More Like this

1. Using Novel Observations to Evaluate Background Ensemble Forecasts at the Air‐Sea Interface from a Self‐Cycled HAFS Data Assimilation System for Hurricane Fiona (2022)

   https://doi.org/10.1029/2025JD044344  

   Li, Tsung‐Han; Wang, Xuguang; Lu, Xu; Yang, Yue; Zhang, Jun A; Kim, Hyun‐Sook (March 2026, Journal of Geophysical Research: Atmospheres)

   Abstract A field campaign during Hurricane Fiona (2022) collected uniquely coordinated observations at the air‐sea interface using paired dropsondes, Airborne EXpendable BathyThermographs, gliders, and saildrones. These platforms measured key atmospheric and oceanic parameters, including wind speed, temperature, and humidity, to quantify air‐sea fluxes and characterize the atmospheric boundary layer and upper ocean. This study is the first to leverage these observations to evaluate the background ensemble at the air‐sea interface from a self‐cycled Hurricane Analysis and Forecast System data assimilation (DA) system. Verification against observations demonstrated that the background deterministic forecast captures key spatial and temporal variability of surface air temperature (SAT), humidity, wind speed, and sea surface temperature (SST), although systematic biases in humidity and SST were identified due to a combination of model and initialization errors. Evaluation of the 6‐hr background ensemble showed skillful spread for SAT, humidity, and wind speed relative to the forecast error variance. In contrast, the SST ensemble spread was markedly underestimated and inversely related to forecast error variance, emphasizing the need for sampling the oceanic state uncertainty. For longer background forecasts, temporal cross‐correlation of SST with SAT and surface wind speed revealed physically coherent time‐lagged relationships, with atmospheric changes preceding SST responses by approximately 24–36 hr. Similar correlations were preserved in ensemble perturbations, reflecting the ability of the background ensemble in maintaining the dynamic consistency across the coupled interface. These findings provide insights into improving the coupled background ensemble covariances, paving the way for the development of strongly coupled ocean‐atmosphere DA for hurricane forecasting. 

   more » « less
2. Robustness of the Ensemble Score Filter to the Type of Assimilated Observation Networks

   https://doi.org/10.1002/asl.70004  

   Xiong, Zixiang; Liang, Siming; Bao, Feng; Zhang, Guannan; Chipilski, Hristo G (January 2026, Atmospheric Science Letters)

   ABSTRACT Recent advances in data assimilation (DA) have focused on developing more flexible approaches that can better accommodate nonlinearities in models and observations. However, it remains unclear how the performance of these advanced methods depends on the observation network characteristics. In this study, we present initial experiments with the surface quasi‐geostrophic model, in which we compare a recently developed ensemble filter using score‐based diffusion models with the standard Local Ensemble Transform Kalman Filter (LETKF). Our results show that the analysis solutions respond differently to the number, spatial distribution, and nonlinear fraction of assimilated observations. We also find notable changes in the multiscale characteristics of the analysis errors. Given that standard DA techniques will eventually be replaced by more advanced methods, we hope this study sets the ground for future efforts to reassess the value of Earth observing systems in the context of newly emerging algorithms. 

   more » « less
3. Innovation-Based Methods for Online Estimates of Observation Error Variances within Ensemble Data Assimilation Cycles

   https://doi.org/10.1175/MWR-D-24-0242.1  

   Santer, Henry; Poterjoy, Jonathan; El_Gharamti, Mohamad (September 2025, Monthly Weather Review)

   Abstract Many data assimilation methods require knowledge of the first two moments of the background and observation errors to function optimally. To ensure the effective performance of such methods, it is often advantageous to estimate the second moment of the observation errors directly. We examine three different strategies for doing so, focusing specifically on the case of a single scalar observation error variance parameterr. The first method is the well-known Desroziers et al. “diagnostic check” iteration (DBCP). The second method, described in Karspeck, adapts the “spread–error” diagnostic—used for assessing ensemble reliability—to observations and generates a point estimate ofrby taking the expectation of various observation-space statistics and using an ensemble to model background error statistics explicitly. The third method is an approximate Bayesian scheme that uses an inverse-gamma prior and a modified Gaussian likelihood. All three methods can recover the correct observation error variance when both the background and observation errors are Gaussian and the background error variance is well specified. We also demonstrate that it is often possible to estimatereven when the observation error is not Gaussian or when the forward operator mapping model states into observation space is nonlinear. The DBCP method is found to be most robust to these complications; however, the other two methods perform similarly well in most cases and have the added benefit that they can be used to estimaterbefore data assimilation. We conclude that further investigation is warranted into the latter two methods, specifically into how they perform when extended to the multivariate case. Significance StatementObservations of the Earth system (e.g., from satellites, radiosondes, aircraft, etc.,) each have some associated uncertainty. To use observations to improve model forecasts, it is important to understand the size of that uncertainty. This study compares three statistical methods for estimating observation errors, all of which can be continuously implemented whenever new observations are used to correct a model. Our results suggest that all three methods can improve forecast outcomes, but that, if observations are believed to have highly biased or skewed errors, care should be taken in choosing which to use and interpreting its results. Future studies should investigate robust methods for estimating more complicated types of errors. 

   more » « less
4. Challenges for Inline Observation Error Estimation in the Presence of Misspecified Background Uncertainty

   https://doi.org/10.1175/MWR-D-22-0298.1  

   Walsworth, Andrew; Poterjoy, Jonathan; Satterfield, Elizabeth (September 2023, Monthly Weather Review)

   Abstract For data assimilation to provide faithful state estimates for dynamical models, specifications of observation uncertainty need to be as accurate as possible. Innovation-based methods based on Desroziers diagnostics, are commonly used to estimate observation uncertainty, but such methods can depend greatly on the prescribed background uncertainty. For ensemble data assimilation, this uncertainty comes from statistics calculated from ensemble forecasts, which require inflation and localization to address under sampling. In this work, we use an ensemble Kalman filter (EnKF) with a low-dimensional Lorenz model to investigate the interplay between the Desroziers method and inflation. Two inflation techniques are used for this purpose: 1) a rigorously tuned fixed multiplicative scheme and 2) an adaptive state-space scheme. We document how inaccuracies in observation uncertainty affect errors in EnKF posteriors and study the combined impacts of misspecified initial observation uncertainty, sampling error, and model error on Desroziers estimates. We find that whether observation uncertainty is over- or underestimated greatly affects the stability of data assimilation and the accuracy of Desroziers estimates and that preference should be given to initial overestimates. Inline estimates of Desroziers tend to remove the dependence between ensemble spread–skill and the initially prescribed observation error. In addition, we find that the inclusion of model error introduces spurious correlations in observation uncertainty estimates. Further, we note that the adaptive inflation scheme is less robust than fixed inflation at mitigating multiple sources of error. Last, sampling error strongly exacerbates existing sources of error and greatly degrades EnKF estimates, which translates into biased Desroziers estimates of observation error covariance. Significance StatementTo generate accurate predictions of various components of the Earth system, numerical models require an accurate specification of state variables at our current time. This step adopts a probabilistic consideration of our current state estimate versus information provided from environmental measurements of the true state. Various strategies exist for estimating uncertainty in observations within this framework, but are sensitive to a host of assumptions, which are investigated in this study. 

   more » « less
5. Random Forest Regression Using Autonomous In Situ Ocean Observations: Inferring Small-Scale Nutrient Variability during a Southern Ocean Field Experiment

   https://doi.org/10.1175/AIES-D-24-0048.1  

   Song, Sangmin; Lavin, Paige D; Gray, Alison R; Nachod, Zachary; Balwada, Dhruv; Dove, Lilian A; Thompson, Andrew F (April 2026, Artificial Intelligence for the Earth Systems)

   Abstract Small-scale ocean processes are known to modulate nutrient availability in the upper ocean, yet nutrient variability at scales ofandhas remained challenging to characterize from observations. Here, we show that small-scale nutrient information can be inferred from the 2019 Southern Ocean Glider Observations of Submesoscales (SOGOS) field experiment using random forest regression (RFR). In this experiment, a Biogeochemical-Argo float with coarse-resolution nutrient-sensing capabilities was co-deployed with two rapid-sampling Seagliders to autonomously observe a turbulent region by the Southwest Indian Ridge over ∼80 days in austral winter. Since the Seagliders did not measure nitrate directly, we used RFR to estimate nitrate from temperature, salinity, pressure, and oxygen data. The RFR was locally trained using nearby ship-based and Biogeochemical-Argo float observations to capture region-specific relationships between these variables. When tested on nitrate data from the SOGOS float, our RFR reproduced these measurements with high fidelity (95% of absolute errors ≤ 0.72μmol kg⁻¹) and outperformed existing, widely used global nutrient estimation routines trained primarily on ship-based data. Applying the RFR to the Seaglider observations then generated novel nitrate distributions extending to 1000 m with a horizontal resolution of ∼1.5 km and 2–3 h compared to ∼75 km and 5 days for the SOGOS float. Our analysis of high-resolution machine learning estimates of Seaglider nitrate, coupled with Seaglider observations of other properties, suggests that vigorous small-scale stirring enhances upper-ocean nutrient availability in this region and highlights the importance of oceanic turbulence in mediating biological productivity. Significance StatementRapid fluctuations in ocean nutrient content can be intense but difficult to observe directly. Here, we infer high-frequency nutrient information from a Southern Ocean field experiment by using regression to leverage multiple types of instruments. We utilize local nitrate measurements from two observing platforms (ship and autonomous Argo floats) to train a random forest regression that will estimate unknown nitrate concentrations along the tracks of different autonomous instruments with significantly faster sampling rates (Seagliders). Our targeted application produces novel nitrate estimates at high enough resolution to observe strong, small-scale nutrient stirring as the Seagliders traverse a turbulent oceanic region. Increased observational coverage of the ocean, combined with machine learning approaches, can thus enable ocean analyses at increasingly fine scales. 

   more » « less