arXiv subjects

Shunji Kotsuki

Publications and source records attributed to Shunji Kotsuki.

10 recordsLinked to original sources

arXiv2026

An extended Perron-Frobenius operator filter for nonlinear state estimation

We propose an extended Perron--Frobenius Operator Filter (PFOF) for nonlinear state estimation. The method learns the Perron--Frobenius operator, an infinite-dimensional linear operator fully preserving properties of a nonlinear dynamical system, using the extended Dynamic Mode Decomposition (eDMD). This enables us to explicitly account for non-Gaussian distributions exhibited by the nonlinear system within a linear-operator representation, while retaining the freedom to choose basis functions in eDMD. Through two numerical examples, we show that the extended PFOF achieves high computational efficiency and high estimation accuracy by exploiting the flexibility in the choice of basis functions in eDMD.

eess.SY↗

arXiv2026

A Mutual Information-Based Ensemble Kalman Filter

Ensemble Kalman filters (EnKFs) are widely used for data assimilation in geophysical systems. Among various implementations, the local ensemble transform Kalman filter (LETKF) has gained popularity because of its computational efficiency. However, the deterministic EnKF such as the LETKF is known to be less robust than the stochastic EnKF in strongly nonlinear regimes. We generalize the LETKF such that it contains a stochastic term and includes the stochastic EnKF within it. We adaptively optimize the parameter that determines the weight of the stochastic term based on an identity of mutual information, which is satisfied by the Kalman filter in linear Gaussian systems. As the analysis perturbation equations of EnKFs are decomposed into a system of equations for modes that are uncorrelated with each other, the application of mutual information is easily achieved. The generalized LETKF thus optimized is named the mutual information-based ensemble Kalman filter (MI-EnKF). The MI-EnKF indirectly uses the third- and fourth-order moments of the forecast ensemble through entropy. To speed up calculations of entropy, we create a lookup table based on maximum entropy distributions. We conduct data assimilation experiments using the Lorenz-96 model to confirm the validity of the optimization method of MI-EnKF. When the observation operator is linear, the ML-EnKF shows the same analysis accuracy as the LETKF. When the observation operator is strongly nonlinear, the MI-EnKF is more accurate than both LETKF and stochastic EnKF regardless of ensemble size. Optimizing just the first mode can lead to significant improvements, but positive impacts of increasing the number of optimized modes are not observed unless the ensemble size is large. The optimized parameter values indicate that the optimal EnKF lies between the deterministic EnKF and the stochastic EnKF.

physics.ao-ph↗

arXiv2026

A Real-Time Remote-Sensing-Guided Decision-Support Framework for Cloud-Seeding Operations: A Field Demonstration Using Himawari-9 and C-band Phased Array Weather Radar

This study proposes a real-time remote-sensing-guided decision-support framework for cloud-seeding operations using high frequency geostationary satellite and ground weather radar observations. The framework integrates cloud assessment, human-in-the-loop decision support, and aircraft operation to translate high-frequency remote-sensing information into actionable guidance for seeding aircraft. We demonstrate the framework using 2.5-min Himawari-9 geostationary satellite observations and 60-s C-band phased-array weather radar (C-PAWR) observations during the preliminary dry-ice cloud-seeding field campaign conducted over Toyama Bay, Japan, in January 2026. In the 13 January case, the framework enabled the ground team to identify a developing cumulus cloud with a lifetime of approximately 20 min, communicate guidance to the aircraft, and conduct seeding immediately before the cloud began to dissipate naturally. Candidate seedable clouds were identified from Himawari-9 infrared indices, and their selection was supported by near-real-time C-PAWR observations of precipitation echoes. Because the released dry-ice amount was limited to 30 kg, this study does not attempt to attribute subsequent cloud evolution to seeding effects. Instead, the results demonstrate that rapid-scan satellite and ground radar observations can support real-time target selection and aircraft guidance for responsible, operationally feasible weather-intervention field experiments.

physics.ao-ph↗

arXiv2026

Sparse Sensor Placement for Reducing Forecast Errors in Ensemble Kalman Filtering

Designing efficient observation networks for reducing forecast errors is a fundamental challenge in numerical weather prediction. Data-driven sparse sensor placement (SSP) and ensemble-based data assimilation via the Ensemble Kalman Filter (EnKF) have each addressed this challenge independently, yet their mathematical connections have not been systematically formalized. This study presents a unified theoretical framework integrating SSP and EnKF through optimal experimental design, providing new theoretical and algorithmic results. While conventional SSP methods aim to reduce analysis errors, this study extends the SSP to target forecast error reduction by using a tangent linear model approximated by an ensemble forecast. We derive the Fisher information matrices in the ensemble and model spaces for the EnKF, and clarify the mathematical interpretations of A-, D-, and E-optimality in terms of forecast error reduction. A-optimality in the model space minimizes the mean forecast error variance; D-optimality is ill-defined in the model space due to rank deficiency and is therefore formulated in the ensemble space, where it maximizes the Shannon information content of assimilated observations; and E-optimality in the model space minimizes the worst-case forecast error variance. We further propose a fast greedy algorithm for selecting observation locations under A-optimality in the model space, avoiding matrix inversion at each greedy step and substantially reducing computational cost. Numerical experiments using the Lorenz-96 model support these theoretical findings. Among the three optimality criteria, A-optimality in the model space most consistently reduces forecast spread and root-mean-square error, and yields stable incremental improvements consistent with post-assimilation observation impact diagnostics.

physics.ao-ph↗

arXiv2025

Bridging Artificial Intelligence and Data Assimilation: The Data-driven Ensemble Forecasting System ClimaX-LETKF

While machine learning-based weather prediction (MLWP) has achieved significant advancements, research on assimilating real observations or ensemble forecasts within MLWP models remains limited. We introduce ClimaX-LETKF, the first purely data-driven ML-based ensemble weather forecasting system. It operates stably over multiple years, independently of numerical weather prediction (NWP) models, by assimilating the NCEP ADP Global Upper Air and Surface Weather Observations. The system demonstrates greater stability and accuracy with relaxation to prior perturbation (RTPP) than with relaxation to prior spread (RTPS), while NWP models tend to be more stable with RTPS. RTPP replaces an analysis perturbation with a weighted blend of analysis and background perturbations, whereas RTPS simply rescales the analysis perturbation. Our experiments reveal that MLWP models are less capable of restoring the atmospheric field to its attractor than NWP models. This work provides valuable insights for enhancing MLWP ensemble forecasting systems and represents a substantial step toward their practical applications.

cs.LG↗

arXiv2025

Exploring Ultra Rapid Data Assimilation Based on Ensemble Transform Kalman Filter with the Lorenz 96 Model

Ultra-rapid data assimilation (URDA) is a method that rapidly updates preemptive forecasts derived from observations without integrating a dynamical model each time additional observations become available. Due to its computational efficiency, we anticipate that URDA will be beneficial for application to numerical weather prediction (NWP); however, the properties of URDA in nonlinear models and its applicability to NWP have not been sufficiently elucidated. Therefore, this study investigates the analytical properties of URDA in nonlinear models and explores inflation and localization that effectively enhance its performance, both of which are generally essential for NWP. We first analytically demonstrate that preemptive forecasts obtained by URDA in nonlinear models are approximately equivalent, under the tangent linear approximation, to forecasts integrated from the analysis. Furthermore, we conduct numerical experiments using the 40-variable Lorenz 96 model. The results show that multiplicative inflation that deliberately deflates (i.e., using an inflation factor less than 1) the forecast ensemble perturbations used to compute the ensemble transform matrix of URDA improves forecast accuracy and inflates ensemble spread moderately. This is presumably attributable to the fact that deflating the forecast ensemble perturbations brings the ensemble transform matrix closer to the identity matrix and reduces the increment of the ensemble mean. With regard to localization, we show that, although R-localization is crucial, advective localization that accounts for the advection of the influence of observations is more effective.

physics.geo-ph↗

arXiv2025

Conditional Diffusion Models for Global Precipitation Map Inpainting

Incomplete satellite-based precipitation presents a significant challenge in global monitoring. For example, the Global Satellite Mapping of Precipitation (GSMaP) from JAXA suffers from substantial missing regions due to the orbital characteristics of satellites that have microwave sensors, and its current interpolation methods often result in spatial discontinuities. In this study, we formulate the completion of the precipitation map as a video inpainting task and propose a machine learning approach based on conditional diffusion models. Our method employs a 3D U-Net with a 3D condition encoder to reconstruct complete precipitation maps by leveraging spatio-temporal information from infrared images, latitude-longitude grids, and physical time inputs. Training was carried out on ERA5 hourly precipitation data from 2020 to 2023. We generated a pseudo-GSMaP dataset by randomly applying GSMaP masks to ERA maps. Performance was evaluated for the calendar year 2024, and our approach produces more spatio-temporally consistent inpainted precipitation maps compared to conventional methods. These results indicate the potential to improve global precipitation monitoring using the conditional diffusion models.

cs.LG↗

arXiv2025

Wasserstein GAN-Based Precipitation Downscaling with Optimal Transport for Enhancing Perceptual Realism

High-resolution (HR) precipitation prediction is essential for reducing damage from stationary and localized heavy rainfall; however, HR precipitation forecasts using process-driven numerical weather prediction models remains challenging. This study proposes using Wasserstein Generative Adversarial Network (WGAN) to perform precipitation downscaling with an optimal transport cost. In contrast to a conventional neural network trained with mean squared error, the WGAN generated visually realistic precipitation fields with fine-scale structures even though the WGAN exhibited slightly lower performance on conventional evaluation metrics. The learned critic of WGAN correlated well with human perceptual realism. Case-based analysis revealed that large discrepancies in critic scores can help identify both unrealistic WGAN outputs and potential artifacts in the reference data. These findings suggest that the WGAN framework not only improves perceptual realism in precipitation downscaling but also offers a new perspective for evaluating and quality-controlling precipitation datasets.

cs.LG↗

arXiv2024

Ensemble data assimilation to diagnose AI-based weather prediction model: A case with ClimaX version 0.3.1

Artificial intelligence (AI)-based weather prediction research is growing rapidly and has shown to be competitive with the advanced dynamic numerical weather prediction models. However, research combining AI-based weather prediction models with data assimilation remains limited partially because long-term sequential data assimilation cycles are required to evaluate data assimilation systems. This study proposes using ensemble data assimilation for diagnosing AI-based weather prediction models, and marked the first successful implementation of ensemble Kalman filter with AI-based weather prediction models. Our experiments with an AI-based model ClimaX demonstrated that the ensemble data assimilation cycled stably for the AI-based weather prediction model using covariance inflation and localization techniques within the ensemble Kalman filter. While ClimaX showed some limitations in capturing flow-dependent error covariance compared to dynamical models, the AI-based ensemble forecasts provided reasonable and beneficial error covariance in sparsely observed regions. In addition, ensemble data assimilation revealed that error growth based on ensemble ClimaX predictions was weaker than that of dynamical NWP models, leading to higher inflation factors. A series of experiments demonstrated that ensemble data assimilation can be used to diagnose properties of AI weather prediction models such as physical consistency and accurate error growth representation.

cs.LG↗

arXiv2024

Convex Optimization of Initial Perturbations toward Quantitative Weather Control

This study proposes introducing convex optimization to find initial perturbations of atmospheric states to realize specified changes in subsequent weather. In the proposed method, we formulate and solve an inverse problem to find effective perturbations in atmospheric variables so that controlled variables satisfy specified changes at a specified time. The proposed method first constructs a sensitivity matrix of controlled variables, such as accumulated precipitation, to the initial atmospheric variables, such as temperature and humidity, through sensitivity analysis using a numerical weather prediction (NWP) model. Then a convex optimization problem is formulated to achieve various control specifications involving not only quadratic functions but also absolute values and maximum values of the controlled variables and initial atmospheric variables in the cost function and constraints. The proposed method was validated through a benchmark warm bubble experiment using the NWP model. The experiments showed that the identified perturbations successfully realized specified spatial distributions of accumulated precipitation.

physics.ao-ph↗