active and inactive periods. However, it remains unknown whether such kinds of decadal variations are predictable. Here, we develop a dynamic-statistic model for the decadal predictions of the tropical cyclone genesis frequency (TCGF) and accumulated cyclone energy (ACE) index of the NWP TCs. The dynamic-statistic model is a combination of decadal prediction experiments by coupled general circulation models (CGCM) from the CMIP6 Decadal Climate Prediction Project (DCPP) and multiple linear regression models based on the correlation relationships between the NWP TCGF (ACE) and large-scale variability modes of sea surface temperature (SST) anomalies in observations. For the TCGF, we first calculate anomalous SST intensities associated with Atlantic multidecadal variability (AMV), Pacific decadal oscillation (PDO), and global mean SST (GMSST) predicted by the decadal prediction experiments. Then, they are substituted into the regression model trained by the historical observational TCs and SST to predict the NWP TCGF. For the ACE, one more predictor, viz. the anomalous SST in the NWP, is involved in its regression model. The dynamic-statistic model can be applied for both deterministic and probabilistic predictions with multi-model ensemble mean, and individual members of the decadal prediction experiments used, respectively. Retrospective predictions for the past 50 years show that the correlation skill of the deterministic predictions for the NWP TCGF (ACE) in the future 2–5 and 6–9 years reach 0.71 and 0.59 (0.59 and 0.41), respectively. The results of this dynamic-statistic model will provide decision-makers of the western Pacific Rim countries with valuable information to adapt to variations in NWP TC activity over the next 10 years.
Global medium-range weather forecasting is critical to decision-making across many social and economic domains. Traditional numerical weather prediction uses increased compute resources to improve forecast accuracy, but does not directly use historical weather data to improve the underlying model. Here, we introduce “GraphCast,” a machine learning-based method trained directly from reanalysis data. It predicts hundreds of weather variables, over 10 days at 0.25° resolution globally, in under one minute. GraphCast significantly outperforms the most accurate operational deterministic systems on 90% of 1380 verification targets, and its forecasts support better severe event prediction, including tropical cyclones tracking, atmospheric rivers, and extreme temperatures. GraphCast is a key advance in accurate and efficient weather forecasting, and helps realize the promise of machine learning for modeling complex dynamical systems.
Extreme precipitation is a considerable contributor to meteorological disasters and there is a great need to mitigate its socioeconomic effects through skilful nowcasting that has high resolution, long lead times and local details1–3. Current methods are subject to blur, dissipation, intensity or location errors, with physics-based numerical methods struggling to capture pivotal chaotic dynamics such as convective initiation⁴ and data-driven learning methods failing to obey intrinsic physical laws such as advective conservation⁵. We present NowcastNet, a nonlinear nowcasting model for extreme precipitation that unifies physical-evolution schemes and conditional-learning methods into a neural-network framework with end-to-end forecast error optimization. On the basis of radar observations from the USA and China, our model produces physically plausible precipitation nowcasts with sharp multiscale patterns over regions of 2,048 km × 2,048 km and with lead times of up to 3 h. In a systematic evaluation by 62 professional meteorologists from across China, our model ranks first in 71% of cases against the leading methods. NowcastNet provides skilful forecasts at light-to-heavy rain rates, particularly for extreme-precipitation events accompanied by advective or convective processes that were previously considered intractable.
Weather forecasting is important for science and society. At present, the most accurate forecast system is the numerical weather prediction (NWP) method, which represents atmospheric states as discretized grids and numerically solves partial differential equations that describe the transition between those states¹. However, this procedure is computationally expensive. Recently, artificial-intelligence-based methods² have shown potential in accelerating weather forecasting by orders of magnitude, but the forecast accuracy is still significantly lower than that of NWP methods. Here we introduce an artificial-intelligence-based method for accurate, medium-range global weather forecasting. We show that three-dimensional deep networks equipped with Earth-specific priors are effective at dealing with complex patterns in weather data, and that a hierarchical temporal aggregation strategy reduces accumulation errors in medium-range forecasting. Trained on 39 years of global data, our program, Pangu-Weather, obtains stronger deterministic forecast results on reanalysis data in all tested variables when compared with the world’s best NWP system, the operational integrated forecasting system of the European Centre for Medium-Range Weather Forecasts (ECMWF)³. Our method also works well with extreme weather forecasts and ensemble forecasts. When initialized with reanalysis data, the accuracy of tracking tropical cyclones is also higher than that of ECMWF-HRES.
Despite continuous improvements, precipitation forecasts are still not as accurate and reliable as those of other meteorological variables. A major contributing factor to this is that several key processes affecting precipitation distribution and intensity occur below the resolved scale of global weather models. Generative adversarial networks (GANs) have been demonstrated by the computer vision community to be successful at super‐resolution problems, that is, learning to add fine‐scale structure to coarse images. Leinonen et al. (2020, https://doi.org/10.1109/TGRS.2020.3032790) previously applied a GAN to produce ensembles of reconstructed high‐resolution atmospheric fields, given coarsened input data. In this paper, we demonstrate this approach can be extended to the more challenging problem of increasing the accuracy and resolution of comparatively low‐resolution input from a weather forecasting model, using high‐resolution radar measurements as a “ground truth.” The neural network must learn to add resolution and structure whilst accounting for non‐negligible forecast error. We show that GANs and VAE‐GANs can match the statistical properties of state‐of‐the‐art pointwise post‐processing methods whilst creating high‐resolution, spatially coherent precipitation maps. Our model compares favorably to the best existing downscaling methods in both pixel‐wise and pooled CRPS scores, power spectrum information and rank histograms (used to assess calibration). We test our models and show that they perform in a range of scenarios, including heavy rainfall.
Uncertainties in ocean-mixing parameterizations are primary sources for ocean and climate modeling biases. Due to lack of process understanding, traditional physics-driven parameterizations perform unsatisfactorily in the tropics. Recent advances in the deep-learning method and the new availability of long-term turbulence measurements provide an opportunity to explore data-driven approaches to parameterizing oceanic vertical-mixing processes. Here, we describe a novel parameterization based on an artificial neural network trained using a decadal-long time record of hydrographic and turbulence observations in the tropical Pacific. This data-driven parameterization achieves higher accuracy than current parameterizations, demonstrating good generalization ability under physical constraints. When integrated into an ocean model, our parameterization facilitates improved simulations in both ocean-only and coupled modeling. As a novel application of machine learning to the geophysical fluid, these results show the feasibility of using limited observations and well-understood physical constraints to construct a physics-informed deep-learning parameterization for improved climate simulations.
Forecasting the trajectory and intensity of tropical cyclones (TCs) is important in disaster mitigation as TC usually causes huge damages. However, it remains a substantial challenge due to the limited understanding of TC complexity. Still, TCs have been observed and recorded for several decades, and so can be predicted if viewed as a spatial‐temporal prediction problem with a huge amount of existing data. We propose a novel TC trajectory and intensity short‐term prediction method: Multi‐Modal Spatial‐temporal Networks (MMSTN). It not only predicts the TC's central pressure, winds, and the location of its center, but also forecasts the TC's varied possible tendencies. Experiments were conducted on the China Meteorological Administration Tropical Cyclone Best Track Dataset. Experimental results show that the proposed MMSTN outperformed state‐of‐the‐art methods as well as the official prediction method of the China Central Meteorological Observatory, in intensity prediction and 6 hr trajectory prediction.
Hurricanes and tropical storms are natural hazards with the capacity to devastate coastal regions. Understanding this catastrophic potential is critical to support informed decision-making at local, state, and federal levels. The primary objective of this research is to use machine learning to rigorously explore and quantify the relationship of tropical cyclone characteristics to their destructive outcomes on the coast of the USA. Historical data on hurricanes and tropical storms are identified and curated to support supervised learning. A novel storm damage ratio is introduced to address the inherent challenge of comparing damage to regions with distinct assets and population. Multiple mathematical models to predict economic impacts from tropical events are created using machine learning methods, and the results are compared. Finally, the storm features that most influence the accuracy of predictions are identified and ranked.
In the last few years, the deep learning (DL) computing paradigm has been deemed the Gold Standard in the machine learning (ML) community. Moreover, it has gradually become the most widely used computational approach in the field of ML, thus achieving outstanding results on several complex cognitive tasks, matching or even beating those provided by human performance. One of the benefits of DL is the ability to learn massive amounts of data. The DL field has grown fast in the last few years and it has been extensively used to successfully address a wide range of traditional applications. More importantly, DL has outperformed well-known ML techniques in many domains, e.g., cybersecurity, natural language processing, bioinformatics, robotics and control, and medical information processing, among many others. Despite it has been contributed several works reviewing the State-of-the-Art on DL, all of them only tackled one aspect of the DL, which leads to an overall lack of knowledge about it. Therefore, in this contribution, we propose using a more holistic approach in order to provide a more suitable starting point from which to develop a full understanding of DL. Specifically, this review attempts to provide a more comprehensive survey of the most important aspects of DL and including those enhancements recently added to the field. In particular, this paper outlines the importance of DL, presents the types of DL techniques and networks. It then presents convolutional neural networks (CNNs) which the most utilized DL network type and describes the development of CNNs architectures together with their main features, e.g., starting with the AlexNet network and closing with the High-Resolution network (HR.Net). Finally, we further present the challenges and suggested solutions to help researchers understand the existing research gaps. It is followed by a list of the major DL applications. Computational tools including FPGA, GPU, and CPU are summarized along with a description of their influence on DL. The paper ends with the evolution matrix, benchmark datasets, and summary and conclusion.
The global navigation satellite system (GNSS) radio occultation (RO) technique is an atmospheric sounding technique that originated in the 1990s. The data provided by this approach are playing a consistently significant role in atmospheric research and related applications. This paper mainly summarizes the applications of RO to numerical weather prediction (NWP) generally and specifically for tropical cyclone (TC) forecast and outlines the prospects of the RO technique. With advantages such as high precision and accuracy, high vertical resolution, full-time and all-weather, and global coverage, RO data have made a remarkable contribution to NWP and TC forecasts. While accounting for only 7% of the total observations in European Centre for Medium-Range Weather Forecasts’ (ECMWF’s) assimilation system, RO has the fourth-largest impact on NWP. The greater the amount of RO data, the better the forecast of NWP. In cases of TC forecasts, assimilating RO data from heights below 6 km and from the upper troposphere and lower stratosphere (UTLS) region contributes to the forecasting accuracy of the track and intensity of TCs in different stages. A statistical analysis showed that assimilating RO data can help restore the critical characteristics of TCs, such as the location and intensity of the eye, eyewall, and rain bands. Moreover, a non-local excess phase assimilation operator can be employed to optimize the assimilation results. With denser RO profiles expected in the future, the accuracy of TC forecast can be further improved. Finally, future trends in RO are discussed, including advanced features, such as polarimetric RO, and RO strategies to increase the number of soundings, such as the use of a cube satellite constellation.
Within the Copernicus Climate Change Service (C3S), ECMWF is producing the ERA5 reanalysis which, once completed, will embody a detailed record of the global atmosphere, land surface and ocean waves from 1950 onwards. This new reanalysis replaces the ERA‐Interim reanalysis (spanning 1979 onwards) which was started in 2006. ERA5 is based on the Integrated Forecasting System (IFS) Cy41r2 which was operational in 2016. ERA5 thus benefits from a decade of developments in model physics, core dynamics and data assimilation. In addition to a significantly enhanced horizontal resolution of 31 km, compared to 80 km for ERA‐Interim, ERA5 has hourly output throughout, and an uncertainty estimate from an ensemble (3‐hourly at half the horizontal resolution). This paper describes the general set‐up of ERA5, as well as a basic evaluation of characteristics and performance, with a focus on the dataset from 1979 onwards which is currently publicly available. Re‐forecasts from ERA5 analyses show a gain of up to one day in skill with respect to ERA‐Interim. Comparison with radiosonde and PILOT data prior to assimilation shows an improved fit for temperature, wind and humidity in the troposphere, but not the stratosphere. A comparison with independent buoy data shows a much improved fit for ocean wave height. The uncertainty estimate reflects the evolution of the observing systems used in ERA5. The enhanced temporal and spatial resolution allows for a detailed evolution of weather systems. For precipitation, global‐mean correlation with monthly‐mean GPCP data is increased from 67% to 77%. In general, low‐frequency variability is found to be well represented and from 10 hPa downwards general patterns of anomalies in temperature match those from the ERA‐Interim, MERRA‐2 and JRA‐55 reanalyses.
Over the past 100 years, the collaborative effort of the international science community, including government weather services and the media, along with the associated proliferation of environmental observations, improved scientific understanding, and growth of technology, has radically transformed weather forecasting into an effective global and regional environmental prediction capability. This chapter traces the evolution of forecasting, starting in 1919 [when the American Meteorological Society (AMS) was founded], over four eras separated by breakpoints at 1939, 1956, and 1985. The current state of forecasting could not have been achieved without essential collaboration within and among countries in pursuing the common weather and Earth-system prediction challenge. AMS itself has had a strong role in enabling this international collaboration.
We present Multi-Source Weighted-Ensemble Precipitation, version 2 (MSWEP V2), a gridded precipitation P dataset spanning 1979–2017. MSWEP V2 is unique in several aspects: i) full global coverage (all land and oceans); ii) high spatial (0.1°) and temporal (3 hourly) resolution; iii) optimal merging of P estimates based on gauges [WorldClim, Global Historical Climatology Network-Daily (GHCN-D), Global Summary of the Day (GSOD), Global Precipitation Climatology Centre (GPCC), and others], satellites [Climate Prediction Center morphing technique (CMORPH), Gridded Satellite (GridSat), Global Satellite Mapping of Precipitation (GSMaP), and Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TMPA) 3B42RT)], and reanalyses [European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-Interim) and Japanese 55-year Reanalysis (JRA-55)]; iv) distributional bias corrections, mainly to improve the P frequency; v) correction of systematic terrestrial P biases using river discharge Q observations from 13,762 stations across the globe; vi) incorporation of daily observations from 76,747 gauges worldwide; and vii) correction for regional differences in gauge reporting times. MSWEP V2 compares substantially better with Stage IV gauge–radar P data than other state-of-the-art P datasets for the United States, demonstrating the effectiveness of the MSWEP V2 methodology. Global comparisons suggest that MSWEP V2 exhibits more realistic spatial patterns in mean, magnitude, and frequency. Long-term mean P estimates for the global, land, and ocean domains based on MSWEP V2 are 955, 781, and 1,025 mm yr−1, respectively. Other P datasets consistently underestimate P amounts in mountainous regions. Using MSWEP V2, P was estimated to occur 15.5%, 12.3%, and 16.9% of the time on average for the global, land, and ocean domains, respectively. MSWEP V2 provides unique opportunities to explore spatiotemporal variations in P, improve our understanding of hydrological processes and their parameterization, and enhance hydrological model performance.
Precipitation nowcasting is one of the fundamental challenges in natural hazard research. High-intensity rainfall, especially the rainstorm, will lead to the enormous loss of people’s property. Existing methods usually utilize convolution operation to extract rainfall features and increase the network depth to expand the receptive field to obtain fake global features. Although this scheme is simple, only local rainfall features can be extracted leading to insensitivity to high-intensity rainfall. This letter proposes a novel precipitation nowcasting framework named Rainformer, in which, two practical components are proposed: the global features extraction unit and the gate fusion unit (GFU). The former provides robust global features learning ability depending on the window-based multi-head self-attention (W-MSA) mechanism, while the latter provides a balanced fusion of local and global features. Rainformer has a simple yet efficient architecture and significantly improves the accuracy of rainfall prediction, especially on high-intensity rainfall. It offers a potential solution for real-world applications. The experimental results show that Rainformer outperforms seven state of the arts methods on the benchmark database and provides more insights into the high-intensity rainfall prediction task.
Precipitation nowcasting is an important task in operational weather forecasts. The key challenge of the task is the radar echo map extrapolation. The problem is mainly solved by an optical-flow method in existing systems. However, the method cannot model rapid and nonlinear movements. Recently, a convolutional gated recurrent unit (ConvGRU) method is developed, which aims to model such movements based on deep learning techniques. Despite the promising performance, ConvGRU tends to yield blurring extrapolation images and fails to multi-modal and skewed intensity distribution. To overcome the limitations, we propose in this letter a generative adversarial ConvGRU (GA-ConvGRU) model. The model is composed of two adversarial learning systems, which are a ConvGRU-based generator and a convolution neural network-based discriminator. The two systems are trained by playing a minimax game. With the adversarial learning scheme, GA-ConvGRU can yield more realistic and more accurate extrapolation. Experiments on real data sets have been conducted and the results demonstrate that the proposed GA-ConvGRU significantly outperforms state-of-the-art extrapolation methods ConvGRU and optical flow.
There are many possible verification measures when deterministic binary forecasts are made. In deciding which to use, various properties of the measures are of interest. One type of property explores what happens when certain types of change are made to the forecasts. In this paper, two types of change, and their effects on various verification measures are examined. Underlying some desirable properties of measures, such as propriety, equitability and consistency, is the desire that it should be impossible to improve the value, or expected value, of a measure by changing the original forecast from the forecaster's “true belief”. This behaviour is often called “hedging”. Among the types of changes that ought not lead to improved forecasts is random changes of subsets of the true binary forecasts. The effect of such changes is discussed here.
A second type of change examined here arises when a continuous forecast is thresholded to produce the binary forecast. When this type of “thresholding” is practiced a posteriori, it can be seen as a fine tuning of the forecasting system, in order to improve its verification performance. In the thresholding framework, one can optimize some verification performance measures with respect to others. In particular, this paper discusses the optimization of bias and Pierce Skill Score.
See http://onlinelibrary.wiley.com/doi/10.1002/qj.3050/full
This paper introduces a network for volumetric segmentation that learns from sparsely annotated volumetric images. We outline two attractive use cases of this method: (1) In a semi-automated setup, the user annotates some slices in the volume to be segmented. The network learns from these sparse annotations and provides a dense 3D segmentation. (2) In a fully-automated setup, we assume that a representative, sparsely annotated training set exists. Trained on this data set, the network densely segments new volumetric images. The proposed network extends the previous u-net architecture from Ronneberger et al. by replacing all 2D operations with their 3D counterparts. The implementation performs on-the-fly elastic deformations for efficient data augmentation during training. It is trained end-to-end from scratch, i.e., no pre-trained network is required. We test the performance of the proposed method on a complex, highly variable 3D structure, the Xenopus kidney, and achieve good results for both use cases.
Equitable skill scores are based on scoring matrices, which assign scores to the various combinations of forecast and observed events. The basic tenets of equitability require that i) all constant forecasts - and random forecasts - receive the same expected score, and ii) the elements of scoring matrices do not depend on the elements of performance matrices. Scoring matrices are assumed here to be symmetric and to possess other reasonable properties related to the nature of the underlying variable. To scale the elements of scoring matrices, the expected scores for constant and random forecasts are set equal to zero and the expected score for perfect forecasts is set equal to one. To obtain a unique scoring matrix, additional conditons must be imposed or some scores must be specified a priori. -from Authors
An integrated form of the diabatic wind profile based on similarity theory is used to estimate surface stress from measured winds and temperatures. It is shown that excellent estimates of stress can be made, given the roughness length, an estimate of the Richardson number and an accurate wind at one level. The theory can further be applied to estimate the roughness length from relatively few observations of wind and temperature not necessarily under neutral conditions.
Suggestions by Taylor and Deacon for the determination of surface stress from autocovariance functions are tested on O'Neill observations. The results show that fair stress estimates can be made if instrumental response is taken into account.
ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate
- Ecmwf
ECMWF. 2017. ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate. https://cds.climate.
copernicus.eu/cdsapp#!/home. Accessed: 2024-07-24.
Prediff: Precipitation nowcasting with latent diffusion models
- Z Gao
- X Shi
- B Han
- H Wang
- X Jin
- D Maddix
- Y Zhu
- M Li
- Y B Wang
Gao, Z.; Shi, X.; Han, B.; Wang, H.; Jin, X.; Maddix, D.;
Zhu, Y.; Li, M.; and Wang, Y. B. 2024. Prediff: Precipitation nowcasting with latent diffusion models. Advances in
Neural Information Processing Systems, 36.
Earthformer: Exploring space-time transformers for earth system forecasting
- Z Gao
- X Shi
- H Wang
- Y Zhu
- Y B Wang
- M Li
- D.-Y Yeung
Gao, Z.; Shi, X.; Wang, H.; Zhu, Y.; Wang, Y. B.; Li, M.;
and Yeung, D.-Y. 2022. Earthformer: Exploring space-time
transformers for earth system forecasting. Advances in Neural Information Processing Systems, 35: 25390-25403.
Denoising diffusion probabilistic models. Advances in neural information processing systems
- J Ho
- A Jain
- P Abbeel
Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion
probabilistic models. Advances in neural information processing systems, 33: 6840-6851.
Video diffusion models
- J Ho
- T Salimans
- A Gritsenko
- W Chan
- M Norouzi
- D J Fleet
Ho, J.; Salimans, T.; Gritsenko, A.; Chan, W.; Norouzi, M.;
and Fleet, D. J. 2022. Video diffusion models. Advances in
Neural Information Processing Systems, 35: 8633-8646.
MGTCF: multi-generator tropical cyclone forecasting with heterogeneous meteorological data
- C Huang
- C Bai
- S Chan
- J Zhang
- Y Wu
Huang, C.; Bai, C.; Chan, S.; Zhang, J.; and Wu, Y. 2023.
MGTCF: multi-generator tropical cyclone forecasting with
heterogeneous meteorological data. In Proceedings of
the AAAI Conference on Artificial Intelligence, volume 37,
5096-5104.
The international best track archive for climate stewardship (IBTrACS) unifying tropical cyclone data
- K R Knapp
- M C Kruk
- D H Levinson
- H J Diamond
- C J Neumann
Knapp, K. R.; Kruk, M. C.; Levinson, D. H.; Diamond,
H. J.; and Neumann, C. J. 2010. The international best track
archive for climate stewardship (IBTrACS) unifying tropical cyclone data. Bulletin of the American Meteorological
Society, 91(3): 363-376.
Multi-layer networks for ensemble precipitation forecasts postprocessing
- F Xu
- G Li
- Y Du
- Z Chen
- Y Lu
Xu, F.; Li, G.; Du, Y.; Chen, Z.; and Lu, Y. 2021. Multi-layer
networks for ensemble precipitation forecasts postprocessing. In Proceedings of the AAAI Conference on Artificial
Intelligence, volume 35, 14966-14973.