Research activities in Earth System Modelling. Report No. 55. WCRP Report No. 08/2025

Research Activities in Earth System Modelling WGNE Blue Book Report No. 55 WCRP Report No. 08/2025 November 2025 © 2025, World Meteorological Organization Bibliographic information This report should be cited as “Working Group on Numerical Experimentation (WGNE), 2025. Research activities in Earth system modelling . Report No. 55. WCRP Report No. 08/2025. WMO, Geneva“ About WGNE The Working Group on Numerical Experimentation (WGNE) to foster collaborative development of models of the Earth system (design, implementation, error diagnosis and model revision) across the full range of temporal and spatial scales. WGNE contributes to the mission of the WCRP core project ESMO. About ESMO The Earth System Modelling and Observations (ESMO) core project coordinates, advances, and facilitates all modelling, data assimilation and observational activities within WCRP. Website: wcrp-esmo.org Editors Bimochan Niraula (ESMO IPO) and members of the Working Group on Numerical Experimentation. Design cover Sara Pasqualetto (ESMO IPO) Copyright notice This report is published by the Earth System Modelling and Observations (ESMO) Project under a Creative Commons Attribution 3.0 IGO License (CC BY 3.0 IGO, www.creativecommons.org/licenses/by/3.0/igo) and thereunder made available for reuse for any purpose, subject to the license’s terms, including proper attribution. NOTE The designations employed and the presentation of material in this publication do not imply the expression of any opinion status of any whatsoever on the part of the Secretariat of the World Meteorological Organization concerning the legal country, territory, city or area, or of its authorities, or concerning the delimitation of its frontiers or boundaries. Editorial note: This report has been produced without editorial revision by the WMO secretariat. It is not an official WMO publication and its distribution in this form does not imply endorsement by the Organization of the ideas expressed. Notes from the Editors A From the Editor, Earth System Modeling and Numerical Experimentation are advancing at a significant rate. The use of newer and emerging technologies is moving the field forward, either directly through novel methodologies or through associated growth in existing areas such as coupling, forcing response, parameterization, and verification. We are pleased to present a collection of articles describing such developments and research in this publication. This publication, “Research Activities in Earth System Modelling,” often referred to as the “WGNE Bluebook,” has been a longstanding activity carried out by the WCRP Working Group on Numerical Experimentation (WGNE) since 1970. When WGNE joined the WCRP core project Earth System Modelling and Observation (ESMO), it was decided that the ESMO IPO would assume editorial and publication responsibility for the Bluebook. After a successful first year, we are pleased to publish the second WGNE Bluebook under ESMO’s stewardship and continue to fulfill the publication’s original intention: to provide an accessible avenue for information exchange among modeling researchers. Remarkably, this year marks the 40th anniversary of WGNE’s formation as a working group, initially under the joint supervision of the Commission for Atmospheric Sciences and the World Climate Research Programme and later reporting to the WMO Research Board. The technological progress and transformations over these four decades are immense and this publication stands as a testament to the work of thousands of scientists and researchers who continuously advance the field. Producing this publication required the contributions of many individuals. I would like to thank the panel members of WGNE and Dr. Fanny Adloff for their careful review of the articles. Special thanks go to Sara Pasqualetto for her excellent work on the design and formatting of the final book. Last but not least, I extend my gratitude to all the authors who contributed to this collection. Through the research published in this volume, their work will inform and inspire many others to pursue further research and development in Earth System Modelling. Sincerely, Dr. Bimochan Niraula Scientific Officer, WCRP Earth System Modelling and Observations (ESMO) International Project Office Table of Contents B Title Authors Page Number Section 1 - Development and studies of coupled models and Earth System Models 7 Generic Co-processing for Earth System Models: An Implementation within the Unified Forecast System Weather Model (UFS-WM) U. Turuncoglu 8 Development of High-Resolution Regional Projections of Indian Ocean Chlorophyll-a A. P. Joshi, P. K. Ghoshal, and K. Chakraborty 10 A Comparative Analysis of Bias-correction Methods for CMIP6 Atmospheric State Variables Over the Indian Ocean A. P. Joshi, S. Bhagat, P. K. Ghoshal, and K. Chakraborty 12 Upgrade of the JMA Sub-Seasonal and Seasonal Ensemble Prediction System (JMA/MRI-CPS4) Y. Kubo, K. Ochi, J. Chiba, T. Yoshida, T. Takakura, R. Sekiguchi, Y. Adachi, M. Deushi, S. Hirahara 14 The convection permitting configuration of the Regional Earth System Model RegCM-ES: simulation of the seasonal precipitation over the Northern Adriatic Region M. Reale, G. Giuliani, F. Giordano, M. Garcia Valdecasas-Ojeda, L. Vargas-Heinz , S. Querin, E. Coppola, C. Solidoro, S. Salon 16 Section 2 - Global and regional climate models: response to forcing, impact studies, sub-seasonal and seasonal forecasting 18 Assessing the Indian Ocean Dipole Predictability in MMCFSv1: Role of Subsurface Ocean Dynamics A. Alone, A. Srivastava 19 Climate-Driven Variability of Groundwater Recharge in Arid and Semi-Arid Africa: A Regional Modeling Perspective Wael F. Galal 21 Streamflow Simulation using WRF-Hydro Model over the Cauvery River Basin A. Soni, A. Srivastava 23 Assessing Precipitation Climatology Using a High-Resolution Coupled Regional Ocean–Atmosphere Model (RSMROMS) C. B. Jayasankar, Vasubandhu Misra 25 Section 3 - Advances in forecast / NWP models: case studies, predictability, ensembles 27 Evaluation of WRF Model Performance in Simulating Track and Rainfall of Super Cyclones Gonu (2007), and Amphan (2020) SH.Babji , K.Nagalakshmi 28 Operational KIM4.0 Upgrade for Global Medium-Range Prediction E.-H. Lee, J. Kim, I.-H. Kwon, S. Bae, M.-S. Koo, K.-H. Seol, I.-H. Cho, H.- J. Choi 30 Upgrade of JMA's Operational Global Numerical Weather Prediction System M. Kawaguchi, T. Kinami, Y. Kuroki, N. Shimokawa, K. Sutou, H. Yonehara, H. Yoshimura 32 Introduction of Stochastic Humidity Profile for Convective parametrization (SHPC) method in JMA’s Global Ensemble Prediction System Y. Ota 34 Upgrade of JMA’s Global Ensemble Prediction System Y. Ota, K. Ochi, J. Chiba, H. Oashi, T. Takakura 36 Operational Use of AMSU-A and ATMS Window Channels in JMA’s NWP Systems T. Urata, H. Shimizu 38 Implementing Isochoric-Isobaric Transformations in the MCV Model's Dynamic-Physical Coupling D. Chen, X. L. Li, X. S. Shen 40 Numerical experiments on snow particle characteristics in Furano, Japan A. Hashimoto 42 Section 4 - Parameterization of physical processes in Earth System Models or their components 44 The impact of the mixing length constraint on tropical cyclone simulations of HAFS W.Wang, B. Liu, Z. Zhang 45 A Simplified Lake Model for Seasonal Prediction Y. Adachi, K. Ochi, S. Hirahara, Y. Kubo, R. Sekiguchi 47 Section 5Forecast verification; novel methodologies to diagnose and measure systematic errors 49 Power Spectra Verification of Machine Learning Weather Prediction Barbara Casati, Leo Separovic, Syed Z. Husain 50 Section 6 - Uncoupled and coupled data assimilation for integrated earth system analysis and prediction; methodology and data impact sensitivity studies 52 Update of the Radiative Transfer Model to RTTOV-13 in JMA's NWP Systems H. Shimizu, A. Ando, N. Kamekawa, K. Kondo, N. Kusano, H. Murata, M. Toyokawa, T. Urata 53 Updating Surface Humidity Observation Error in JMA's Regional NWP Systems R. Saito, R. Toguchi 55 Section 7 - Developments in ocean, sea-ice, and wave modeling 57 MOM6-CICE6 Forecasting using the Unified Forecast System-Weather Model Framework Z. Garraffo, J. Cummings, A. Wallcraft, A. Bozec, E. Chassignet, H.C. Kim, S. Akella, D. Iredell, D. Dukhovskoy, A. Mehra, N. Barton 58 Upgrade of the Real-Time Ocean Forecast System to version 2.5 S. Akella, Z. Garraffo, D. Iredell, J. Cummings, A. Mehra 60 Section 8 - Reanalysis datasets and statistical post-processing 62 Comparison of annual average climatology of key oceanographic parameters in the Bay of Bengal and Arabian Sea Gopikrishna N, Sujata K. Mandke, Susmitha Joseph 63 Annual cycle of surface salinity, SST, precipitation, MLD and BLT: Comparison between Bay of Bengal and Arabian Sea Gopikrishna N, Sujata K. Mandke, Sushmita Joseph 65 Conventional Observation Reanalysis (CORe) Datasets L. Zhang, W. Ebisuzaki, A. Kumar, and W. Wang 67 Section 9 - Numerical/computational techniques and model resolution, physics-dynamics and physics-physics cross-component coupling 69 Development of Regional Short-range prediction Systems Based on KIM E.-H. Lee, J. Kim, H. Cho, S. Y. Bae, K.-H. Seol 70 Section 10 - Machine learning and AI in weather prediction and climate modeling 72 Assessment of Regional Acidification and its Amplitude Variability in the Indian Ocean Using an Improved MLBased Surface pCO2 Data Product A. P. Joshi, P. K. Ghoshal, and K. Chakraborty 73 High resolution rainfall predictions over the Indian region through hybrid integration of dynamical and artificial intelligence models D. Trivedi, S. Pattnaik, O. Sharma, N. B. Puhan 75 An SPI-SVM Approach for Early Warning of Rainfall Extremes: Demonstration over a Regional Testbed A. Mahapatra, G. P. Samal 77 Application of artificial intelligence models in improving the intensity forecast of Tropical Cyclones over North Indian Ocean Basin D. Trivedi, O. Sharma, S. Pattnaik, H. Kumar, S. Bansal, N. B. Puhan, 79 Development and studies of coupled models and Earth System Models 1 7 Generic Co-processing for Earth System Models: An Implementation within the Unified Forecast System Weather Model (UFS-WM) Ufuk Turuncoglu National Center for Atmospheric Research, Climate and Global Dynamics Laboratory Email: [email protected] 1. Introduction As the complexity of the earth system modeling applications increases significantly in terms of represented physical processes and their nonlinear interactions, our ability to process and interact with the vast amount of data produced by those applications is fundamentally changed and pushed the community to develop new and novel data processing approaches such as co-processing and in situ visualization. The Unified Forecast System Weather Model (UFS-WM; Worthen et al., 2024) is one of the real examples of such a complex, multi-component modeling system that aims to replace existing NOAA’s operational model suite and applications by unifying different modeling applications and configurations. The main objective of this study is to increase interoperability between Earth system models and novel data-processing tools by providing an easy-to-use, efficient, and standardized modeling environment for interacting with large data. The newly developed data processing component (GeoGate) interacts with any ESMF/NUOPC-based model component to bring co-processing capability and allow gaining insight from data flowing from multiple Earth system model components and using it to make timely, data-driven decisions. The NUOPC Layer mainly defines conventions and a set of generic components for building coupled models using the Earth System Modeling Framework (ESMF; Theurich et al., 2016). The new component is initially integrated with the UFS Coastal modeling system and currently testing to support model development efforts as a part of the existing regression testing (RT) framework. In addition, its plugin-based design enables it to extend its functionality such as Interacting with different programming languages (i.e. Python) to process the data, and having online interaction with AL/ML and data processing tools while the model is running. 2. Results The GeoGate is designed to connect ESMF/NUOPC-compliant Earth system model components to harness the data provided by each component (Fig. 1a). The GeoGate co-processing component acts like a regular physical model component, but it is specialized for large-scale data interaction. It can process data flowing from several model components or produce data via interacting with AI/ML-based models to drive other physical model components in a hybrid coupled modeling configuration. In this design, the GeoGate co-processing component mirrors the export state (ESMF's building block to store fields provided by the component) of each connected component (Fig. 1a) as its import state. Then, it creates a separate data channel (Conduit nodes; Harrison et al., 2022) for each individual connection (e.g., ATM, OCN, ICE, and WAV, as shown in Fig. 1b). Then, GeoGate passes data to the desired predefined plugin (I/O, Python, or ParaView Catalyst) to interact with the data. To demonstrate its capabilities, the newly developed GeoGate component is integrated with the ESMF/NUOPC-based UFS-WM. The UFS-WM is a fully coupled earth system model for shortand medium-range research and operational forecasts. For demonstration, the multi-component fully coupled UFS-WM configuration (cpld_control_gefs, including FV3, MOM6, CICE, and WW3 model components) is used. This configuration showcases the ParaView Catalyst (v2) plugin and 8 CPS4 takes over the role of backing the issuance of One-month Forecasts from the Global Ensemble Prediction System (GEPS). Operational prediction will run every day with five members as per CPS3, but on Tuesdays and Wednesdays the ensemble is enhanced to 25 members up to the one-month lead. Similarly, five-member re-forecasts are enhanced to 13 members up to the one-month lead. 3. Verification Based on Re-forecast Experimentation (1991 – 2020) Figure 1 shows SST biases. CPS4 reduces cold biases in the eastern Indian Ocean (IO) and mid-latitudes for boreal summer and in the eastern Equatorial Pacific due to enhancement of the cloud and cumulus convection scheme. Figure 2 shows a composite of outgoing longwave radiation (OLR) flux starting from an initial date when the convectively active phase of the Madden-Julian Oscillation (MJO) is present in the IO. In relation to GEPS, CPS4 reproduces the slow eastward progression of the MJO as well as CPS3. References Adachi, Y. et al, 2025: A simplified lake model for seasonal prediction. WGNE Res. Activ. Earth Sys. Model., submitted. Deushi, M. and K. Shibata, 2011: Development of a Meteorological Research Institute Chemistry-Climate Model version 2 for the study of tropospheric and stratospheric chemistry. Pap. Meteorol. Geophys., 62, 1–46. Hirahara, S. et al, 2023: Japan Meteorological Agency/Meteorological Research Institute Coupled Prediction System version 3 (JMA/MRI-CPS3). J. Meteor. Soc. Japan, 101, 149–169. Kawai, H. et al, 2017: Interpretation of Factors Controlling Low Cloud Cover and Low Cloud Feedback Using a Unified Predictive Index. J. Climate, 30, 9119-9131. Ota, Y., 2025: Introduction of Stochastic Humidity Profile for Convective parametrization (SHPC) method in JMA’s Global Ensemble Prediction System. WGNE Res. Activ. Earth Sys. Model., submitted. Sakamoto, K. et al, 2023: Reference manual for the Meteorological Research Institute Community Ocean Model version 5 (MRI.COMv5). Tech. Rep. MRI, 87. Smith, R. N. B., 1990: A scheme for predicting layer clouds and their water content in a general circulation model. Quart. J. Roy. Meteor. Soc., 116, 435-460. Yonehara, H. et al, 2020: Upgrade of JMA’s Operational Global Model. WGNE Res. Activ. Earth Sys. Model., 50, 6.19-6.20. Figure 1: SST biases of (a, d) CPS4, (b, e) CPS3, and (c, f) related differences. (a – c) JJA and (d – f) DJF with 1-month prediction lead. Figure 2: Longitude-time composite for OLR anomalies averaged at 5°S – 5°N for predictions starting with the MJO’s convectively active phase in the IO, based on 5 members for each initial date in 305 cases. 15 The convection permitting configuration of the Regional Earth System Model RegCM-ES: simulation of the seasonal precipitation over the Northern Adriatic Region M.Reale (1,*), G. Giuliani (2,3), F. Giordano(2,1), M. Garcia Valdecasas-Ojeda(4), L. Vargas-Heinz (2,3), S. Querin(1), E. Coppola(3), C. Solidoro(1), S. Salon(1) (1) National Institute of Oceanography and Applied Geophysics-OGS, Trieste, Italy, (2) University of Trieste, Trieste, Italy, (3) Abdus Salam ICTP, Trieste, Italy, (4) University of Granada, Granada, Spain *Email: [email protected] Introduction The Northern Adriatic region is characterized by a complex topography with a presence of several chain mountains (rectangular box in ATM in Figure 1, left panel), a relatively flat valley with a complex river network (River in Figure 1, left panel), significant air-sea interactions which drives deep water formations processes in the Northern Adriatic Sea (OCE in Figure 1, left panel) as well as severe thunderstorms over the land. Here we describe the convection permitting configuration of the Regional Earth System Model RegCMES (Sitz et al., 2017; Reale et al., 2020) and show its added value in simulating the seasonal precipitation over the region. The Regional Earth System Model RegCM-ES The RegCM-ES (Figure 1, right panel) is composed by the Regional Climate Model v5 (RegCM5/CLM5/CLMU, Giorgi et al., 2023) with a horizontal resolution of 3 km and 41 vertical levels (convection permitting, ATM), the ocean module MITgcm (OCE, Marshall et al., 1997a,b) with an horizontal resolution of approximately 700 m and 59 vertical levels (non-hydrostatic), and (iii) a river discharge module CHyM (RIVER, Coppola et al, 2007) with an horizontal resolution of approximately 1 km. The run of each component of the system and the exchanges of the fields is managed by a driver based on the Earth System Modeling Framework (ESMF). The model uses as initial and boundary conditions the ECWMF ERA5 reanalysis for ATM (Hersbach et al., 2020) and Copernicus Marine Service Physical Reanalysis for OCE (Escudier et al., 2021). The model has been run in evaluation mode starting from August,1st 1987 onwards. Figure 1 Results Figure 2 shows the simulated total precipitation (in mm) in winter (DJF, a), spring (MAM, c), summer (JJA, e) and autumn (SON, g) in the EURO4AM dataset (Isotta et al., 2014). Panel b, d, f, h shows the minimum biases (contour lines) between the simulated values and the observations, while in shaded colors the added 16 values (AV) defined as 100* abs(BiasRegCM5)-abs(BiasRegCM-ES)/abs(BiasRegCM5). When AV>0 the coupled model improves the simulation of the precipitation. It is clear that the coupled system improves the simulation of the precipitation over the domain mainly in SON (that is the wettest season over the domain) in the Po Valley and at lee of chain mountains (Figure 2h). This improvement is likely linked to a better representation in the coupled system of the air-sea interaction and transport of moisture from the sea to the inland. Figure 2 References Coppola, E., Tomasetti, B., Mariotti, L., Verdecchia, M., & Visconti, G. (2007). Cellular automata algorithms for drainage network extraction and rainfall data assimilation. Hydrological Sciences Journal, 52(3), 579–592. https://doi.org/10.1623/hysj.52.3.579 Escudier, R., Clementi, E., Cipollone, A., Pistoia, J., Drudi, M., Grandi, A., & Pinardi, N. (2021). A high resolution reanalysis for the Mediterranean Sea. Frontiers in Earth Science, 9, 702285 Giorgi, F., Coppola, E., Giuliani, G., Ciarlo`, J. M., Pichelli, E., Nogherotto, R., et al. (2023). The fifth-generation regional climate modeling system, RegCM5: Description and illustrative examples at parameterized convection and convection-permitting resolutions. Journal of Geophysical Research: Atmospheres, 128, e2022JD038199. https://doi.org/10.1029/2022JD038199 Hersbach, H., Bell, B., Berrisford, B., Hirahara, S., Horanyi, A., Munoz-Sabater, J., et al. (2020). The ERA5 global reanalysis. Quarterly Journal of the Royal Meteorological Society, 146(730), 1999–2049. https://doi.org/10.1002/qj.3803 Isotta, F. A., Frei, C., Weilguni, V., Percec Tadic, M., Lassegues, P., Rudolf, B., et al. (2014). The climate of daily precipitation in the Alps: Development and analysis of a high-resolution grid dataset from pan-alpine rain-gauge data. International Journal of Climatology, 34(5), 1657–1675. https://doi.org/10.1002/joc.3794 Marshall, J., Adcroft, A., Hill, C., Perelman, L., & Heisey, C. (1997). A finite.volume, incompressible Navier Stokes model for studies of the ocean on parallel computers. Journal of Geophysical Research, 102(C3), 5753–5766. https://doi.org/10.1029/96JC02775 Marshall, J., Hill, C., Perelman, L., Heisey, C., & Adcroft, A. (1997). Hydrostatic, quasi.hydrostatic and nonhydrostatic ocean modeling. Journal of Geophysical Research, 102(C3), 5733–5752. https://doi.org/10.1029/96JC02776 Reale, M., Giorgi, F., Solidoro, C., Di Biagio, V., Di Sante, F., Mariotti, L., Farneti R., Sannino G. (2020).” The Regional Earth System Model RegCMES: Evaluation of the Mediterranean climate and marine biogeochemistry” Journal of Advances in Modeling Earth Systems,12, e2019MS001812. https://doi.org/10.1029/2019MS001812 Sitz, L. E., Di Sante, F., Farneti, R., Fuentes-Franco, R., Coppola, E., Mariotti, L., Reale, M., Sannino, G., Barreiro, M., Nogherotto, R., Giuliani, G., Graffino, G., Solidoro, C., Cossarini, G. and Giorgi, F. (2017), “Description and evaluation of the Earth System Regional Climate Model (RegCM-ES)” J. Adv. Model. Earth Syst. doi:10.1002/2017MS000933 17 Global and regional climate models: response to forcing, impact studies, subseasonal and seasonal forecasting 2 18 Assessing the Indian Ocean Dipole Predictability in MMCFSv1: Role of Subsurface Ocean Dynamics Ashish Alone1,*, Ankur Srivastava1 1Indian Institute of Tropical Meteorology, Ministry of Earth Sciences, Dr. Homi Bhabha Road, Pashan, Pune, Maharashtra-411008, India Email*: [email protected] Introduction Indian Ocean Diapole (IOD) plays an important role in the tropical Indian Ocean climate variability, characterized by a zonal gradient of sea surface temperature (SST) anomalies between the western (WEIO) and eastern equatorial Indian Ocean (EEIO). The positive IOD event is marked by anomalously warm WEIO and cooler EEIO which results in above-normal rainfall to East Africa and droughts over Indonesia and Australia, while strongly modulating the Indian summer monsoon. Conversely, a negative IOD event with cooler WEIO and warmer EEIO exerts opposite impacts by flipping the rainfall and drought patterns. With the strong influence of IOD on rainfall, agriculture, and socio-economic activities across the Indo-Pacific rim, accurate prediction of IOD events is a key requirement for regional climate services. Predicting IOD remains a challenge due to its strong coupling with both ocean and atmosphere processes. The seasonal cycle of the thermocline - the boundary between warmer surface water and cooler deep water, along with the role of deeper ocean currents, and the feedback processes collectively play a role in IOD evolution. Model IOD forecast often encounters biases, especially in simulating thermocline depth, SST gradients and rainfall teleconnections, limiting the skill of IOD forecasts. Previous studies highlighted that coupled forecast systems tend to show excessive amplitude in simulated IOD events, exhibiting a weaker link between surface and subsurface ocean processes. Studies have shown errors in the representation of climate pattern like El Ni˜ no-Southern Oscillation (ENSO)–IOD interactions, thereby reducing reliability at longer lead times. The Monsoon Mission Coupled Forecast System version 1 (MMCFSv1) has been widely used for seasonal prediction of monsoon variability. While MMCFSv1 has demonstrated skill in simulating large-scale climate modes such as ENSO, its performance in forecasting IOD events has not been extensively evaluated. This study evaluates the seasonal predictive capability of MMCFSv1 for IOD during the period 1982–2016. Study uses diagnostics methods that combine surface SST anomalies, subsurface ocean thermocline depth, and coupled feedback mechanisms. Analysis examines the model’s skill in reproducing the Dipole Mode Index (DMI), the leading modes of SST variability, the role of subsurface anomalies, and the correlations between observed and simulated fields. With the comparative analysis of model and observed data the study aims to identify key strengths and limitations of the model in representing IOD dynamics also the IOD evolution, with strong seasonal dependence between (JuneAugust) JJA onset and (September-November) SON for improving seasonal forecasts. Model and Data MMCFSv1 is a fully coupled ocean–atmosphere general circulation model developed for extended range and seasonal prediction under the Indian Monsoon Mission initiative. The atmospheric component is based on NCEP CFSv2, with a horizontal resolution of T382 ( 38 km). The ocean component is based on the MOM4p0d ocean model, with a horizontal resolution of 0.25° ×0.25° near the equator. For this study, we use May Initial Conditions (ICs) to generate seasonal forecasts for June–August (JJA) and September–November (SON) seasons for the period 1982–2016. These lead times capture the evolution of IOD events during their developing (JJA) and mature (SON) phases. Study uses observations and reanalysis datasets essential for evaluating the simulations from the MMCFSv1. The ERA5 reanalysis daily sea surface temperature (SST), SST anomalies, and U and V wind components at a 0.25° resolution, covering the period from 1982 to 2016 are used. For the observed thermocline depth, the data comes from the Global Ocean Data Assimilation System (GODAS) monthly temperature profiles, where thermocline depth is defined as the depth of the 20°C isotherm. In the model, thermocline depth is extracted from MMCFSv1 subsurface temperature fields by identifying the depth where the temperature crosses 20°C. This interpolation is carried out for every grid point and time step, both in the model and the observed temperature profiles. Results In Figure 1, the result highlights the ability of MMCFSv1 May IC to capture the interannual variability of the Indian Ocean Dipole (IOD). The JJAS DMI detrended time series (Fig. 1a) shows that the model successfully reproduces most positive and negative IOD events during 1982-2016, compared to observations. Although MMCFSv1 slightly 19 Figure 1: IOD Diagnostics from MMCFSv1 May IC and Observations for 1982–2016: (a) Detrended JJAS: DMI Time Series, (b) EOF Modes of JJAS: SST Anomalies, and (c) SST–Wind Correlation Patterns during Positive IOD Events. underestimates the amplitude of extreme events, the overall correlation between the modeled and observed DMI remains significant, indicating a skilled seasonal forecasting ability from the initial May condition. The spatial patterns of SST variability derived from the EOF analysis (Fig. 1b) further demonstrate the model’s ability to reproduce the canonical IOD mode. The first EOF mode in MMCFSv1 shows a dipole structure with warm anomalies over the western equatorial Indian Ocean (WEIO) and cool anomalies over the eastern equatorial Indian Ocean (EEIO), closely resembling the observed mode. The SST-wind correlations (Fig. 1c) also show consistency with the physical processes of the development of the IOD, with enhanced easterly wind anomalies over the central equatorial Indian Ocean coinciding with the cooling of SST in the EEIO, suggesting that the model captures the coupled air–sea feedbacks. The results in Figure 2 focus on the relationship between SST and thermocline depth (D20), a key driver of IOD evolution. Panel (a) shows that the correlation between observed SST and observed D20 is high in the EEIO during JJA and SON, confirming that shallow thermocline conditions favour cold SST anomalies during positive IOD events. MMCFSv1 replicates this relationship, though with slightly weaker correlations. Panel. It signifies the shallow thermocline conditions are closely linked to cooler SSTs, a key mechanism driving positive IOD events due to negative correlations EEIO. The MMCFSv1 model captures this relationship reasonably well, suggesting that while the model reproduces the essential thermocline–SST feedback, it may underestimate its intensity. Panel (b) further illustrates a clear linear relationship between detrended SST anomalies and D20 anomalies, with higher IOD index values corresponding to stronger negative D20 anomalies in the EEIO. This highlights the importance of accurately simulating thermocline depth variability to improve IOD prediction skill. Further analysis using MMCFSv2 is planned to evaluate whether advancements in model physics and resolution can reduce coupled biases, improve thermocline representation, and enhance the seasonal predictability of the IOD. Figure 2: Relationship Between SST and D20 for IOD Diagnostics: (a) Correlation of Observed SST with Observed and Model D20 during JJA and SON (1982–2016), and (b) Scatter plots of SST and D20 Anomalies with IOD Index for JJA and SON (Detrended). 20 Climate-Driven Variability of Groundwater Recharge in Arid and Semi-Arid Africa: A Regional Modeling Perspective Wael F. Galal Geology Department, Faculty of Science, Assiut University Email: [email protected] Introduction In Africa, Groundwater is an essential water source for domestic, agricultural, and industrial activities in arid and semi-arid regions. However, groundwater recharge in these regions is extremely sensitive to climate variability and change. Limited precipitation, high evapotranspiration, and infrequent rainfall lead to complex and variable recharge systems. Predicting groundwater recharge in their changing climate conditions is critical for planning and managing water resources. Advances in coupled Regional Climate Models (RCMs) and hydrological models permits an assessment and projection of regional groundwater recharge variability. Nonetheless, challenges remain in climate projection downscaling, estimating recharge in data-sparse regions, and representation of land-atmosphere interactions. Methodological Overview Recharge variability in arid and semi-arid Africa is assessed using RCMs (e.g. CORDEX Africa) or GCMs, downscaled for use in hydrological models such as SWAT and WaterGAP, and biascorrected to remove model errors (Ali et al., 2022). The coupled models analyze spatiotemporal recharge variability as a function of soils, land use and water input data, and validated using mediocre satellite data (e.g. GRACE, SMAP) or comparable data in some cases. Climate Forcing and Response Forecasted changes in the intensity, duration, and interval of rainfall have a direct effect on recharge events. In this regard, increased temperatures raise evapotranspiration, and subsequent net recharge, even with infrequent extraordinary rainfall events, drops. Research shows that recharge across semi-arid Africa is closely linked to variations in seasonal precipitation associated with larger climate systems such as the El Niño Southern Oscillation (ENSO) and Indian Ocean Dipole (Taylor et al., 2018). Recharge tends to occur in short bursts after intense storms, which supports investigating sub-seasonal and seasonal climate forecasts to see how the principal rainfall and recharge systems follow a similar behavior and pattern. The uncertainty in an RCM's projected thunderstorms and convective rain performance detracts from overall confidence in future recharge estimates. Impact Studies Recharge instability endangers sustainable water supplies and jeopardizes ecosystems and livelihoods, especially those of farmers using shallow aquifers. Reductions in water table levels have reversed good actions taken to protect wetlands and oases. Unplanned recharge can cause unintended consequences and unravel during droughts and water scarcity crises. Scientists call for immediate action regarding groundwater management, advocating for the adoption of managed aquifer recharge (MAR) and climate adaptive policies (Zhang et al., 2021). 21 Sub-seasonal and Seasonal Forecasting Regional climate models can reasonably identify season aperture precipitation forecast values, but repeatedly struggle to provide an accurate forecast of groundwater recharge, involving precipitation bias, precipitation timing, weak soil moisture feedback, and problematic incorporation of groundwater modelling physically. Throughout Africa, standalone, real-time time series hydroclimatic data (e.g. soil moisture, vegetation indices, evapotranspiration) was successfully employed in regional recharge forecasting in select short-term forecasts. Where to from here leads to forecasting, modelling and utilization of seasonal climate predictions combined with land surface models to further improve yet to be imminent drought early warning systems. Conclusion Significant obstacles to sustainable water resource management in entire Africa’s arid and semiarid regions due to uncertainty regarding climate-induced groundwater recharge are ongoing. Though recent progress using regional climate models has revealed some factors influencing recharge dynamics, challenges remain for sub-seasonal and seasonal variability. Remediation of these knowledge gaps will involve a more composite approach that extends observational networks, downscaling approaches, and the integration of socio-hydrological processes. Researchers must also invest in hybrid modelling frameworks that combine process-based understanding with datadriven analysis to improve actionable groundwater recharge forecasts, a necessary action to improve water security for the most vulnerable populations in Africa. References ● Ali, R., et al, (2022). Climate change impacts on groundwater recharge in Africa. Science of the Total Environment, 159765. https://doi.org/10.1016/j.scitotenv.2022.159765 ● Evaristo, J., & Savenije, H. H. G. (2022). Linking seasonal climate forecasts to groundwater recharge. Journal of Hydrology: Regional Studies, 101076. https://doi.org/10.1016/j.ejrh.2022.101076 ● Taylor, R. G., et al, (2018). Groundwater and climate in Africa. Hydrogeology Journal, 26, 1081-1094. https://doi.org/10.1007/s10040-018-1898-8 ● Zhang, Y., et al, (2021). Managed aquifer recharge under climate change. Journal of Hydrology, 126602. https://doi.org/10.1016/j.jhydrol.2021.126602 22 Streamflow Simulation using WRF-Hydro Model over the Cauvery River Basin Aarti Soni and Ankur Srivastava Indian Institute of Tropical Meteorology, Ministry of Earth Sciences, Dr. Homi Bhabha Road, Pashan, Pune, Maharashtra-411008, India Email: [email protected] Background Understanding the hydrological cycle is crucial for managing water resources and assessing risks associated with water, such as drought and floods. (Liu et al., 2008). In this study the large-scale hydrologic modeling system WRF-Hydro is used to assess the streamflow simulation. This modeling framework was developed by the National Center for Atmospheric Research (NCAR) to simulate the movement of water through the land surface and river channels (Gochis et al., 2018). WRF-Hydro is the hydrologic extension of Weather Research Forecasting (WRF) model, which is a next generation numerical weather prediction system. The model can be run in two modes standalone and coupled with the WRF atmospheric model to understand feedback between land and atmosphere. It incorporates key hydrological processes such as surface and subsurface runoff, channel routing, and land-atmosphere interactions, making it suitable for large-scale hydrologic application (i.e., streamflow forecasting, flood modeling, and water resources analysis). To estimate land surface fluxes such as sensible, latent heat, and net radiation in the 1-dimensional column land surface model, WRF-Hydro provides various physics options. The land surface model also estimates soil moisture dynamics in four different layers, canopy interception, canopy transpiration and rainfall partitioning at land into surface runoff and infiltration. In the present study, standalone WRF-Hydro model is set-up with one domain over the entire Cauvery basin boundary (Figure 1a). The parent domain covers the Southern part of India with a spatial resolution of 5 km. Figure 1(a) Study area showing reservoir stations marked as black dots from upstream to downstream (1=Kudige, 2=Kollegal, 3=Bilungundulu, 4=Urachikottai, 5=Kodumudi, 6=Musuri) located in sub-basins. (b) Timeseries of daily streamflow estimates (m3/s) over six stations in the Cauvery River basin for 2003. Model Setup and Preprocessing In this study, GLDAS version 2.1 data sets are used as forcing data during the period of 2001-2022 in the WRF-Hydro model, which is available with a spatial resolution of 0.25 and a temporal resolution of 3 hours. GLDAS data provided by the National Aeronautics and Space Administration (NASA) Goddard Space Flight Streamorder (b) (a) 23 Center (GSFC) Hydrological Sciences Laboratory (HSL) are the combination of satellite and ground-based observational data products (Rodell et al., 2004). Meteorological forcings for the model run, 3-hourly dataset are prepared air temperature, shortwave and longwave radiation, surface pressure, specific humidity, nearsurface U and V wind speed components and precipitation. To create hydrological grids for the NOAH-MP module, we use digital elevation model (DEM) raster and forecast points. In addition to meteorological forcing datasets, the model also utilizes soil properties and general information is available at the NCAR. The spatial soil variability data are prepared using R utility script provided by the NCAR. Furthermore, streamflow in situ data is obtained for 6 stations in the Cauvery River Basin from the Central Water Commission (CWC), India. Daily rainfall observations at 0.25° spatial resolution are acquired from India Meteorology Department (IMD). For setting up the hydrologic model domain, high-resolution terrain data such as flow direction, flow accumulation, and stream networks derived from Digital Elevation Model (DEM), land use land cover, soil type, and watershed boundaries and streamorder, are processed and resampled to match the routing grid. To create the watershed delineation boundaries, the latitude and longitude of each station located within the basin boundary are used. Further, the delineation tool identifies the upstream area that contributes runoff to the discharge location, thereby defining the watershed boundary. This confirms that the modeled watershed aligns with the observed location for accurate streamflow comparison and calibration. Results Figure 1b shows the streamflow comparison between WRF-Hydro-simulated streamflow (blue line) and observed streamflow (green line) across six stations for the year 2003. Results show that across all stations, especially at the downstream sites, the model consistently overestimates streamflow, with peak flows and baseflows both appearing higher than observed values. Lower Cauvery Basin covers more than 50% of total area of the basin of the total area of the basin, which lies mainly in the Tamil Nadu state and receives rainfall mainly from the Northeast Monsoon (October to December). In contrast, upper Cauvery basin located in Karnataka, is primarily dependent on the Southwest Monsoon during June to September. The spatial variation in rainfall patterns impacts the hydrological response of the basin. The time series of streamflow observation during October to December illustrates that streamflow is comparatively higher at the station present in the lower basin due to the rainfall contribution from the Northeast Monsoon. The overestimation in model simulation suggests insufficient infiltration or storage in the model, or uncalibrated routing parameters that lead to excessive runoff. The model will be calibrated to improve the accuracy of the streamflow predictions and will be ultimately utilized to subseasonal to seasonal (S2S) streamflow prediction using forcings derived from operational S2S forecasts. References Rodell, M., Houser, P. R., Jambor, U. E. A., Gottschalck, J., Mitchell, K., Meng, C. J., ... & Toll, D. (2004). The global land data assimilation system. Bulletin of the American Meteorological society, 85(3), 381-394. Liu, C., Wang, Z., Zheng, H., Zhang, L., & Wu, X. (2008). Development of hydro-informatic modelling system and its application. Science in China Series E: Technological Sciences, 51(4), 456-466. Gochis, D. J., Barlage, M., Dugger, A., FitzGerald, K., Karsten, L., McAllister, M., et al. (2018). The WRFHydro modeling system technical description, version 0. In NCAR Technical Note (p. 107). 24 3. Evaluation and forecast performance KIM4.0 demonstrates overall superior performance in global simulations, as summarized in the scorecard in Fig. 1. Specifically, there is an approximately ~3.5% reduction in error for +5-day geopotential height forecasts at 500 hPa over the northern extratropics, based on the monthly average. Slight degradation is observed in the southern hemisphere, likely due to the discarding of some operational physics optimizations, which will be addressed in future upgrades. In terms of precipitation, KIM4.0 shows notable improvements, particularly in simulating strong signals associated with heavy rainfall events. This improvement is partially reflected in the statistical precipitation skill scores shown in Fig. 1b, where the new version outperforms KIM3.9 across all precipitation thresholds. The enhanced simulation performance is especially evident in terrain-following features over the Korean Peninsula, such as snowfall along eastern mountain regions (not shown). Currently, the KMA plans to upgrade the next version, focusing on enhancements in data assimilation and physics revisions. KIAPS is also developing a next-generation KIM with new configurations, such as ~km scale short-range prediction systems for regional domains and extended-prediction systems with coupled models. There advancements are scheduled for release after the completion of the KIAPS’ second project in 2026. Figure 1: Performance evaluation of KIM4.0 versus KIM3.9. (a) scorecard comparing global RMSE against analysis for January 2023 (00 UTC), with green indicating improvement and red indicating degradation. (b) Frequency bias and equitable threat score of 24-h accumulated precipitation (+3-day prediction) over South Korea and Asia, verified against gauge observations in August 2022. References Han, J.-Y. (2025). Impact of different scale-aware cumulus parameterizations on precipitation forecasts over Korea. Atmos. Res., 317, 107990. doi:10.1016/j.atmosres.2025.107990. Lee, E.-H., K.-H. Seol, H.-J. Park, S. Cho, K.-H. Cho, G. Lee, J. Jung, J. Lee, E. Lee, I.-J. Choi, J. Jang (2025) A Forecast Verification/Evaluation System for Development of the Global Weather Prediction System: Korean Integrated Model Analysis/Evaluation Tool (KAT), Atmosphere. Korean Meteorological Society, 35(3), 395-412, doi:/10.14191/Atmos.2025.35.3.395 (in Korean with English Abstract) 31 Upgrade of JMA’s Operational Global Numerical Weather Prediction System KAWAGUCHI Masashi1, KINAMI Teppei1, KUROKI Yukihiro1, SHIMOKAWA Naofumi1, SUTOU Kouhei1, YONEHARA Hitoshi1, and YOSHIMURA Hiromasa2 1Numerical Prediction Division, Japan Meteorological Agency 2Meteorological Research Institute, Japan Meteorological Agency Email: m.kaw[email protected] Introduction In March 2025, the Japan Meteorological Agency (JMA) upgraded its operational global numerical weather prediction system by introducing a revised version of its Global Spectral Model (GSM). The revision involved refinements of parallelization and data structure to reduce computational resource usage, and parametrized radiation and land surface processes, thereby improving forecasting over the previous version (Yonehara et al. 2023), particularly for the midand high latitudes of the Northern Hemisphere. This report outlines individual components of the upgrade and related verification results. Main Updates Parallelization and Data Structure The two-dimensional decomposition method adopted for parallelization among MPI processes was enhanced. The previous GSM used different domain decompositions for semi-Lagrangian advection scheme calculation and other grid-point based calculations such as physical parametrization (JMA 2025). In the updated model, all grid-point based processes are computed in the same domain with fewer MPI communications. In association, the decomposition approach in the data assimilation system was also enhanced. These updates led to a reduction in computational resources such as memory usage. Additionally, the structure for grid-point model data in the GSM was upgraded, enabling more flexible adaptation to computing architecture characteristics. Radiation The globally uniform climatological annual mean used for carbon dioxide concentration was updated from 396.0 ppmv (the 2013 observation value) to 417.9 ppmv (the 2022 value)(WMO 2023). This reduced temperature biases in the stratosphere through strengthened cooling due to increased longwave radiation emissions. Radiation scheme calculations were also optimized. Land Surface The Leaf Area Index (LAI) monthly climatology was updated to a gridded version, as opposed to the previous average over tropical, temperate and boreal latitudinal zones for each vegetation type. The updated data are derived from a more recent MODIS product (Myneni et al. 2002) featuring higher resolution and greater accuracy than the version used in the previous model (Yan et al. 2016 a, b). The update enables more realistic representation of LAI distribution, leading to a reduction in temperature and relative humidity biases in the lower troposphere through enhanced forecast accuracy for sensible and latent heat fluxes. Verification results The effects of parallelization enhancement were evaluated through two experiments run from a single initial condition. Figure 1 shows reduced memory usage for 132-hour forecasts and 4D-Var. The two performance evaluation experiments (TEST for the updated system and CNTL for the previous one) covered July to September 2023 (summer) and December to February 2023/2024 (winter). Figure 2 shows vertical profiles of mean errors (ME) in temperature for CNTL and TEST averaged over summer. Cold biases in the lower troposphere and warm biases in the stratosphere were reduced, mainly due to the update of LAI monthly climatology and carbon dioxide concentration climatology, respectively. 32 References Japan Meteorological Agency, 2025: Outline of The Operational Numerical Weather Prediction at JMA. Japan Meteorological Agency, Tokyo, Japan. Myneni, R. B., S. Hoffman, Y. Knyazikhin, J. L. Privette, J. Glassy, Y. Tian, Y. Wang, X. Song, Y. Zhang, G. R. Smith, A. Lotsch, M. Friedl, J. T. Morisette, P. Votava, R. R. Nemani, and S. W. Running, 2002: Global products of vegetation leaf area and fraction absorbed PAR from year one of MODIS data. Remote Sens. Environ.,83, 214–231. WMO, 2023: The state of Greenhouse Gases in the Atmosphere Based on Global Observations through 2022. WMO Greenhouse Gas Bulletin,19, 1–11. Yan, K., T. Park, G. Yan, C. Chen, B. Yang, Z. Liu, R. R. Nemani, Y. Knyazikhin, and R. B. Myneni, 2016a: Evaluation of MODIS LAI/FPAR Product Collection 6. Part 1: Consistency and Improvements. Remote Sens.,8, 359. Yan, K., T. Park, G. Yan, Z. Liu, B. Yang, C. Chen, R. R. Nemani, Y. Knyazikhin, and R. B. Myneni, 2016b: Evaluation of MODIS LAI/FPAR Product Collection 6. Part 2: Validation and Intercomparison. Remote Sens.,8, 460. Yonehara, H., Y. Kuroki, M. Ujiie, C. Matsukawa, T. Kanehama, R. Nagasawa, K. Ochi, M. Higuchi, Y. Ichikawa, R. Sekiguchi, and S. Hirahara, 2023: Upgrade of JMA’s Operational Global Numerical Weather Prediction System. WGNE Res. Activ. Earth Sys. Modell., 6.15-6.16. Figure 1: Maximum per-node memory usage [MiB] of jobs (132-hour forecasts (FcEf, Tq959) and 4D-Var (Fdv, Tl319)) among all computing nodes. Blue: CNTL; red: TEST. Figure 2: Vertical profiles of mean errors for temperature [K] against radiosonde observations in the 20 – 90°N region for the summer experiment. Blue: CNTL; red: TEST. Lines represent ME at different forecast lead times for days 1 to 11. 33 Introduction of Stochastic Humidity Profile for Convective parametrization (SHPC) method in JMA’s Global Ensemble Prediction System OTA Yoichiro Japan Meteorological Agency Email: [email protected] 1. Introduction The Japan Meteorological Agency (JMA) introduced a new model ensemble scheme called the Stochastic Humidity Profile for Convective parametrization (SHPC) into its Global Ensemble Prediction System (GEPS) in March 2025. The SHPC perturbs humidity profile input for convective parametrizations to represent uncertainties in convective activity. It mitigates the issue of under-dispersiveness in the tropics and makes it possible to reduce the excessive amplitude of initial perturbations made from the Singular Vector (SV) method in the tropics. 2. Formulation and Settings Based on Tompkins and Berner (2008), the SHPC perturbs relative humidity profile input for convective parametrization. Perturbations are generated from random patterns with horizontal and temporal correlations based on the combination of spherical harmonic functions such as the Stochastic Perturbation of Physics Tendency (SPPT) scheme. The perturbation amplitude is at its maximum at the lowest model level, and is exponentially reduced with the logarithm of vertical pressure levels. Figure 1 shows an example of random patterns used for the SHPC. In the experiments described below, the maximum wave number, temporal correlation scale and vertical e-folding scale of perturbations are set as 20, 72 hours and 0.8 (in logarithm with pressure levels), respectively. The perturbation standard deviation at the lowest model level is approximately 4%, and perturbations are truncated at +/-10%. Perturbation amplitude is adjusted so as not to induce supersaturation and negative humidity in the input humidity profile. 3. Experiments To verify system performance for medium-range forecasts with lead times of up to 11 days, retrospective experiments covering periods of one month or more for summer 2021 (12UTC initials from 21 July 2021 to 11 September 2021) and winter 2021/22 (12UTC initials from 21 December 2021 to 11 February 2022) were conducted. The CNTL experiment involved the use of the operational GEPS as of March 2024, with the number of ensemble members reduced to 13. In the TEST experiment, the SHPC was applied to CNTL and the amplitude of initial perturbations made from SV in the tropics was reduced to 60%. Figure 2 shows the spread of zonal wind speed at 250 hPa (U250) normalized using the ensemble mean RMSE in the tropics (20°S – 20°N). The spread was reduced by up to three or four days and increased beyond four days. The initial spread reduction observed was due to the lower initial perturbation amplitude. The SHPC mitigated the underdispersiveness of medium-range forecasts in the tropics. Figure 3 shows Continuous Ranked Probability Scores (CRPSs) of U250 for the tropics. Probabilistic forecast skill was enhanced with better spread-skill relationships. The effect for the extra-tropics was smaller than that for the tropics (not shown). References Tompkins, A. M. and J. Berner (2008). A stochastic convective approach to account for model uncertainty due to unresolved humidity variability. J. Geophys. Res. (113), D18101. DOI: 10.1029/2007JD009284. 34 Figure 1: Example of a random pattern used for the SHPC. Figure 2: Spread of zonal wind speed at 250 hPa (U250) normalized using the ensemble mean RMSE for the tropics (20°S – 20°N) for summer 2021 (left) and winter 2021/22 (right). Green and red lines show CNTL and TEST, respectively. Figure 3: As per Figure 2, but for CRPS against analysis (unit: m/s). Purple lines and yellow triangles show CRPS change ratios ((TEST-CNTL)/CNTL; %, right axis) and statistically significant enhancement at a 95% confidence level with bootstrap application, respectively. 35 Upgrade of JMA’s Global Ensemble Prediction System OTA Yoichiro*, OCHI Kenta, CHIBA Jotaro, OASHI Hiroaki, and TAKAKURA Toshinari Japan Meteorological Agency Email: [email protected] 1. Introduction The Japan Meteorological Agency (JMA) upgraded its Global Ensemble Prediction System (Global EPS) on March 18 2025 to incorporate a new model ensemble scheme, reduced amplitude of initial perturbations in the tropics, recent Global Spectral Model (GSM) developments and revised sea surface temperature (SST) perturbations. 2. Major Updates (1) New model ensemble scheme A new model ensemble scheme called the Stochastic Humidity Profile for Convective parametrization (SHPC, Ota (2025)) was introduced to represent uncertainty in convective activity and mitigate underdispersiveness in the tropics. (2) Reduced amplitude of initial perturbations in the tropics The amplitude of initial perturbations made from Singular Vector (SV) data for the tropics was reduced to 60% of the previous level. (3) Recent GSM developments The forecast model was upgraded to a low-resolution version of the latest GSM (Kawaguchi et al. 2025). The required computational resources are reduced with enhancements in MPI-based parallelization and data transfer. (4) Revised SST perturbations SST perturbations were revised so that the ensemble mean SST coincides with the unperturbed SST for all lead times. 3. Verification To verify system performance for medium-range forecasts with lead times of up to 11 days, retrospective experiments covering periods of three months or more in summer 2023 and winter 2023/24 were conducted. The results showed enhanced spread-skill relationships and continuous ranked probability scores (CRPSs), especially in the tropics (20°S – 20°N). Figure 1 shows the spread normalized by the ensemble mean RMSE, and CRPSs of 250 hPa zonal wind speed (U250) for the tropics in summer. Brier skill scores for precipitation forecasts in Japan were also enhanced for summer (not shown). Hindcast experiments were also conducted for the 30-year period from 1991 to 2020 with atmospheric initial conditions from the latest Japanese reanalysis (JRA-3Q; Kosaka et al. 2024). Reduced amplitude of initial perturbations and implementation of the SHPC mitigate excessive and insufficient ensemble spreads, respectively (Figure 2). MJO forecast skill was almost neutral for all lead times (not shown). 36 Figure 1: Spread normalized by the ensemble mean RMSE (left) and CRPS (right; unit: m/s) of zonal wind speed at 250 hPa (U250) in the tropics (20°S – 20°N) for summer 2023. Green and red lines represent verification results for the previous (CNTL) and new (TEST) Global EPS. The purple line and yellow triangles in the panel on the right represent score change ratios ([TEST-CNTL]/CNTL, right axis; unit: %) and statistically significant enhancements at a 95% confidence level with bootstrap application, respectively. Figure 2: Ensemble spread (left; unit: x106 m2/s) and spread normalized by the ensemble mean RMSE (right) of velocity potential at 200 hPa in the tropics (20°S – 20°N) for winter 1991 – 2020. Black and red lines represent verification results for the previous (CNTL) and new (TEST) Global EPS, respectively. References Kawaguchi, M., T. Kinami, Y. Kuroki, N. Shimokawa, K. Sutou, H. Yonehara, and H. Yoshimura (2025). Upgrade of JMA’s Operational Global Numerical Weather Prediction System. Res. Activ. Earth Sys. Modell. (55), submitted. Kosaka, Y., S. Kobayashi, Y. Harada, C. Kobayashi, H. Naoe, K. Yoshimoto, M. Harada, N. Goto, J. Chiba, K. Miyaoka, R. Sekiguchi, M. Deushi, H. Kamahori, T. Nakaegawa, T. Y. Tanaka, T. Tokuhiro, Y. Sato, Y. Matsushita, and K. Onogi (2024). The JRA-3Q reanalysis. J. Meteor. Soc. Japan, 102, 49-109, https://doi.org/10.2151/jmsj.2024-004. Ota, Y. (2025). Introduction of Stochastic Humidity Profile for Convective parametrization (SHPC) method in JMA’s Global Ensemble Prediction System. Res. Activ. Earth Sys. Modell. (55), submitted. 37 Operational Use of AMSU-A and ATMS Window Channels in JMA’s NWP Systems URATA Tomoya and SHIMIZU Hiroyuki Japan Meteorological Agency Email: [email protected] 1. Introduction Radiance data collected by satellite-based microwave sounders mainly provide information on the vertical distribution of atmospheric temperature and humidity, as assimilated in JMA’s Global Analysis (GA), Meso-scale Analysis (MA) and Local Analysis (LA) systems. AMSU-A and ATMS microwave sounders are equipped with window channels that have higher transmittance in the atmosphere than sounding channels and are sensitive to humidity at lower-tropospheric levels. These channels are used only for Quality Control (QC), but are considered capable of contributing to more accurate weather forecast by enhancing analysis humidity fields. 2. Methodology 2.1. Channel Selection In this study, window channels with frequencies of 23.8 GHz (AMSU-A/ch1 and ATMS/ch1) and 31.4 GHz (AMSU-A/ch2, ATMS/ch2) were assimilated. 89 GHz channels (AMSU-A/ch15 and ATMS/ch16) are also sensitive to humidity at lower-tropospheric levels, but were not assimilated because of their higher sensitivity to ice clouds and precipitation, which complicates QC processing. 2.2. QC A clear-sky method was applied in which only data unaffected by clouds or precipitation are assimilated. The same cloud detection scheme used for channels sensitive to temperature at lower-tropospheric levels (Okamoto et al., 2005) using retrieved cloud liquid water and scattering index values was implemented. Only data from areas over oceans are assimilated in consideration of the greater uncertainty of surface emissivity and temperature in the first guess from areas over land and sea ice. Data from areas near land or sea ice are not used in order to avoid contamination of such surfaces within the ocean-based observation footprint. Data at the edge of scans where the observation footprint is larger are also excluded to avoid contamination. 2.3. Observation Errors Observation errors are defined for each satellite based on the standard deviation of O-B (Observation minus Background) calculated from 15 days of GA statistics (Table 1). These values are inflated in analysis systems because inter-channel and spatial observation error correlations are not considered in assimilation. Error inflation factors are set to match those for microwave imager channels whose frequencies are close to those of the window channels (3.0 for GA, 4.0 for MA and 6.0 for LA). 3. NWP Effects and Results Numerical experiments were conducted to evaluate the effects of window channels on the global, meso-scale and local NWP systems. Here, the results of experiments involving the meso-scale NWP system for July 2023 and January 2024 are reported. Window channel assimilation enhanced consistency between the first guess and humidity-sensitive observations such as AMSR2 and the humidity channels of ATMS and CrIS (Fig. 1), which implies better first-guess accuracy. The false alarm and miss ratios compared to radar/raingauge-analyzed precipitation data were reduced for relatively weak precipitation forecasts up to 5 mm/3 h for the summer period. Several case studies also indicated that additional information on lower-tropospheric 38 humidity provided by the window channels of AMSU-A and ATMS near precipitation areas contributed to better forecast accuracy (Fig. 2). Similar enhancements were observed in the global and local NWP systems. Based on these results, assimilation of the window channels of AMSU-A and ATMS was adopted in MA and LA in March 2025, and is scheduled for adoption in GA in autumn 2025. Satellite/Sensor ch1 ch2 Metop-C/AMSU-A 2.80 2.40 NOAA-15/AMSU-A 3.00 2.60 NOAA-18/AMSU-A 2.80 2.40 NOAA-19/AMSU-A 2.80 2.40 Suomi-NPP/ATMS 2.80 2.20 NOAA-20/ATMS 3.00 2.70 NOAA-21/ATMS 3.00 2.70 References Okamoto, K., Kazumori, M., and Owada, H. (2005). The assimilation of ATOVS radiances in the JMA Global Analysis System. J. Meteor. Soc. Japan, 83, 201-217. https://doi.org/10.2151/jmsj.83.201 Table 1: Window channel observation error settings [K] Fig. 1: Ratio of change [%] in the standard deviation of OB resulting from window channel assimilation in MA. (a) Microwave imagers: GMI, AMSR2 and SSMIS. (b) CrIS hyperspectral infrared sounder. The red and green lines represent experiment results for July 2023 and January 2024, respectively. Fig. 2: 3-hour cumulative precipitation [mm]. (a) Forecast without window channels. (b) Forecast with window channels. (c) Forecast difference resulting from window channel assimilation: (b) – (a). (d) Radar/raingauge-analyzed precipitation data. The forecast time is 3 hours from 15 UTC on 7th July 2023. 39 Implementing Isochoric-Isobaric Transformations in the MCV Model's Dynamic-Physical Coupling Dong Chen 1, Xingliang Li 1*, Xueshun Shen 1 1 CMA Earth System Modeling and Prediction Center (CEMC), Beijing 100081, China *Email: [email protected], che[email protected]ov.cn 1. Introduction Dynamic-physical coupling in atmospheric models enables accurate characterization of multi-scale atmospheric interactions through the integration of dynamic and physical processes. The design of the coupling strategy directly influences energy conservation, the efficiency of physical feedbacks, and the coordination of multi-scale interactions. The accuracy of coupling processes further affects the enhancement of simulation and the stability of runtime performance (Gross et al., 2018). Beljaars et al. (2004) demonstrated that improved numerical coupling mechanisms within dynamic-physical interactions lead to a reduction in prediction errors. The nextgeneration regional/global unified nonhydrostatic Multi-moment Constrained finite Volume (MCV) model by CMA (Li et al., 2013; Chen et al., 2014; 2023), achieves efficient dynamic-physical coupling via innovative framework design. However, current MCV model still exhibits inherent biases in isochoric-isobaric coordinate thermodynamic conversion within the coupling process, which may result in imbalances in temperature tendencies, biases in precipitation simulations, and reduced predictability scores. To address this issue, this study aims to implement isochoric-isobaric transformations within the dynamic-physical coupling strategy of the MCV model. 2. Method and results 2.1 Isochoric-isobaric coupling method The physics-dynamics coupling in MCV model differs from many existing atmospheric models in that it is performed at constant volume (density) rather than constant pressure. However, the physical parameterization schemes is built at pressure-based coordinates. Thermodynamically, this corresponds to the dynamic core being associated with an isochoric process and the physical processes with an isobaric process. During dynamics-physics coupling, the update of temperature tendencies must incorporate the conversion between these two distinct thermodynamic processes to ensure heat conservation. Previously, the MCV model’s coupling scheme lacked consideration of this conversion, resulting in biases in the potential temperature fed back to the dynamic core following physical process computations. This, in turn, led to systematic underestimation of total and grid-scale precipitation in tropical regions. To address this issue, we redesigned the temperature conversion based on energy conservation and thermodynamic principles. Remember that total heat must be strictly conserved, then the dynamical temperature tendency that fed back from physics becomes ∆Qdyn=∆Qphy ⟹∆Tdyn=cp cvm ∆Tphy (1) where Q represents the heat per mass, Tis temperature, cpand cvmare specific capacity at constant pressure and moist specific capacity at constant volume, respectively. 2.2 Results and analysis Case simulations implementing the dynamics-physics conversion formulation (1) demonstrate that, under identical resolution and initial conditions, total precipitation and grid-scale precipitation exhibit enhanced physical reasonableness (Fig. 1 a), with biases in temperature and humidity throughout the atmospheric column significantly reduced (figure 40 A Simplified Lake Model for Seasonal Prediction ADACHI Yukimasa, OCHI Kenta, HIRAHARA Shoji, KUBO Yutaro, SEKIGUCHI Ryohei Meteorological Research Institute / Japan Meteorological Agency, Tsukuba, Japan Email: [email protected] 1. Abstract The reproducibility of interannual variations in the lower atmosphere and the conditions of land surfaces with which it interacts are important in seasonal prediction. To enhance related data over previous JMA seasonal prediction systems (which relied on monthly climatology for lake-related variables such as surface water temperature and ice concentration), a lake model was introduced into the system (CPS3; Hirahara et al. 2023) and modified in the next iteration (CPS4; Kubo et al. 2025). As variability in air temperature above lakes is high in winter and closely associated with ice presence, particular emphasis is placed on the treatment of lake ice. Accordingly, the lake model predicts only lake surface conditions and lake ice variables required as the lower boundary for the atmospheric model, and is highly efficient in operational usage. Since there is no entirelake targeting, lake depth data are not required; only monthly climatology for surface temperatures (which are based on satellite observation and therefore considered highly accurate) are needed. The model helps to make data on the amplitude of interannual variations in surface air temperature over lakes more realistic. 2. Lake Model Overview Based on vertical column thermodynamics, the model predicts lake ice formation and variations in water temperature throughout water phase changes and heat transfer among water, ice, and snow. It consists of three layers of lake water, four layers of lake ice and one layer of snow on lake ice. Water, ice and snow densities are constant, and the water temperature in the bottom (third) layer is set as a constant annual-mean climatology value because lower water temperature changes little throughout the year. The energy of water’s upper two layers, lake ice and snow on lake ice is conserved, but the variability of lake water mass is not taken into account. The thickness of lake water layer is constant. The thermal diffusion coefficient between the first and second layers is set to be globally uniform as the inverse of an arbitrary relaxation time. This time between the second and third layers depends on the observed amplitude of the seasonal cycle in surface temperature, with shorter relaxation times associated with shallow lake conditions. The seasonal amplitude is defined as the difference between the maximum and minimum climatology monthly mean surface temperature based on MODIS/Terra Land Surface Temperature data (MOD11C3_v006; Wan, Z. et al, 2015). The bottom water temperature is set to a proportional value between the annual mean of the observed surface temperature and the temperature where the density of water reaches its maximum (3.98°C), as bottom water temperature is close to the annual minimum in non-frozen lakes and the maximum water density temperature in frozen lakes. The model uses a fixed water density, and the annual mean (rather than the annual minimum) is applied. Thus, it acts in a pendulum-like manner, unlike the thermostat behavior observed around the annual minimum value, or 3.98°C in nature. The number of lake ice layers is constant, and the upper and lower limits of ice thickness are set individually to keep the upper layers thinner than the lower ones. Lake ice concentration and thickness vary independently during freezing and melting, respectively, enabling representation of both wide/thin and narrow/thick conditions. 47 3. CPS4 Enhancements In CPS3, a lower limit of 0°C was set for the observed monthly mean temperature for calculation of annual mean and seasonal amplitudes. This limit is removed in CPS4, resulting in larger lake ice indications due to the lower annual mean temperature and larger seasonal amplitude in frozen lakes. Bottom water temperature has also been modified to the average of the annual mean and annual minimum, resulting in higher water temperature indications in ice-free lakes such as the Caspian Sea. 4. Results Figure 1 shows seasonal changes in interannual surface air temperature variation on the Great Lakes. Variability is large for Nov. – Mar. in reanalysis (JRA-55; Kobayashi et al. 2015). Application of the lake model enables appropriate lake ice representation, with amplitudes of interannual variation much closer to those of JRA-55 than those of climatology lake ice. Figure 2 shows seasonal surface air temperature biases relative to reanalysis (JRA-3Q; Kosaka et al. 2024) and the effects of model enhancement. Warm biases for North America and Siberia in boreal winter and cold biases in the Caspian Sea are reduced. References Hirahara, S. et al, 2023: Japan Meteorological Agency/Meteorological Research Institute Coupled Prediction System version 3 (JMA/MRI-CPS3). J. Meteor. Soc. Japan, 101, 149–169. Kobayashi, S. et al, 2015: The JRA-55 reanalysis: General specifications and basic characteristics. J. Meteor. Soc. Japan, 93, 5–18. Kosaka, Y. et al, 2024: The JRA-3Q reanalysis. J. Meteor. Soc. Japan, 102, 49–109, doi:10.2151/jmsj.2024-004. Kubo, Y. et al, 2025: Upgrade of the JMA Sub-Seasonal and Seasonal Ensemble Prediction System (JMA/MRI-CPS4). WGNE Res. Activ. Earth Sys. Model., submitted. Wan, Z. et al, 2015: MOD11C3 MODIS/Terra Land Surface Temperature/Emissivity Monthly L3 Global 0.05Deg CMG V006. NASA Land Processes Distributed Active Archive Center, doi:10.5067/MODIS/MOD11C3.006. (a) (b) Figure 1: Monthly interannual variation of surface air temperature on the Great Lakes in (a) climatology lake ice and (b) the lake model. Colored lines represent model results for initial data, and black lines represent JRA-55. (a) (c) (b) (d) Figure 2: (a) Climatological mean fields of surface air temperatures (contours) and biases relative to JRA-3Q (shading) with the current lake model, and (b) effects of enhancements in CPS4 for boreal summer. (c) and (d): as per (a) and (b) but for boreal winter. 48 Forecast verification; novel methodologies to diagnose and measure systematic errors 5 49 Power Spectra Verification of Machine Learning Weather Prediction Barbara Casati, Leo Separovic, Syed Z. Husain Meteorological Research Division, Environment and Climate Change Canada Email: [email protected] Efficient yet meaningful verification for comparing machine learning weather prediction (MLWP) against traditional physicsbased numerical weather prediction (NWP) is crucial for scientific credibility and operational uptake. Traditional summary statistics, like MSE, bias, variance or correlations alone, can be hedged by smoother forecasts or climatology-induced false skill (Hamill and Juras, 2006). Moreover, traditional metrics can mask unphysical behaviors and canceling errors. Power spectra diagnostics have been recently leveraged, to enable scale-aware physically meaningful evaluation of both MLWP and NWP models (e.g. Husain et al, 2025; Rodwell et al, 2025). This note wishes to highlight some of the key features of the power-spectra diagnostic approaches, with recommendations for operational verification practices. The diagnostic capabilities of the power-spectra verification statistics are illustrated for global forecasts from the Environment and Climate Change Canada’s (ECCC) operational NWP-based Global Deterministic Prediction System (GDPS) and two other ECCC experimental systems: the GEML (Global Environmental eMuLator) MLWP model, and the hybrid NWP-MLWP system based on Spectral Nudging (GDPS-SN). Predictions from all these systems are verified against the operational GDPS analysis. Figure 1: (Left) Forecast activity ratio as a function of lead time and (right) 120-h spectral activity ratio for 500-hPa geopotential height anomalies for (black) the GDPS analysis, (blue) GDPS, (red) GDPS-SN and (gray) GEML, averaged over 60 cases of JFM 2022. Climatology and anomalies: Atmospheric fields from the global forecasts and verifying analysis are separated into transient anomalies and stationary climatology, by subtracting the ERA5 30-year climatology. This enables focusing on weather phenomena and removing artificial skill due to persistent climatological features. Power Spectral Decomposition: Gridded atmospheric fields are decomposed spectrally using truncated spherical harmonic expansions (e.g., up to n=800 for ~0.25˚ resolution). This spectral representation enables evaluation as a function of total Figure 2: Spectral activity ratio of 72-h screen-level temperature for GDPS (blue) and GEML (red), averaged over 62 cases of JJA 2022. The ratio is shown for the stationary JJA 2022 sample climatology (dotted), transient anomalies (dashed) and the total/integral screen-level temperature fields (solid). 50 wavenumber (n), corresponding to physical scales (L = 2πR/n). Such scale decomposition is critical for determining the forecast "effective resolution" and diagnosing compensating errors at different scales. Furthermore, spectral decomposition permits the assessment of bias, error, and skill at different scales separately (e.g. Casati et al, 2023). In verification, bias and accuracy/skill are key and complementary attributes of the forecast quality: the bias compares forecastversus-observation marginal distributions, whereas the accuracy pertains to the forecast-observation joint distribution (Jolliffe and Stephenson, 2012). In the power spectra framework, these attributes can be assessed by computing forecast and analysis variances and their covariance. Activity and coherence: the activity, defined as the standard deviation of the anomalies, is associated with the amount (and intensity) of active weather (Ben Bouallègue et al, 2024). The forecast versus analysis activity ratio is a measure of the marginal bias in the forecast-versus-analysis variability, or of the amplitude error for the spherical harmonics. The anomaly correlation (named coherence for the power spectra) can provide a complementary measure of accuracy/skill (or phase error in spectral space). When evaluated for the spectral components, activity ratio and coherence can assess scale-dependent bias/amplitude errors and accuracy/skill for each spatial scale, separately. While in this short note we illustrate solely the activity ratio, measures of both bias and accuracy/skill should always be evaluated in tandem, to provide a more complete assessment of forecast performance (e.g. Figure 2 of Husain et al, 2025). Figure 1 shows that the seemingly almost neutral bias of GEML when the activity ratio is evaluated on the anomaly fields only (left), is affected by unphysical compensating effects on meso-α and meso-β scales, which are detected solely when using powerspectral diagnostics (right). Figure 2 shows that the spectral activity ratio captures GEML variance deficiency for mesoscales, typical of the overly smooth ML model outputs. Moreover, the stationary versus transient signal separation reveals that GEML resolves well the climatological (stationary) variability, but that this compensates for a large negative bias associated with the transient anomalies. This is an expected behavior for any MLWP model trained to minimize MSE, which leads to fine-scale smoothing to reduce the “double penalty” effect (Subich et al, 2025). In conclusion, for operational verification practices we recommend: i) computing both activity ratio and anomaly correlation to assess marginal bias/amplitude error and forecast skill/accuracy; ii) separate stationary climatology from transient anomalies, to distinguish skill in reproducing climate versus actual weather; iii) apply spectral decomposition to analyze separately the distinct model behaviors at different scales, to detect unphysical error compensation, and to assess the “effective resolution”, particularly for deterministic ML models that are prone to smoothing. References Ben Bouallègue, Z., M.C. A. Clare, L. Magnusson, E. Gascón, M. Maier-Gerber, M. Janoušek, M. Rodwell, F. Pinault, J. S. Dramsch,S.T. K. Lang, Baudouin Raoult, F. Rabier, M. Chevallier, I. Sandu, P. Dueben, M. Chantry, F. Pappenberger, 2024: “The Rise of Data-Driven Weather Forecasting”. Bulletin of the American Meteorological Society, E864-E883, https://doi.org/10.1175/BAMS-D-23-0162.1 Casati, B., C. Lussana, and A. Crespi, 2023: “Scale-separation diagnostics and the symmetric bounded efficiency for the intercomparison of precipitation reanalyses”. Int. J. Climatol., 43, 2287–2304, https://doi.org/10.1002/joc.7975 Hamill, T., and Juras J. , 2006: “Measuring forecast skill: Is it real skill or is it varying climatology?” Quart. J. Roy. Meteor. Soc., 132 , 2905– 2923, https://doi.org/10.1256/qj.06.25 Husain, S.Z., L. Separovic, J-F Caron, R. Aider, M. Buehner, S. Chamberland, E. Lapalme, R. McTaggart-Cowan, C. Subich, P.A. Vaillancourt, J. Yang, A. Zadra. 2025: “Leveraging Data-Driven Weather Models for Improving Numerical Weather Prediction Skill through Large-Scale Spectral Nudging”. Weather & Forecasting 40, n. 9: 1749-1771, https://doi.org/10.1175/WAF-D-24-0139.1 Jolliffe, I.T. and Stephenson, D.B. (2012) Forecast Verification: A Practitioner’s Guide in Atmospheric Science. 2nd Edition, Wiley-Blackwell, Oxford, ISBN:9781119960003Rodwell, M. J., M. C. A. Clare, S.-J. Lock, K. Lonitz, and M. Chevallier, 2025: “ Power Spectra of PhysicsBased and Data-Driven Ensembles.” Met. Apps. 32, no. 5: e70071. https://doi.org/10.1002/met.70071. Rodwell, M. J., M. C. A. Clare, S.-J. Lock, K. Lonitz, and M. Chevallier, 2025: “ Power Spectra of Physics-Based and Data-Driven Ensembles.” Met. Apps. 32, no. 5: e70071. https://doi.org/10.1002/met.70071. Subich, C., Husain, S.Z., Separovic, L., and Yang, J. (2025): Fixing the double penalty in data-driven weather forecasting through a modified spherical harmonic loss function. https://arxiv.org/abs/2501.19374 51 Uncoupled and coupled data assimilation for integrated earth system analysis and prediction; methodology and data impact sensitivity studies 6 52 Update of the Radiative Transfer Model to RTTOV-13 in JMA's NWP Systems SHIMIZU Hiroyuki, ANDO Akira, KAMEKAWA Norio, KONDO Keiichi, KUSANO Naoto, MURATA Hidehiko, TOYOKAWA Masakazu, URATA Tomoya Japan Meteorological Agency E-mail: shimizu_[email protected].go.jp 1. Introduction Various types of satellite radiance data are assimilated to determine accurate initial conditions for numerical weather prediction (NWP) models. The RTTOV fast radiative transfer model (Eyre 1991), developed by the EUMETSAT Satellite Application Facility on Numerical Weather Prediction (NWP SAF), is used as the observation operator for assimilating satellite radiance data in JMA’s NWP systems (Global, meso-scale and local NWP). As a first step in updating RTTOV10.2 (Saunders et al. 2012) to RTTOV-13 (Saunders et al. 2020), the minimal changes necessary to ensure secure implementation with no significant impact on NWP accuracy (such as changes to modules and relevant file formats) were made in 2022. The optical depth coefficient files (referred to here simply as “coefficient files”) and sea surface emissivity models in RTTOV were subsequently updated to the new versions, and relevant quality control procedures were adjusted accordingly. This report briefly summarizes the effects of the changes on NWP systems. 2. Modification of Quality Control Procedures The coefficient files for sensors already in use were updated to Version 13 other than for CrIS, which showed no enhancement from coefficient file updating. Sea surface emissivity models (FASTEM-6 (Kazumori and English 2015) for microwave observation and IREMIS for infrared observation) are now available in RTTOV13. FASTEM-6 models the dependence of sea surface emissivity on relative wind direction (RWD) more appropriately. Global analysis indicated smaller O-B (observation minus background; i.e., the difference between observed and calculated brightness temperature) biases in microwave imagers dependent on RWD. However, biases dependent on surface wind speed were larger for surface-sensitive channels of AMSU-A and ATMS. Accordingly, surface wind speed was added as a predictor in variational bias correction (VarBC) for ch4 and ch5 of AMSU-A and ch6 of ATMS. FASTEM-6 was not adopted in meso-scale and local analysis because no enhancement was observed in relation to a FASTEM update from version 4 to 6 and addition of surface wind speed as a predictor in VarBC. The statistical characteristics of calculated brightness temperature changed as a result of the updates to the coefficient files and the sea surface emissivity models described above. The static scan bias correction value and the precipitation screening parameters for AMSU-A, which were estimated in advance from statistics on observed and calculated brightness temperature, were updated for the new calculation characteristics. Retrieved cloud water amounts for AMSU-A quality control are also larger than before, mainly due to the updating of scan bias correction values. As a result, the amount of data considered to represent cloud increased, and the amount of data used decreased. In reference to O-B statistics, the threshold of cloud water content used for cloud detection was increased from 100 to 120 g/m2 to make the number of data used comparable to that before the update. 3. Effects of Updating Coefficient Files and Sea Surface Emissivity Models on NWP Systems Data assimilation experiments were conducted to evaluate the effects of updating the coefficient files and sea surface emissivity models to the new versions in JMA’s NWP systems. This report details the results of experiments in the global NWP system for the periods from June to October 2023 and November 2023 to March 2024. The control experiments (CNTL) had the same configuration as JMA’s operational global NWP system as of March 2025, and the test experiments (TEST) were as per CNTL except for the modifications described in Section 2. Figure 1 shows changes in the standard deviation and mean of FG departures (another expression of OB) for radiosonde temperature against CNTL. Those of TEST were closer to observations than CNTL, which indicates an enhanced first guess for the temperature field. Meanwhile, negative biases in 500 hPa 53 geopotential height against radiosonde observations over the tropics in CNTL showed a slightly increased tendency in TEST (Figure 2). This change corresponded closely to that in the O-B of ATMS ch6 (not shown), which was likely the result of a bug fix affecting sensors that have channels influenced by the 184 GHz ozone absorption line (Saunders et al. 2017). The results for precipitation forecasts and tropical cyclone track and intensity forecasts were almost neutral. Experiments with meso-scale and local NWP systems showed no significant changes in forecast accuracy. 4. Summary Updating for the coefficient files and sea surface emissivity models of RTTOV to the new version in JMA’s NWP systems showed an almost neutral effect on NWP, while stratospheric temperature fields were enhanced in the global NWP system. The modification was implemented in the meso-scale and local NWP systems in March 2025, and is scheduled for implementation in the global NWP system in autumn 2025. Figure 1: (a) Ratios of change [%] in the standard deviation of FG departures for radiosonde temperature observation. Error bars show confidence levels of 95%. (b) Mean of FG departures [K] for radiosonde temperature observation. Solid and dotted lines represent the results of TEST and CNTL experiments. Red and green lines represent the results of summer and winter experiments. Figure 2: Mean error of 500 hPa geopotential height [gpm] in 1-day forecasting against radiosonde observations. (a) represents CNTL results, and (b) represents differences between TEST and CNTL. References Eyre, J. R., 1991: A fast radiative transfer model for satellite sounding systems. ECMWF Tech. Memo., 176, 28pp. Kazumori, M. and S. J. English, 2015: Use of the ocean surface wind direction signal in microwave radiance assimilation. Quart. J. Roy. Meteor. Soc., 141, 1354–1375, doi:10.1002/qj.2445. Saunders, R., J. Hocking, P. Rayer, M. Matricardi, A. Geer, N. Bormann, P. Brunel, F. Karbou, and F. Aires, 2012: RTTOV-10 science and validation report. Tech. rep., EUMETSAT NWP SAF, 31 pp. https://nwpsaf.eumetsat.int/oldsite/deliverables/rtm/rtm_rttov10.html. Saunders, R., J. Hocking, D. Rundle, P. Rayer, S. Havemann, M. Matricardi, A. Geer, C. Lupu, P. Brunel, J. Vidot, 2017: RTTOV-12 science and validation report. Tech. Rep., EUMETSAT NWP SAF, URL https://nwp-saf.eumetsat.int/site/download/documentation/rtm/docs_rttov12/rttov12_svr.pdf. Saunders, R., J. Hocking, E. Turner, S. Havemann, A. Geer, C. Lupu, J. Vidot, P. Chambon, C. Köpken-Watts, L. Scheck, O. Stiller, C. Stumpf, E. Borbas, 2020: RTTOV-13 science and validation report. Tech. rep., EU METSAT NWP SAF, URL https://nwp-saf.eumetsat.int/site/download/documentation/rtm/docs_rttov13/rtto v13_svr.pdf. 54 Updating Surface Humidity Observation Error in JMA's Regional NWP Systems SAITO Riku, TOGUCHI Ryo Japan Meteorological Agency Email: riku.s[email protected]p 1. Introduction Water vapor in the lower troposphere has a significant influence on the occurrence and development of torrential rainfall caused by stationary linear mesoscale convective systems. Accordingly, assimilation of observation data that capture the inflow of such vapor is critical for optimal torrential-rain forecasting. In this context, the Japan Meteorological Agency (JMA) began progressively installing hygrometers at its Automated Meteorological Data Acquisition System (AMeDAS) facilities on a national basis in March 2020. JMA assimilates screen-level relative humidity as observed by these hygrometers into its regional NWP systems (mesoscale and local; JMA, 2025). As statistical examination of these hygrometers indicated a tendency for reduced observation accuracy under high-humidity conditions, JMA applied modification to increase the observation error of surface specific humidity under these conditions to its regional NWP systems in February 2025. This report describes findings on hygrometer data quality and observation error characteristics, along with related effects on regional NWP systems. 2. Hygrometer Data Quality The Rotronic MP-102H and Vaisala HMP155 hygrometers deployed at AMeDAS stations differ in measurement characteristics due to structural variations, such as heater functionality and distinct internal processing specifications. Statistical comparison of humidity data from these instruments showed that the MP102H frequently records 100% relative humidity (RH), whereas the HMP155 rarely does so (Fig. 1). This is often seen with the MP-102H at stations near rice paddies or other consistently humid areas. In contrast, the HMP155 rarely exceeds 97% RH, even during rainfall or similar conditions. In dry conditions, such as sunny daytime periods, both instruments show similar readings. These differences suggest that the data characteristics of AMeDAS hygrometers vary significantly in high-humidity conditions, indicating larger observation uncertainties. 3. Update of Observation Error Settings As AMeDAS hygrometers still provide useful data despite lower accuracy in humid conditions, it is deemed appropriate to utilize these data while accounting for the shortcomings. Accordingly, a method is applied to increase the observation error of surface specific humidity when RH exceeds 90%. The error is currently set as 0.7 – 0.75 g/kg (equivalent to approximately 3.6 – 3.8% RH at 25°C) for the regional NWP system. In the mesoscale NWP system, the standard deviation of climatological forecast error for the first model layer over land during summer is approximately 5% RH. Considering the larger observation error in high-humidity conditions and the representativeness error of water vapor, the observation error at 100% RH was set at three times the current value – roughly twice the model forecast error. The observation error increases in linear progression from the current value at 90% to three times the current value at 100%. 55 4. Effects on Analysis and Prediction To evaluate the effects of the updated observation error settings on regional NWP systems, numerical experiments were conducted based on the system as of March 2023 (CNTL) with updated observation error settings (TEST) for June 26 – July 25, 2023. In the mesoscale NWP system, increasing the observation error in high-humidity conditions mitigated the initial bias of surface specific humidity and reduced root mean square errors (RMSEs) (Fig. 2). In precipitation prediction, false alarms for light rain decreased, leading to enhanced equitable threat scores (ETS). In the local NWP system, the initial dry bias of surface specific humidity was mitigated, while effects on precipitation and other elements were small. 5. Summary The study suggested that JMA’s AMeDAS hygrometers exhibit lower accuracy in high humidity. By increasing the observation error settings for such conditions, accuracy for surface specific humidity and precipitation forecasts in the mesoscale NWP system was enhanced. JMA implemented these changes in its regional NWP systems in February 2025. Figure 1: Observation frequency distribution of relative humidity by instrument type based on statistics from March 10 2023 to February 29 2024. (a) Rotronic MP-102H (157 stations), (b) Vaisala HMP155 (208 stations). Figure 2: Mean errors and RMSEs for observation of specific humidity [g/kg] over the experiment period of June 26 – July 25 2023 in the mesoscale NWP system. Blue lines: CNTL; red lines: TEST. References Japan Meteorological Agency, 2025: Data Assimilation Systems, OUTLINE OF THE OPERATIONAL NUMERICAL WEATHER PREDICTION AT THE JAPAN METEOROLOGICAL AGENCY, 13-56. 56 Comparison of annual average climatology of key oceanographic parameters in the Bay of Bengal and Arabian Sea N. Gopikrishna1, Sujata K. Mandke1, *, Susmitha Joseph1 1 Indian Institute of Tropical Meteorology, Ministry of earth Sciences, India. Email: *[email protected] 1. Introduction The Bay of Bengal (BoB) and Arabian Sea (AS) are adjacent sub-basins of the north Indian Ocean (NIO), with markedly different upper ocean stratification due to contrasting freshwater inputs and wind-driven mixing (5). The thermohaline structure of the NIO plays a pivotal role in regulating air–sea heat and moisture exchanges, which in turn influence the onset, intensity, and variability of the Indian monsoon (4, 5). To enhance our understanding of upperocean stability, vertical mixing processes, and their coupling with atmospheric circulation, a comprehensive climatology of parameters such as sea surface temperature (SST), salinity, mixed layer depth (MLD), isothermal layer depth (ILD), and barrier layer thickness (BLT) over the NIO is an essential prerequisite for the modelling community in developing ocean–atmosphere coupled models. In the present study, we compare the annual climatological thermohaline structure of BoB and AS. 2. Data and Methodology Monthly gridded SST and salinity data from the ARMOR3D dataset (2) (Copernicus Marine Service) are used. The MLD is directly obtained, while the ILD is calculated as the depth where temperature decreases by 0.2 °C from the 10 m reference depth (Boyer et al., 2004). The BLT is then computed as (ILD − MLD) (1&6). Annual climatology is estimated by averaging the annual means of the variable over the period 1993–2022. 3. Results and Discussion A comparative description of the annual climatology of SST, salinity, MLD, ILD, and BLT, considering their vertical, latitudinal, and longitudinal variations over the BoB and AS is provided below. 3.1 Vertical T–S Structure (Fig. 1) the annual climatology of surface salinity is lower in BOB than AS, as freshwater input through rainfall and river discharge markedly exceeds evaporation in BOB. In contrast, the evaporation exceeds precipitation and river discharge is limited in AS. The salinity increases with depth, showing a rapid rise in the upper 100 meters in BOB. In contrast, a slow increase to a subsurface maximum at ~75 meters, then a very gradual decrease below in AS. The annual temperature climatology is slightly warmer in BOB than AS at the surface, gradually decreasing with depth in both basins. This vertical T-S structure leads to a higher Brunt–Väisälä frequency in the BoB due to high surface layer stratification, supporting the development of a stable barrier layer(BL), that inhibits vertical mixing (3,6). The annual climatology of BOB and AS highlights following differences (corresponding values for AS are presented in parentheses): BLT: ~20 m (8m); MLD: ~20 m (29.6m); ILD: ~35 m (29.5m). 3.2 Latitudinal variation (Fig. 2) zonal average over the BoB and AS depicts SST’s are warmer and surface salinity is lower in BOB than AS, north of 20ºN, whereas, BoB SST’s are cooler than the AS between 0°-20°N. The MLD remains shallower in the BoB than in the AS, while the ILD extends deeper in both basins, except from 0°-4°N. Notably, near the equator (0-5°N), MLD is deeper in BOB than the AS due to weaker freshwater stratification and stronger wind-driven mixing (3, 4, 6, and 8). BLT exhibits strong latitudinal variation in BOB, exceeding 20 m between 15–20°N due to freshwater input that enhances stratification, suppresses mixing, and results in a shallow MLD. Conversely, lower freshwater and stronger mixing in AS maintain a shallow BLT (0-10 m) across most latitudes, similar to earlier findings (1, 6). However, north of 20°N in the BoB, both the ILD and BLT approach zero due to limited bathymetric depth in near-coastal regions, where mixing dominates over stratification. The latitudinal gradient of SST, BLT, ILD and surface salinity are much stronger north of 20°N in BOB than AS, due large inflow of freshwater from precipitation and river runoff. 3.3 Longitudinal variation (Fig. 3) the meridional-averaged variation illustrate that SST’s in BOB are warmer and surface salinity is less than in AS over all longitudes. SST’s gradually increases from west to east in BoB 63 (5), while in AS, SST decreases from west to east in the western (40°-50°E) and eastern (72°-78°E) basin, while increases from west to east in the central 50°-72°E basin. The western AS (near 45°-50°E) is notably cooler than the eastern, due to strong winds causing upwelling in the summer monsoon (5). Surface salinity decreases eastward in BOB owing to freshwater input (4, 8). The AS is characterized by high surface salinity throughout, with slightly higher salinity in the east than in the west. The BoB is charecterized by a distinct pattern: cooler (warmer) SSTs in the western (eastern) BoB correspond to higher (lower) salinity - a signature of freshwater-induced stratification. The ILD also deepens eastward, but variation with longitude is less than the MLD, resulting in a pronounced BL (BLT up to ~20-30 m) in the BoB and a very shallow BLT in the AS (6). 4. Conclusion In the vertical T–S structure, the BoB has lower surface salinity and higher surface temperature than the AS, with salinity increasing and temperature decreasing more sharply with depth in the BoB. Across most latitudes, the MLD is shallower and the BLT is thicker in the BOB than in the AS, associated with lower surface salinity that enhances stratification and suppresses vertical mixing. The BOB exhibits a pronounced east-west contrast, with lower SST and higher salinity in the west than east. This zonal asymmetry in the thermal-haline structure promotes the formation of a substantially thicker BL (20-30 m) in the BoB. In contrast, the AS remains well-mixed with a consistently shallow BLT across all longitudes, lacking a pronounced BL. These results provide an updated baseline for understanding surface ocean stratification in the NIO. References 1. Boyer, T. P., et al. (2004). Linear trends in salinity for the world ocean, 1955–1998. Geophysical Research Letters. 2. Guinehut, S., Drecourt, J. P., Le Traon, P. Y., & Larnicol, G. (2012). Validation of large‐scale high‐resolution sea surface temperature fields from ARMOR3D. Journal of Atmospheric and Oceanic Technology. 3. Lukas, R., & Lindstrom, E. (1991). The mixed layer of the western equatorial Pacific Ocean. Journal of Geophysical Research. 4. Rao, R. R., & Sivakumar, R. (2003). Seasonal variability of sea surface salinity and salt budget of the mixed layer of the north Indian Ocean. Journal of Geophysical Research: Oceans. 5. Shenoi, S. S. C., Shankar, D., & Shetye, S. R. (2002). Differences in heat budgets of the near-surface Arabian Sea and Bay of Bengal: Implications for the summer monsoon. Journal of Geophysical Research: Oceans. 6. Sprintall, J., & Tomczak, M. (1992). Evidence of the barrier layer in the surface layer of the tropics. Journal of Geophysical Research: Oceans. 7. Thadathil, P., Ghosh, A. K., Ramesh Kumar, M. R., & Ravichandran, M. (2007). Surface layer temperature inversion in the Bay of Bengal during winter. Geophysical Research Letters. 8. Vinayachandran, P. N., Murty, V. S. N., & Ramesh Babu, V. (2002). Observations of barrier layer formation in the Bay of Bengal during summer monsoon. Journal of Geophysical Research: Oceans. Fig. 1: Annual climatological vertical variation of temperature, Salinity, MLD, ILD, and BLT averaged over BoB (solid) and AS (dashed). Fig. 2: Annual climatological zonal variation of SST, surface Salinity, MLD, ILD, and BLT averaged over the BoB (solid) and AS (dashed). Fig. 3: Annual climatological meridional variation of SST, Salinity, MLD, ILD, and BLT, averaged over the BoB (solid) and AS (dashed). 64 Annual cycle of surface salinity, SST, precipitation, MLD and BLT: Comparison between Bay of Bengal and Arabian Sea N. Gopikrishna1, Sujata K. Mandke1, *, Sushmita Joseph1 1 Indian Institute of Tropical Meteorology, Ministry of earth Sciences, India. Email: *[email protected] 1. Introduction Seasonal variations in the North Indian Ocean (NIO) are driven by the interplay of sea surface salinity (SSS), sea surface temperature (SST), precipitation, and mixed layer dynamics, which together shape air–sea interactions and climate feedbacks. The Bay of Bengal (BoB) and Arabian Sea (AS), though part of the same monsoon system, respond differently to these drivers because of their contrasting freshwater inputs, evaporation–precipitation balance, and wind-driven mixing (Shenoi et al., 2002). These processes critically influence upper-ocean stratification, mixed layer depth (MLD), and barrier layer thickness (BLT), which are central to monsoon variability and regional climate (Rao & Sivakumar, 2003; de Boyer Montégut et al., 2007). This study examines the climatological annual cycle of key oceanic variables across BoB and AS, the two sub-regions of the NIO, using high-resolution satellite and reanalysis datasets, with emphasis on the seasonal evolution of MLD and BLT. 2. Data and Methodology The weekly climatology is constructed using the following datasets: SST: Daily 0.25° × 0.25° NOAA OISST v2.1, Precipitation: Daily 1° × 1° Global Precipitation Climatology Project (GPCP). Salinity and MLD: Weekly ARMOR3D Level 4 reanalysis (Copernicus Marine Service), and BLT: Computed as isothermal layer depth (ILD) − MLD, following the Boyer et al. (2004). ILD is estimated as the depth at which temperature decreases by 0.2 °C from the 10 m reference depth. Two regions are defined for spatial averaging: 1. BoB: - 78°– 100°E, 0°–25°N, 2. AS: - 40°–78°E, 0°–25°N. Each dataset is first averaged into weekly means for each year. Weekly climatology for each of the 53 weeks is then estimated by averaging weekly means across the entire period from 1993-2023. 3. Results and Discussion The annual cycle of SSS, SST, precipitation, MLD and BLT over BoB and AS, based on the weekly climatology as illustrated in Figure 1, are compared and discussed below. SSS: - In the BoB, SSS peaks during the pre-monsoon season, and then declines sharply as the summer monsoon intensifies due to heavy rainfall and river discharge (Shenoi et al. 2002), reaching a minimum in the post-monsoon. The AS remains more saline than the BoB year-round. The minimum SSS in AS is during the pre-monsoon, then increases through the monsoon season as evaporation exceeds precipitation, peaking in the post-monsoon when precipitation subsides and persists till winter. SST: In the BoB, SST has annual minimum during winter and maximum in pre-monsoon. During the monsoon, SST declines gradually due to reduced solar insolation from increased cloud cover, enhanced latent heat loss from evaporation, and surface cooling from heavy precipitation (Rao & Sivakumar, 2003), and warm SST >28°C persists till post monsoon. Evaporation rates in the BoB are lower than in the AS, and the shallow MLD limits vertical mixing (Shenoi et al., 2002), facilitating relatively warm and stable SSTs through the post-monsoon. In the AS, SST has two peaks, one primary peak in the pre-monsoon and secondary peak during post monsoon, followed by minimum in winter. Stronger winds during the summer monsoon season cause deeper mixing, and substantially higher evaporation rates in AS than in the BoB, resulting in greater latent heat loss and a more pronounced cooling (Shenoi et al 2002). Precipitation: - The precipitation increases sharply with the onset of the summer monsoon in late April to first week of May and maintained till post monsoon in both BoB and AS. SSTs above convective threshold (>28 °C) enhance evaporation causing increased atmospheric moisture, thereby creating favourable conditions for convection and rainfall (Gadgil, 2003). In the BoB, low SSS due to freshwater input strengthens surface stratification, and shallow MLD which in turn maintains these 65 warm SSTs during post monsoon seasons, resulting in more precipitation (Sprintall & Tomczak, 1992). A thick BLT reinforces this stability by trapping heat and moisture in the upper layer, providing sustained energy for deep convection and heavy precipitation (Sprintall & Tomczak, 1992; Thadathil et al., 2007). Thus, precipitation is more in BoB than AS throughout the year. MLD: - In the BoB, the MLD is shallower than in the AS throughout the year. The seasonal evolution of the MLD in BoB is characterized by shoaling in winter and post monsoon, at its minimum in the pre-monsoon and subsequent deepening in the monsoon. However, the concurrent increase in BLT indicates that freshwater stratification from rainfall and river discharge keeps the effective MLD shallow (Sprintall & Tomczak, 1992; de Boyer Montégut 2007). In the AS, the MLD is deepest in winter due to convective overturning and surface cooling, shoals in the pre-monsoon, and then deepens again during the monsoon under wind-driven mixing (Shenoi et al 2002). This deep mixing entrains cooler, saltier subsurface water, lowering SSTs and reducing atmospheric moisture availability for rainfall (Schott et al., 2009). BLT: - In the BoB, BLT evolves seasonally with a spring minimum (April–May), late-winter maximum (January–February), and a summer– early winter transition (June–December). A thick BLT insulates the surface from cooler thermocline waters, sustaining warm SSTs and enhancing monsoon rainfall (Thadathil et al., 2007). BLT remains very shallow throughout the year in AS than BoB. The BLT in AS features annual maxima during late winter and remains shallow rest of the year; the shallow BLT is insufficient to significantly alter SST or precipitation (Shenoi et al., 2002). The seasonal cycle of BLT is driven by forcing mechanism of MLD and ILD variability (ILD is controlled by Ekman pumping, net surface heating and propagating long waves. MLD is forced by transfer of turbulent fluxes of heat, mass, and momentum and surface circulations) (Thadathil et al., 2007). Conclusion The seasonal evolution of most parameters is broadly similar across the two basins, but there is a large difference in the seasonal cycle of SSS, driven by freshwater input and surface circulation (Shenoi et al., 2002). In the BoB, low SSS combined with a shallow MLD and a thick BLT result in strong upper-ocean stratification that traps heat and moisture. This stability maintains SSTs above the convective threshold, supporting intense and prolonged precipitation. In contrast, the AS is characterized by higher salinity, a deeper MLD, a thinner BLT, and stronger wind-driven mixing (Thadathil et al., 2007), which together promote cooler SSTs and consequently weaker precipitation. The stark contrast between the stratified BoB and the well-mixed AS underscores the critical role of upper-ocean processes in modulating monsoon precipitation. References 1. Boyer, T. P., et al. (2004). Linear trends in salinity for the world ocean, 1955–1998. GRL, 31(1), L01303. 2. De Boyer Montégut, C., et al. (2007). Simulated seasonal and interannual variability of mixed layer depths in the tropical Indian Ocean. JGR: Oceans, 112(C6). 3. Gadgil, S. (2003). The Indian monsoon and its variability. Annual Review of Earth and Planetary Sciences, 31, 429–467. 4. Shenoi, S. S. C., Shankar, D., & Shetye, S. R. (2002). Differences in heat budgets of the near-surface Arabian Sea and Bay of Bengal: Implications for the summer monsoon. JGR: Oceans, 107(C7). 5. Sprintall, J., & Tomczak, M. (1992). Evidence of the barrier layer in the surface layer of the tropics. JGR: Oceans, 97(C5). 6. Thadathil, P., et al. (2007). Surface layer temperature inversion in the Bay of Bengal during winter. GRL, 34(3). Fig 1: Annual cycle of SSS, SST, Precipitation, MLD and BLT averaged over BoB and AS, depicted using weekly climatology (53 weeks) 66 Conventional Observation Reanalysis (CORe) Datasets Li Zhang2, Wesley Ebisuzaki1, Arun Kumar2, and Wanqiu Wang1 1NOAA/NWS/NCEP/CPC, 2ERT at NOAA/NWS/NCEP/CPC Email:[email protected] Introduction The Conventional Observation Reanalysis (CORe, Ebisuzaki et al., 2020) is planned to become an operational system at the National Centers for Environmental Prediction (NCEP) in the calendar year of 2025. It is intended to replace the long-running NCEP/NCAR Reanalysis (Kalnay et al., 1996), which is currently used by the Climate Prediction Center (CPC) at NCEP for real-time climate monitoring. CORe is a global atmospheric reanalysis covering the period from 1950 to the present. It’s designed specifically to get better long-term consistency and trends by assimilating only conventional (in situ) observations and Atmospheric Motion Vectors (AMVs). This focus supports CPC’s climate monitoring activities. Data Assimilation System CORe uses an Ensemble Kalman Filter, specifically the Local Ensemble Transform Kalman Filter (LETKF). The system has an ensemble of 80 members and uses incremental updates. For the forecast model, CORe uses a C128, 64 level cubed sphere model (FV3GFS, v15) with an approximate resolution of 0.7 degrees. A unique feature of CORe among reanalyses is its use of a non-hydrostatic model, which may improve the atmospheric representation of certain physical processes such as breaking gravity waves. Results The top panel of Figure 1 shows global temperature anomaly from CORe as a function of pressure (1000 hPa to 10 hPa, y-axis) and year (x-axis), relative to the 1991-2020 climatology. The troposphere exhibits a gradual warming trend, while the stratosphere shows a gradual cooling trend. The warm spikes in the stratosphere correspond to effects of volcanic eruptions which increase the stratospheric aerosols. The middle panel is the same as the top panel except for the JRA-3Q reanalysis (Kosaka et al, 2024), and the bottom panel for ERA-5 (Hersbach et al., 2020). All three reanalyses display quite similar trends, with JRA-3Q exhibiting slightly smaller stratospheric warming from the volcanoes. These comparisons demonstrate a high level of agreement among the latest reanalyses. Figure 1. Global temperature anomalies for CORe, JRA-3Q and ERA-5. Figure 2. 10S-10N precipitation anomalies for DYNAMO period for observations, R1, R2, CFSR and CORe. 67 Figure 2 displays the 10°S-10°N precipitation anomalies during the Dynamics of the MJO (DYNAMO) campaign period from (a) observations (Xie et al., 2017), (b) NCEP/NCAR Reanalysis (R1), (c) NCEP/DOE Reanalysis (R2, Kanamitsu et al., 2002), (d) Climate Forecast System Reanalysis (CFSR, Saha et al., 2010) and (e) CORe. Notably, CORe, despite not using satellite radiance data, successfully captures the Madden-Julian Oscillation (MJO) precipitation signal over the Indian and Western Pacific Oceans, while the older R1 and R2 systems, despite assimilating satellite retrievals, show weaker MJO signals. Datasets The primary CORe data archive contains ensemble means of the atmospheric and land surface state. The analyses are on pressure-level, model levels and special layers. These data are in GRIB version 2 (grib2) format on a 512x256 Gaussian grid which is the data assimilation grid. The secondary level archive includes ensemble members and statistics as well as the observation increments and a subset of model restart files. These files are in grib2, NEMSIO or NetCDF formats. Cloud Archive The main distribution of CORe analyses will be via the NOAA Open Data Dissemination (NODD) cloud platform. The full archive will eventually cover January 1950 to present with a 2-3 day latency. As of September 2025, the archive includes data from January 1950 to December 2021. Available products include 3-hourly, daily and monthly ensemble mean analyses. The archive is expected to expand after CORe becomes operational. These datasets are in grib2 format, and conversion to other grids can be easily done using programs, such as wgrib2, and pygrib. References Ebisuzaki W., and coauthors (2020). A Conventional Observation Reanalysis (CORe) for Climate Monitoring. Science and Technology Infusion Climate Bulletin, 44th NOAA Annual Climate Diagnostics and Prediction Workshop, Durham, NC 22-24 October, 2019, https://doi.org/10.25923/t4qa-ae63. Hersbach, H., and coauthors (2020). The ERA5 Global Reanalysis. Quarterly Journal of the Royal Meteorological Society 146 (730), https://doi.org/ 10.1002/qj.3803. Kalnay, E., and coauthors (1996). The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc. 77(3), 437-472, https://doi.org/10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2 Kanamitusu, M., and coauthors (2002). NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc. 83(11), 1631-1644, https://doi.org/10.1175/BAMS-83-11-1631. Kosaka, Y., and coauthors (2024). The JRA-3Q Reanalysis. J. Meteor. Soc. Japan (102), 49-109. https://doi.org/10.2151/jmsj.2024-004. Saha, S., and coauthors (2010). The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc. (91), 1015–1057, https://doi.org/10.1175/2010BAMS3001.1. Xie, P., and coauthors (2017). Reprocessed, Bias-Corrected CMORPH Global High-Resolution Precipitation Estimates from 1998. Journal of Hydrometeorology (18), 1617-1641, https://doi.org/10.1175/JHM-D-16-0168.1 68 Numerical/ computational techniques and model resolution, physics-dynamics and physics-physics crosscomponent coupling 9 69 Development of Regional Short-range Prediction Systems Based on KIM Eun-Hee Lee, Junghan Kim, Heeje Cho, Soo Ya Bae and Kyung-Hee Seol Korea Institute of Atmospheric Prediction Systems, Seoul, South Korea Email: [email protected] 1. Development of short-range prediction systems based on KIM The Korean Integrated Model (KIM) was originally designed for the global medium-range forecast. Since its adoption as the operational weather prediction model of the Korea Meteorological Administration (KMA), the Korea Institute of Atmospheric Prediction Systems (KIAPS) has expanded its applications to support seamless prediction across scales. A recent advancement is the development of short-range prediction systems by implementing two model functionalities: the variable-resolution grid system and the limited-area model (LAM) configuration. These short-range prediction systems share core components with the global uniform-resolution KIM, including the non-hydrostatic dynamic core which utilizes the spectral element method on a cubedsphere grid and the scale-aware physics. 1.1 Variable Resolution grid The variable-resolution version of KIM was developed using the Schmidt transformation, with increasing horizontal resolutions around the target area. The cubed-sphere framework was extended through an updated transformation matrix, and the time step and numerical viscosity coefficients were adjusted to maintain consistent numerical behavior across resolutions. The degree of grid system deformation is controlled by a stretching factor, S, relative to a reference uniform-resolution grid (S=1). Because explicit time integration schemes are limited by the smallest grid spacing, the time step is proportionally reduced by a factor of 1/S. Evaluation results indicate that the variable-resolution KIM provides forecast skill comparable to highresolution uniform-grid simulations within the targeted domain and significantly outperforms lower-resolution simulations. Figure 1 illustrates various grid configurations of stretched global grids with different S factors. 1.2 Limited Area Model The limited-area (regional) version of KIM was developed by constructing a modelling framework that allows the model to operate on a single panel of the cubed sphere. Because the LAM configuration can be employed with the variable-resolution grid system, our “single-panel” approach is able to narrow the LAM domain by setting the stretching factor S>1. Figure 1 illustrates the domains with different S values. Lateral boundary conditions are specified at all Gauss-Lobatto-Legendre (GLL) points of the outermost elements and are derived through spatial and temporal interpolation from coarser-resolution global KIM simulations. To enhance simulation stability, a relaxation (or nudging) term is applied near the domain boundaries. This term gradually blends the regional model solution toward the background field and follows the same functional form used in WRF and MPAS. 2. Semi-real time operation and performance The latest upgrade of KIAPS KIM includes utilities for two regional-scale prediction systems. KIAPS has initiated the internal semi-real-time testing for two short-range prediction systems: (1) a variable-resolution global grid targeting a 3 km resolution over East Asia (NE1=576, S=2.5), and (2) a limited-area grid targeting 1 Number of elements in each panel of the cubed-sphere. Each element has 3×3 GLL grid points for all model configurations mentioned here. 70 a 1 km resolution over the Korean Peninsula (NE=768, S=5.0). These systems operate in parallel with the global medium-range forecast system with uniform 8-km resolution grid (NE=576) which provides initial and boundary conditions for the two short-range prediction systems. All three systems share the same physics, supporting scale-aware features like convection schemes and gravity wave drag parameterizations. Note that all configurations use the same vertical discretization and resolution. The two short-range prediction systems do not have data assimilation systems, but the 8 km background fields provide observational information. Preliminary evaluations during the summer monsoon season show that the very-high-resolution systems capture intense precipitation signals for heavy rainfall events, closely matching observations. This highlights the potential of regional ~km resolution systems for reliable forecasting and seamless predictions using KIM (Fig. 2). These developments are set to be integrated into operational systems at KMA following the completion of the second KIAPS project (2020–2026). Figure 1: Various grid systems in KIM: Global uniform (left), stretched (center), and limited-area grid (right) Figure 2: 6-h accumulated precipitation (+12-h prediction) simulated by various KIM configurations at 0000 UTC on July 17 2025. Results include simulations from the uniform global 8 km grid (NE576), the global variableresolution model (NE576S2.5), and the 1 km LAM (NE767S5.0) alongside gauge observation (leftmost). 71 Machine learning and AI in weather prediction and climate modeling 10 72 Application of artificial intelligence models in improving the intensity forecast of Tropical Cyclones over North Indian Ocean Basin 1Dhananjay Trivedi, 1Omveer Sharma, 1Sandeep Pattnaik*, 2Hemant Kumar, 2Samarth Bansal, 2Niladri Bihari Puhan 1 School of Earth, Ocean, and Climate Science, Indian Institute of Technology, Odisha, India 2School of Electrical Sciences, Indian Institute of Technology, Odisha, India Email: [email protected] 1.Tropical Cyclone Intensity 1.1 Estimation Tropical cyclones (TC) pose serious risks and inflict major property and human damage; comprehending a TC's various stages is crucial to assessing its effects. It has been more challenging but essential in recent years to create effective, high-performing techniques for cyclonic storm prediction, mitigation, and coexistence. In general, it is extremely difficult for operational agencies to estimate the intensity of TC using physics-based dynamical models, and the situation is highly challenging if a TC goes through a Rapid Intensification (RI) phase. We design a novel attention mechanism for cyclone intensity estimation based on the perceptual Mach band effect that facilitates boosting and suppression of edges in the input image. This is the first work to establish a useful connection between the perceptual Mach band effect and the adaptive feature weighing mechanism intrinsic to the attention model deployed within a deep convolutional neural network (CNN). The Mach band attention model (MBAM) aims to amplify or suppress the prominence of feature locations within the convolutional feature space by leveraging attention weights. We trained and tested our model using INSAT3DR satellite images and India Meteorological Department (IMD) best intensity archive datasets.The proposed model shows the Mean Absolute Error (MAE) of 0.354 (kts) and Root Mean Square Error of 0.415 kts in wind speed estimation. In the cyclone severity classification, the model shows accuracy of 0.9944, recall of 0.9943, and F1-score of 0.9938 shows the excellence of model performance in cyclone’s intensity estimation over NIO region (Bansal et al., 2024). Further, the model has been tested for RI of TCs in NIO basins. The testing of the RI TCs (i.e., Bulbul, Maha, and Nilofar) has been carried out using various attention mechanisms, such as Mach Band and CBAM, with a basic model as ResNet18.The results suggested that the MBAM3 shows the least (2.87 kts) and CBAM4 shows the highest (3.39 kts) RMSE for all TCs. Further, the MBAM3 model is used for RI estimation of the TC. The MBAM model captures the RI phase of the TCs accurately with an RMSE of 1.65 kts for Bulbul, 2.63 kts for Maha, and 1.68 kts for Nilofar (Sharma et al., 2025a). 79 Figure 1. Actual (obs) and Estimated (model) intensity plot for the TC Nilofar (Sharma et al. 2025). 1.2 Prediction We have designed a fused deep learning (DL) network comprised of a Vision Transformer (ViT) and Convolutional Block Attention Model (CBAM) with ResNet, known as HASTVi, for cyclone intensity forecast with a lead time of up to 24 hours. The model has been trained and tested using infrared satellite images obtained from the Indian Space Research Organization (ISRO) and the India Meteorological Department (IMD) best estimates of TC intensity. A total of 38 TCs were used to train and test the model, out of which 35 TCs (7274 images) were used for training and 3 TCs (694 images) were used for testing purposes. The three most devastating TCs, viz. Amphan (2020), Fani (2019), and Tauktae (2021) were used for testing purposes due to their immense socio economic impact. The proposed vision HASTVi model shows the Mean absolute error (MAE) of 2.90 kts, 3.14 kts, and 5.38 kts, for 3hr forecast, which is better than the state-of-the-art Convolutional Long Short Term Memory (ConvLSTM) model, having an MAE of 15.35 kts, 18.99 kts, and 29.90 kts for Taukate, Fani, and Amphan, respectively. In addition, the HASTVi model has been tested for the rapid intensification phase of the TC over the NIO basin (Sharma et al., 2025b). These findings have direct implications for improving the TC early warning systems over the NIO basins. References Bansal, S., Puhan, N. B., & Pattnaik, S. (2024). Novel Perceptual Mach Band-Based Deep Attention Network for Cyclone Intensity Estimation. IEEE Transactions on Instrumentation and Measurement, 73, 1-11. Sharma, O., Kumar, H., Trivedi, D., Goswami, N., Pattnaik, S., & Puhan, N. B. (2025a). Estimation of the rapid intensification of tropical cyclones over the North Indian Ocean using attention-based deep learning models. Natural Hazards, 1-16. Sharma, O., Trivedi, D., Kumar, H., Pattnaik, S., & Puhan, N. B. (2025b). Enhancing intensification forecast of tropical cyclones in the North Indian Ocean Basin using fused Vision transformer and Convolutional block attention with ResNet model. IEEE Transactions on Geoscience and Remote Sensing. 80 WGNE Working Group on Numerical Experimentation ESMO Earth System Modelling and Observations WCRP World Climate Research Programme WORLD METEOROLOGICAL ORGANIZATION