NVIDIA Earth-2 Introduces AI Data Assimilation for Timely Regional and Global Weather Forecasting

NVIDIA Earth-2 Introduces AI Data Assimilation for Timely Regional and Global Weather Forecasting NVIDIA Earth-2 provides AI data assimilation tools within Earth2Studio to help organizations integrate local, proprietary, and remote-sensing observations into weather forecasting pipelines. These techniques bypass the processing delays of numerical data assimilation, allowing faster forecast updates and improved localized accuracy. Score-Based Data Assimilation (SDA) guides diffusion models like CorrDiff and StormCast toward observations at each denoising step without model retraining. Applying SDA reduced wind-speed root-mean-square error by 54% in a CorrDiff regional downscaling test and by an average of 7.2% across six steps in StormCast CONUS forecasts. HealDA combines in situ and satellite observations using an observation encoder and a vision transformer backbone to estimate the global atmospheric state in seconds. Earth2Studio standardizes access to diverse weather data sources, including radar networks, weather stations, and geostationary satellites, through a unified Python interface. Regional Forecasting with Score-Based Data Assimilation Score-Based Data Assimilation (SDA) incorporates point observations into diffusion-based models, such as CorrDiff for downscaling and StormCast for regional forecasting. Diffusion models produce high-resolution outputs through iterative denoising steps. At each step, SDA compares intermediate predictions against observations via an observation operator and nudges the model toward consistency without retraining. This mechanism enables continuous forecast updates that bridge the latency window between numerical analyses and real-time conditions. In experimental runs, SDA lowered wind-speed RMSE by 54% in CorrDiff downscaling over Europe and by 7.2% across six forecast time steps in StormCast over the central United States. Custom Diffusion Guidance via PhysicsNeMo Developers can extend custom models with SDA by utilizing diffusion utilities from the PhysicsNeMo library. Pretrained denoising models can be converted into score-predicting pipelines using DPS guidance classes. For direct point-based station measurements, DataConsistencyDPSGuidance aligns masked grid points with observed values. For more complex inputs, such as proxy measurements or multi-grid observations like wind turbine power outputs, ModelConsistencyDPSGuidance connects custom PyTorch models to map generated samples to simulated observations. Global Weather Estimation with HealDA Traditional numerical data assimilation requires substantial computing power, limiting update frequencies. HealDA offers an AI alternative that estimates global atmospheric states in seconds. The architecture features an observation encoder that converts disparate remote-sensing and in situ observations into point-cloud tokens embedded with geolocation and time metadata. A vision transformer backbone then processes these tokens on a HEALPix grid. HealDA supports custom conventional data natively and allows encoder adaptation for proprietary satellite data streams. Unified Data Access in Earth2Studio Earth2Studio includes a standard interface across weather and climate data platforms. It supports gridded products from radar networks like MRMS and OPERA, geostationary satellites like GOES and Meteosat, reanalysis products such as ERA5, and tabular observation databases including GHCN, ASOS, and NOAA UFS replays. Developers query data source objects with standardized parameters for timestamps and variables, simplifying multi-source data ingestion for training, model initialization, and verification.

NVIDIA Earth-2 推出 AI 数据同化工具：利用最新观测数据加速气象决策 NVIDIA 在 Earth-2 中引入了 AI 数据同化（DA）工具，帮助能源、保险与应急等行业将专有或第三方观测数据高效融入气象预报管线。通过区域尺度上的基于分数的同化（SDA）以及全球尺度上的 HealDA 模型，用户无需重新训练扩散模型即可在数秒至数分钟内融合最新观测数据，提升预测的时效性与局部精准度。 基于分数的同化（SDA）利用扩散模型去噪步骤融入点观测数据，无需重新训练模型即可约束局部预测结果。 在欧洲 CorrDiff 降尺度实验中，SDA 使测试站点的 10 米风速 RMSE 降低了 54%；在 StormCast 预报中使风速 RMSE 平均降低了 7.2%。 全球模型 HealDA 结合观测编码器与 ViT 架构，能在数秒内将异构点云观测转换为 1° HEALPix 全球大气网格状态。 用户可通过 PhysicsNeMo 的 DPS 引导组件为自定义扩散模型添加同化支持，并利用 Earth2Studio 统一接口调用卫星、雷达与地面观测源。 利用 SDA 优化区域模型与下尺度预报 针对风能、光伏与物流等依赖微观环境的行业，SDA 能将近地面测站或遥感点观测直接映射至扩散模型的去噪步骤中。在欧洲 0.25° 降尺度的 CorrDiff 模型测试中，融入 10 米风速观测后，保留测试站点的 RMSE 降低了 54%。在基于 HRRR 初始化、分辨率为 3 公里的美国本土 StormCast 预报中，SDA 阶段填补了分析数据发布的时间差，使 6 个预报时步的风速 RMSE 平均下降 7.2%。 通过 PhysicsNeMo 为自定义模型构建同化管线 对于拥有自研去噪扩散模型的开发者，可利用 PhysicsNeMo 库中的扩散调度器与扩散后验采样（DPS）引导组件实现 SDA。其中 DataConsistencyDPSGuidance 可处理气象站等网格掩码观测，而 ModelConsistencyDPSGuidance 则支持引入自定义 PyTorch 模型，将模拟观测与风电机组输出功率等派生变量直接关联。 全球大气状态估计技术 HealDA 传统数值数据同化（NDA）计算耗时冗长，导致全球预报更新延迟。HealDA 采用观测编码器与 Vision Transformer（ViT）主干网络，将时空窗口内的异构卫星遥感与原位观测视为点云并嵌入为 Token，最终在数秒内输出 1° HEALPix 网格的全球一致大气状态，大幅提升了全球预报初始场生成的时效性。 Earth2Studio 统一观测数据访问接口 Earth2Studio 提供了标准化的数据抽象层，统一对接静止轨道卫星（GOES、Himawari、Meteosat）、雷达网（MRMS、OPERA）、极轨卫星系统以及 GHCN、GDAS 和 ASOS 等常规地基观测数据。用户通过一致的函数调用传参即可获取不同时间跨度与变量的数据帧或网格场，降低了搭建气象数据流水线与验证基准的复杂度。

Weather-sensitive industries increasingly have access to observations that offer an earlier, more local view of changing conditions. Energy companies collect measurements across wind and solar assets, emergency management teams rely on radar and local sensors, and satellite providers continuously observe the Earth. This data helps organizations understand and manage physical risk across sectors such as capital markets, insurance, agriculture, and logistics. With the AI data assimilation tools in NVIDIA Earth-2, you can process these observations more efficiently. By incorporating proprietary or third-party data, you can use these tools to issue forecasts more frequently, keep estimates aligned with real-time conditions, and tailor your forecasting pipeline to specific regions and applications. This tutorial covers two techniques: Constraining diffusion models with point observations, typically used for regional models. Assimilating disparate datasets into a consistent state, typically used for global models. Prerequisites For this tutorial, you will need: A development environment with Earth2Studio installed An NVIDIA RTX PRO or data center GPU A basic knowledge of Python Approximately 30 minutes Improve regional forecasts with observations You might run a regional weather forecasting pipeline for managing energy production and demand, drawing observations from wind and solar parks, transmission corridors, or densely populated areas. With AI data assimilation, you can use these observations to constrain your forecast where local accuracy matters most, helping you improve operational decisions. The same techniques can support other sectors using observations from production sites, event venues, logistics networks, or other assets where local conditions directly drive decisions. You can use Score-Based Data Assimilation (SDA) to incorporate observations into diffusion-based AI downscaling and forecasting models such as CorrDiff and StormCast. SDA guides the model toward predictions that are consistent with your observations without requiring to retrain the model. Figure 1, below, shows how the process works under the hood. Diffusion models generate high-resolution predictions through a sequence of denoising steps. At each step, SDA compares the intermediate prediction with your observations and nudges the model in the right direction. The output of SDA is probabilistic, with less uncertainty near observation locations and a wider spread further away, where predictions are increasingly governed by the other model inputs and the underlying AI simulations. Figure 1. SDA leverages the multi-step denoising process of diffusion models. At each step, the intermediate result is compared with observations to guide the next denoising step. The final output is an observation-informed, high-resolution weather prediction To nudge the model, you define an observation operator, which maps the model output to the quantity you would expect to observe at each measurement location. This is particularly straightforward for in situ measurements of physical quantities such as temperature or wind speed. In this case, the simplest form of an operator interpolates nearby grid values to each observation location. It is also possible to create operators for proxy measurements or observed impacts. For example, the power output of a wind turbine can act as a proxy measurement of wind speed. SDA unlocks two major capabilities: Update forecasts more rapidly. Numerical analyses require substantial processing time and are released on fixed schedules. SDA enables you to incorporate observations continuously. Figure 2, below, shows a concrete example of a pipeline forecasting at one-hour intervals and depending on a global analysis with a six-hour dissemination schedule. Incorporate proprietary, regional or domain-specific observations. Numerical analyses draw on a broad range of observations. SDA lets you incorporate data from your own sources to focus your forecast on specific asset locations or downstream applications. The effectiveness of SDA depends on several factors: the number, spatial distribution, and accuracy of your observations; the characteristic length scales of the field you are predicting; and the quality and well-posedness of the observation operator. Figure 2. Example pipeline using SDA for downscaling (CorrDiff + SDA) and high-resolution forecasting (StormCast + SDA). CorrDiff creates the high-resolution initial conditions for StormCast. SDA bridges the gap between the analysis reference time and current time by incorporating observations into downscaling and forecast steps that lie in the past How to run CorrDiff-SDA in Earth2Studio CorrDiff is a technique for AI-based downscaling. Earth2Studio provides a CorrDiff model pretrained over Europe that turns 0.25° weather fields into 2.2-km predictions. Using AI data assimilation, you can improve these predictions with observations where local accuracy matters. With the refined outputs, you can then initialize a regional forecast or create a reanalysis dataset for calibrating downstream models. Start by loading the pretrained model. We limit the domain to a part of the Netherlands and northwestern Germany and choose to assimilate 10-meter wind speeds. from datetime import datetime from earth2studio.data import GHCNHourly from earth2studio.models.da import CorrDiffCosmoEra5SDA domain = dict(lat_min=50.2, lat_max=53.8, lon_min=4.6, lon_max=10.4) sda = CorrDiffCosmoEra5SDA.load_model( CorrDiffCosmoEra5SDA.load_default_package(), assimilate_variables=("u10m", "v10m"), resolution="rea2", domain=domain, number_of_samples=1, sampler_steps=12, amp=True, ).to("cuda") Fetch the ERA5 data for low-resolution conditioning and GHCN wind observations over the domain. # Fetch and regrid ERA5 inputs onto the high resolution regional grid # Follow the link to the example below for the full implementation init_time = datetime(2024, 1, 26) x = fetch_and_regrid_era5(init_time, domain) # Fetch GHCN hourly 10-m wind observations over the model domain lat, lon = sda.model.lat_output_numpy, sda.model.lon_output_numpy bbox = lat.min(), lon.min(), lat.max(), lon.max() ghcn = GHCNHourly(stations=GHCNHourly.get_stations_bbox(bbox)) obs = ghcn(init_time, ["u10m", "v10m"]).dropna(subset=["observation"]) Lastly, run the model with the input data. We perform two runs to measure how the additional observations affect the results. prior = sda(x) # free downscaling, no observations analysis = sda(x, obs) # guide the diffusion toward the observations For a complete implementation, see the example in Earth2Studio , from which the code above was adapted. Figure 3. CorrDiff-COSMO downscaling with and without SDA. In this example, SDA reduces the wind-speed RMSE at held-out stations by 54%. Left: wind speed downscaled without SDA. Middle: wind speed downscaled with SDA using GHCN-Hourly station observations. Right: difference in wind speed between the SDA and no-SDA experiments How to run StormCast-SDA in Earth2Studio StormCast is a technique similar to CorrDiff but designed for high-resolution, regional forecasting. Earth2Studio includes a StormCast model pretrained over the contiguous U.S. (CONUS) that is initialized with HRRR and makes predictions at a 3-km resolution. The computation and dissemination of a new HRRR analysis takes some time, but you can use SDA to combine the currently available analysis with the latest observations to update your forecast. Start by loading the pretrained model. We limit the domain to the central U.S. import numpy as np from earth2studio.data import GHCNHourly from earth2studio.models.px import StormCastCONUS # Limit the domain to the central U.S. hrrr_lat_lim, hrrr_lon_lim = (305, 785), (595, 1203) model = StormCastCONUS.load_model( StormCastCONUS.load_default_package(), hrrr_lat_lim=hrrr_lat_lim, # comment out for full CONUS domain hrrr_lon_lim=hrrr_lon_lim, # comment out for full CONUS domain num_diffusion_steps=18, num_sda_diffusion_steps=96, # more steps for SDA for better stability sda_std_obs=0.15, sda_gamma=1e-3, ).to("cuda") Next, fetch the HRRR analysis for model initialization and define the observation data source over the model domain. # Fetch HRRR initial conditions # Follow the link to the example below for the full implementation init_time = datetime(2026, 4, 17, 18) x, coords = fetch_hrrr(init_time) # Define GHCN hourly data source for the model domain lat, lon = model.lat, model.lon bbox = lat.min(), lon.min(), lat.max(), lon.max() ghcn = GHCNHourly( stations=GHCNHourly.get_stations_bbox(bbox), time_tolerance=timedelta(minutes=15), ) We can now run the model using observations during the initial rollout steps before transitioning to forecasting without additional observations. Similar to the illustration in Figure 2, above, this approach uses observations to bridge the gap between the latest analysis and current conditions, after which the forecast proceeds independently. For a pipeline initialized with HRRR, only one SDA-informed step is typically relevant before a new analysis arrives. When using a global analysis for initialization, multiple rollout steps can benefit from SDA. # Initialize generator and get the first output (analysis passthrough) gen = model.create_generator(x.clone(), coords.copy()) x, coords = next(gen) # Run the first part of the rollout with SDA for step in range(nsteps_sda): valid_time = np.array( [coords["time"][0] + coords["lead_time"][0] + np.timedelta64(1, "h")] ) obs = ghcn(valid_time, ["u10m", "v10m", "t2m"]) x, coords = gen.send(obs) # advance one step with observations # Run the remaining rollout without SDA for step in range(nsteps_non_sda): x, coords = next(gen) # advance one step without observations You can find a full implementation of the example in the Earth2Studio example library. Figure 4. StormCast-CONUS forecasts with and without SDA. In this example, SDA reduces the wind-speed RMSE at held-out stations by an average of 7.2% across six time steps. The top and bottom rows show the 3- and 6-hour forecasts, respectively. Left: wind-speed prediction without SDA. Middle: wind-speed prediction with SDA using GHCN-Hourly station observations. Right: difference in wind speed between the SDA and no-SDA predictions How to use SDA with your own model You can assimilate observations with a custom model by extending its Earth2Studio model wrapper. To do this, use the diffusion utilities in PhysicsNeMo. We start with x0_predictor , a pre-trained denoising diffusion model that takes a noisy sample and its noise level as inputs and predicts a noise-free sample. Without SDA, the diffusion sampling for the model would be implemented like this: from physicsnemo.diffusion.noise_schedulers import EDMNoiseScheduler from physicsnemo.diffusion.samplers import sample # Construct diffusion scheduler sigma_min, sigma_max = 0.01, 100 scheduler = EDMNoiseScheduler(sigma_min=sigma_min, sigma_max=sigma_max) # Get denoiser from scheduler denoiser = scheduler.get_denoiser(x0_predictor=x0_predictor) # Generate sample latents = sigma_max * torch.randn(shape) sample(denoiser, latents, noise_scheduler=scheduler, num_steps=num_steps) To use SDA, we transform the x0_predictor into a score-predicting model with SDA guidance. We use DataConsistencyDPSGuidance , which associates each masked pixel with a corresponding observed value. You can use it to assimilate observations from weather stations, proprietary sensors, or similar point-based sources. from physicsnemo.diffusion.guidance import ( DataConsistencyDPSGuidance, DPSScorePredictor, ) # Setup SDA guidance guidance = DataConsistencyDPSGuidance( mask=mask, # binary mask that identifies pixels with observations y=y_obs, # gridded observations std_y=sda_std_obs, # the remaining parameters are SDA settings norm=sda_dps_norm, gamma=sda_gamma, sigma_fn=scheduler.sigma, alpha_fn=scheduler.alpha, ) # Convert x0_predictor to score predictor score_predictor = DPSScorePredictor( x0_predictor=x0_predictor, x0_to_score_fn=scheduler.x0_to_score, guidances=guidance, ) denoiser = scheduler.get_denoiser(score_predictor=score_predictor) # Generate sample (identical to non-SDA example) latents = sigma_max * torch.randn(shape) sample(denoiser, latents, noise_scheduler=scheduler, num_steps=num_steps) For more advanced SDA pipelines, use ModelConsistencyDPSGuidance to derive simulated observations from multiple grid points. This approach requires you to provide a PyTorch model that maps each sample to the corresponding simulated observations. With a custom PyTorch model, you can also assimilate observed impacts. For example, you can use a wind power model to assimilate turbine output measurements. For a complete implementation, have a look at the StormCast CONUS wrapper in Earth2Studio , on which the example above is based. Compute the global weather from observations Most global weather forecasting pipelines are initialized with an estimate of the current weather derived through numerical data assimilation. Numerical data assimilation is computationally demanding, which reduces the timeliness and refresh rate of forecasts and makes it harder to integrate custom observations. With an AI-based technique called HealDA , you can estimate the state of the global atmosphere in a matter of seconds. This allows you to issue forecasts closer to current conditions or compute a custom reanalysis. HealDA maps remote-sensing and in situ observations within a time window to a global gridded atmospheric state. It consists of two main components: an observation encoder and a vision transformer (ViT) backbone. The encoder ingests heterogeneous observations as point clouds, embedding each scalar value into a token together with metadata such as geolocation and time. These tokens are then aggregated onto the target grid and processed by the ViT backbone. Figure 5. HealDA combines in situ and remote-sensing observations from different platforms to provide a consistent estimate of the global weather. The observation encoder treats each data stream as a point cloud in space and time and transforms the measurements into tokens, which are processed by a ViT backbone You can use a pretrained global data assimilation model as a starting point. If you have custom conventional observations, you can typically incorporate them without modifying the model. For proprietary satellite data, you can adapt the encoder to support your data sources. This flexibility lets you tailor the data assimilation system to your region or application. You can use the same technique to train a regional instead of a global system. To get started, see the HealDA training pipeline in the open-source Python library PhysicsNeMo . How to run HealDA in Earth2Studio Earth2Studio provides a pretrained global data assimilation model for research purposes. It integrates data from microwave sounders, radio occultation, surface stations, aircraft, buoys, and other sources onto a 1° HEALPix grid (HPX64). First, load the model. from datetime import timedelta import numpy as np from earth2studio.data import UFSObsConv, UFSObsSat, fetch_dataframe from earth2studio.models.da import HealDA model = HealDA.load_model( HealDA.load_default_package(), lat_lon=True, # regrid from HEALPix to regular lat/lon ).to("cuda") Next, fetch the input observations from the NOAA UFS replay repository. We use conventional and satellite observations. # HealDA was trained on the UFS replay window: 21h before to 3h after analysis time time_tolerance = (timedelta(hours=-21), timedelta(hours=3)) analysis_time = np.array([np.datetime64("2024-01-01T00:00")]) # input_coords() returns the schemas the two observation DataFrames must satisfy conv_schema, sat_schema = model.input_coords() # fetch_dataframe attaches the request_time metadata the model needs conv_df = fetch_dataframe( UFSObsConv(time_tolerance=time_tolerance), time=analysis_time, variable=np.array(conv_schema["variable"]), fields=np.array(list(conv_schema.keys())), ) sat_df = fetch_dataframe( UFSObsSat(time_tolerance=time_tolerance), time=analysis_time, variable=np.array(sat_schema["variable"]), fields=np.array(list(sat_schema.keys())), ) Then call the model with the observation data frames. # stateless model - call it directly for a one-shot analysis, or use # create_generator for cycled assimilation analysis = model(conv_obs=conv_df, sat_obs=sat_df) You can find an extended example for running HealDA in the Earth2Studio example library. Access observational data with Earth2Studio Earth2Studio gives you access to a broad range of data sources for developing, initializing, and validating weather models, including observations from different platforms and sensor types. Among these are gridded data from geostationary satellites (GOES, Himawari, Meteosat) and radar networks (MRMS, OPERA), which you can use directly to train and rapidly update regional, high-resolution forecasting models such as StormScope . These sources are especially useful when you want to forecast quantities that depend on insolation or precipitation, like solar power production, cooling processes, and reservoir inflows. For developing and benchmarking a data assimilation system, Earth2Studio also lets you access archives of conventional observations like GHCN/ISD, NNJA, and UFS, as well as operational observations from GDAS and ASOS. These sources provide variables such as temperature and wind speed as data frames. Observations from polar-orbiting satellite systems, including MetOp and JPSS, are also available. Earth2Studio provides a unified interface across all data sources. You instantiate a data source object and call it with a list of timesteps and variable names. Forecast data sources also accept a list of lead times. This consistent interface makes it easy to combine multiple data sources within the same workflow or connect your own observations to a pipeline. era5 = NCAR_ERA5() da_era5 = era5(datetime(2025, 7, 15), ["t2m", "z500"]) print(da_era5.shape) # (1, 2, 721, 1440) ifs = IFS_FX() da_ifs = ifs(datetime(2026, 7, 15), timedelta(hours=48), ["t2m"]) print(da_ifs.shape) # (1, 1, 1, 721, 1440) goes = GOES(satellite="goes19", scan_mode="C") da_goes = goes(datetime(2026, 7, 15), ["abi01c", "abi02c", "abi03c"]) print(da_goes.shape) # (1, 3, 1500, 2500) ghcn = GHCNHourly(stations=["USW00013301"]) df_ghcn = ghcn(datetime(2026, 6, 15), ["t2m", "ws10m"]) print(df_ghcn.shape) # (10, 7) For the full list of supported data sources, see the API reference in the user guide . Get started with Earth2Studio Explore end-to-end AI data assimilation examples in the Earth2Studio example library. To connect your own observations to a pipeline, follow the custom data source example . AI data assimilation lets you issue more accurate, timely forecasts by incorporating the observations that matter to your region or organization. Visit the Earth2Studio user guide to get started with AI data assimilation and explore the broader capabilities of AI weather models.