Assimilation Frequency Research Articles

How frequent and which variables of automatic weather station data should be assimilated into WRF-3DVar model? A case study of a squall line event in Beijing

Accurate precipitation forecasting is one of the most challenging problems in mesoscale Numerical Weather Prediction (NWP) models. The utilization of 3-Dimensional Variational (3DVar) Data Assimilation technique based on automatic weather station (AWS) data can significantly enhance the accuracy of precipitation simulations and forecasts using the Weather Research and Forecasting (WRF) model. However, the impacts of different assimilation frequencies and meteorological variables on the accuracy of precipitation forecasts remain unclear. This study comprehensively evaluated the impact of different assimilation frequencies and meteorological variables on precipitation forecasts, as illustrated by a case study of a squall line event on August 2, 2017, in Beijing. Eight experiments were conducted by assimilating different combinations of meteorological variables (air pressure, temperature, relative humidity, wind speed, and wind direction) at various time intervals (1 h, 3 h, and 6 h). The results indicated that the WRF model roughly simulated the evolution of this event but overestimated the precipitation amount, accompanied by a large deviation in precipitation areas. Assimilating detailed AWS data significantly improves the model performance in precipitation forecasts. The experiment assimilating all variables at a 3 h interval yielded the most accurate forecasts, with the maximum threat score (TS) increasing from 0.02 to 0.56. Higher assimilation frequencies do not guarantee a better performance. In practice, a 3 h assimilation frequency emerges as an optimal choice for AWS data assimilation. The assimilation of various variables enhanced precipitation forecasts to varying degrees, with the optimum results achieved when all variables were assimilated. Wind speed and direction were the most significant factors in dynamically enhancing the simulation of precipitation areas. Relative humidity and temperature influenced the precipitation intensity by affecting the evolution of convective precipitation thermodynamically. The findings of this study can contribute to the development of AWS data assimilation strategies for the WRF-3DVar model, thereby enhancing the precipitation forecast accuracy.

Assimilation of Water Vapor Retrievals from ZDR Columns Using the 3DVar Method for Improving the Short-Term Prediction of Convective Storms

Abstract The differential reflectivity (ZDR) column is a notable polarimetric signature related to updrafts in deep moist convection. In this study, pseudo–water vapor (qυ) observations are retrieved from observed ZDR columns under the assumption that humidity is saturated within the convection where ZDR columns are detected, and are then assimilated within the 3DVar framework. The impacts of assimilating pseudo-qυ observations from ZDR columns on short-term severe weather prediction are first evaluated for a squall-line case. Radar data analysis indicates that the ZDR columns are mainly located on the inflow side of the high-reflectivity region. Assimilation of the pseudo-qυ observations leads to an enhancement of qυ within the convection, while concurrently reducing humidity in no-rain areas. Sensitivity experiments indicate that a tuned smaller observation error and a shorter horizontal decorrelation scale are optimal for a better assimilation of pseudo-qυ from ZDR columns, resulting in more stable rain rates during short-term forecasts. Additionally, a 15-min cycling assimilation frequency yields the best performance, providing the most accurate reflectivity forecast in terms of both location and intensity. Analysis of thermodynamic fields reveal that assimilating ZDR columns provides more favorable initial conditions for sustaining convection, including sustainable moisture condition, a strong cold pool, and divergent winds near the surface, consequently enhancing reflectivity and precipitation. With the optimal configuration determined from the sensitivity tests, a quantitative evaluation further demonstrates that assimilating the pseudo-qυ observations from ZDR columns using the 3DVar method can improve the 0–3-h reflectivity and accumulated precipitation predictions of convective storms.

Improved data assimilation for algal bloom dynamics simulation in the Three Gorges Reservoir using particle filter

Algal blooms have been increasingly prevalent in recent years, especially in lakes and reservoirs; their accurate prediction is essential for preserving water quality. In this study, the observed chlorophyll a (chl-a) levels were assimilated into the Environmental Fluid Dynamics Code (EFDC) of algal bloom dynamics by using a particle filter (PF), and the state variables of water quality and model parameters were simultaneously updated to achieve enhanced algal bloom predictive performance. The developed data assimilation system for algal blooms was applied to Xiangxi Bay (XXB) in the Three Gorges Reservoir (TGR). The results show that the ensemble mean accuracy and reliability of the confidence intervals of the predicted state variables, including chl-a and indirectly updated phosphate (PO4), ammonium (NH4), and nitrate (NO3) levels, were considerably improved after PF assimilation. Thus, PF assimilation is an effective tool for the dynamic correction of parameters to represent their inherent variations. Increased assimilation frequency can effectively suppress the accumulation of model errors; therefore, the use of high-frequency water quality data for assimilation is recommended to ensure more accurate and reliable algal bloom prediction.

Based on historical weather data to predict summer field-scale maize yield: Assimilation of remote sensing data to WOFOST model by ensemble Kalman filter algorithm

Effective support for agricultural production-management strategies relies on the accurate monitoring of crop growth and grain yield estimation at the field scale, making it an urgent need for the development of precision agriculture. Crop models are capable of simulating the growth and development of crops. In addition, Unmanned Aerial Vehicle (UAV) provides high temporal and spatial resolution for remote sensing, which can quickly and accurately obtain the crop growth status of small and medium-sized areas. This research utilized multispectral remote sensing information collected via UAV, alongside the WOFOST (World Food Studies) model, to evaluate the growth of summer maize within the experimental site. By utilizing vegetation indices derived from the UAV remote sensing data, the LAI (Leaf Area Index) of summer maize was ascertained, which further revealed a robust correlation between LAI and NDVI (Normalized Difference Vegetation Index). The model parameters were calibrated using crop and soil data from the Irrigation Experiment Station of Northwest A&F University. Verification indices revealed good consistency with a normalized root mean square error (nRMSE) of around 25 %, indicating that the model was suitable for simulating summer maize growth in this region. The ensemble Kalman filter algorithm was utilized to assimilate the remote sensing observation data with the WOFOST model, and the results showed that assimilation improved the accuracy of the yield simulation for each treatment. The optimal assimilation frequency should be more than three times, and assimilation during the flowering and grain-filling stages of summer maize could obtain better yield simulation results. By assimilating remote sensing observation data with 33 years of historical meteorological data from the Yangling Meteorological Station using the ensemble Kalman filter algorithm, this study predicted the yield and obtained a relative error of less than 5 %. The predicted yield value obtained on September 6 was more accurate than the predicted yield obtained at other times.

Data assimilation of soil moisture and leaf area index effectively improves the simulation accuracy of water and carbon fluxes in coupled farmland hydrological model

Real time status of farmland hydrology and crop growth is essential for agricultural management. Data assimilation is a common method to improve the prediction accuracy of the model by fusing observed and simulated data. For the farmland hydrological processes, evapotranspiration (ET) and net ecosystem productivity (NEP) are widely concerned and strongly affected by crop growth and soil moisture. In this study, data assimilation for soil water content (SWC) and leaf area index (LAI) was combined with a coupled farmland hydrological model, and the potential of Kalman filter (KF) and ensemble Kalman filter (EnKF) methods to enhance model accuracy were explored. Furthermore, the impact of observation density of assimilated data and different assimilation strategies (single-factor or dual-factor assimilation) were analyzed. The findings revealed that both KF and EnKF methods effectively improved the simulation ability of SWC and LAI. When assimilation was performed daily, KF could obtain results comparable to EnKF with assimilation efficiency coefficient (Eff) exceeded 70%. However, with a reduced assimilation frequency for LAI to ten-day interval, EnKF exhibited superior applicability, demonstrating a 13% increase in Eff. The assimilation of soil moisture could positively affect the simulation results of ET with Eff close to 10%, and the assimilation of LAI could improve the simulation accuracy of NEP with Eff close to 15%. Overall, dual-factor assimilation proved to have a more substantial impact than single-factor, even reducing the frequency to ten-day interval. The sensitivity analysis showed that the coupling model could resist the influence of the preset observation error in the filter, with data assimilation effectively mitigating the influence of parameter errors in coupling model. These analyses supply an effective basis to deepen the understanding of improve real time simulation accuracy of farmland hydrological model with data assimilation.

Assimilation of radar‐derived refractivity and radar data in the context of ensemble Kalman filter: Cases study of the Southwest Monsoon Experiment

AbstractRadar‐derived refractivity provides moisture information near the surface, which plays a crucial role in convective initiation and of heavy rain. In this study, the high‐resolution Weather Research and Forecasting local ensemble transform Kalman filter data assimilation system was employed for two cases in the Southwest Monsoon Experiment to conduct two experimental sets. The first set was applied in both cases to investigate the effects of assimilating radar‐retrieved refractivity along with reflectivity and radial wind. The second set was conducted in the second case to examine the benefit of increasing the frequency of refractivity assimilation and investigate the optimal strategy to assimilate refractivity. Results of the two cases revealed that assimilating reflectivity and radial velocity modified near‐surface humidity on the basis of the flow‐dependent error correlation estimated by the ensemble, but the spatial distribution may not be fully accurate, causing underestimation of rainfall. With further refractivity assimilation, stronger convergence and more accurate low‐level moisture, temperature, and wind field corrections were obtained, leading to superior forecasts for both light and heavy rainfall during six hours. The results of the second set indicated that increasing the assimilation interval of refractivity enabled capturing the dramatic moisture variation and enhancing wind convergence, resulting in short‐term forecast improvement. The strategy that solely assimilated refractivity before the appearance of the convection system optimized the correction of environmental moisture, then accurately represented the humidity and strengthened the wind convergence when precipitation occurred. Consequently, the achieved improvement of the short‐term forecast was most noteworthy, particularly for heavy rain.

Impact of Radar Reflectivity Data Assimilation Frequency on Convection-Allowing Forecasts of Diverse Cases over the Continental United States

Abstract The sensitivity of convection-allowing forecasts over the continental United States to radar reflectivity data assimilation (DA) frequency is explored within the Gridpoint Statistical Interpolation (GSI)-based ensemble–variational (EnVar) system. Experiments with reflectivity DA intervals of 60, 20, and 5 min (RAIN60, RAIN20, and RAIN5, respectively) are conducted using 10 diverse cases. Quantitative verification indicates that the degree of sensitivity depends on storm features during the radar DA period. Five developing storms show high sensitivity, whereas five mature or decaying storms do not. The 20-min interval is the most reliable given its best overall performance compared to the 5- and 60-min intervals. Diagnostics suggest that the differences in analyzed cold pools (ACPs) among RAIN60, RAIN20, and RAIN5 vary by storm features during the radar DA period. Such ACP differences result in different forecast skills. In the case where RAIN20 outperforms RAIN60 and the case where RAIN5 outperforms RAIN20, assimilation of reflectivity with a higher frequency commonly produces enhanced and widespread ACPs, promoting broader storms that match better with reality than a lower frequency. In the case where RAIN5 performs worse than RAIN20, the model imbalance of RAIN5 overwhelms information gain associated with frequent assimilation, producing overestimated and spuriously fast-moving ACPs. In the cases where little sensitivity to the reflectivity DA frequency is found, similar ACPs are produced.

ВЛИЯНИЕ СПОСОБОВ УСВОЕНИЯ СПУТНИКОВЫХ ДАННЫХ O ТЕМПЕРАТУРЕ ПОВЕРХНОСТИ МОРЯ НА ВОСПРОИЗВЕДЕНИЕ ГИДРОФИЗИЧЕСКИХ ПОЛЕЙ ЧЕРНОГО, АЗОВСКОГО И МРАМОРНОГО МОРЕЙ В МОДЕЛИ INMOM

The results are analyzed of the simulation of hydrophysical fields of the Black, Azov, and Marmara seas with the Institute of Numerical Mathematics Ocean Model (INMOM) implemented with a spatial resolution of 4 km, with various assimilation technique for sea surface temperature (SST) data from the SEVIRI sensor installed on MSG satellites. The relaxation (so called nudging) and the ensemble optimal interpolation (EnOI) with various assimilation time frequencies (3-, 6-, 12-, and 24-hour assimilation windows) were used as assimilation methods. It was shown that the assimilation of SST data via the EnOI made it possible to reproduce hydrophysical fields more accurately than via nudging or without assimilation at all. Even with the assimilation of SST data irregularly distributed over space and time, a decrease in the calculation error was observed over the entire sea area, and the structures of the zones of temperature rise or drop were more correctly simulated. The best results were achieved with via the EnOI assimilation with an increasing assimilation frequency over time. When SST data were assimilated using the EnOI, the mean error decreased from 0.16°C (24-hour assimilation window) to 0.08°C (3-hour assimilation window); accordingly, the absolute mean error decreased from 1.03 to 0.33°C, and the standard deviation decreased from 1.33 to 0.42°C. In addition, assimilation using the EnOI 3-hour window improved the reproduction of SST during the period of convection cooling. The assimilation of SST data also led to changes in the structure of the surface sea circulation. In some areas, the direction of currents varied within 5°-10°, and the velocity modulus changed by 3-5%. The assimilation of SST data only slightly reduced errors in the structure of the model vertical temperature profile, which can reach 2°C at depths of 30-40 m. At depths greater than 100 m, deviations did not exceed 0.05°C.

Improving the prediction of heavy rainfall with rapid-update dual-resolution hybrid En3DVar assimilation of all-sky AHI infrared water vapor radiances

The Advanced Himawari Imager (AHI) on board Himawari-8 provides continuous high-resolution observations of severe weather phenomena in space and time, providing a good data source for the rapidly updated cycling data assimilation in numerical weather prediction (NWP). In this study, the all-sky AHI water vapor infrared radiances were rapidly assimilated using a dual-resolution hybrid EnVar method to introduce the flow-dependent ensemble information with a reduced ensemble cost in a case study. Considering the greatly increased computational costs for all-sky radiance data assimilation, the sensitivity of the all-sky radiance data assimilation to the resolution of ensemble error covariance as well as the cycling frequency using dual-resolution EnVar was investigated. The impact of all-sky radiance data assimilation using the dual-resolution hybrid EnVar was then compared to the clear-sky radiance data assimilation.From the case study, it was found that the dual-resolution hybrid EnVar and the full-resolution hybrid EnVar generally performed comparable for the all-sky assimilation of the AHI water vapor radiances in this study. Both show great advantages over 3DVar in analyses and forecasts with different leading time for model state variables. However, the former cost much fewer computational resources with only a tiny loss of the accuracy. For the assimilation frequency of all-sky radiances using the dual-resolution hybrid EnVar, it was found that the half-hourly and hourly test both obtain the better results than the other two (namely the 10-minutely test and 3-hourly test). The former two have the comparable performances in rainfall forecast although the half-hourly test is better. Finally, in the dual-resolution hybrid EnVar framework, the all-sky AHI radiance assimilation impact was studied versus the clear-sky radiances as well as the conventional observations in a case study. The results present a generally positive improvement of the rainfall forecast with the larger rainfall thresholds, caused by the assimilation of the cloud-influenced AHI water vapor radiances.

Assimilation of blended in situ-satellite snow water equivalent into the National Water Model for improving hydrologic simulation in two US river basins

This study investigates the potential of assimilating a 1/8° blended in situ-satellite snow water equivalent (SWE) product for improving snow and streamflow predictions of the National Water Model (NWM). The blended product is assimilated into the NWM via a three-dimensional variational (3DVAR) scheme and a direct insertion (DI) scheme, with a daily (1d) and a every 5 days (5d) assimilation frequencies. The experiments are for the Upper Colorado River Basin (UCRB) and Susquehanna River Basin (SRB), which feature seasonal and ephemeral snow covers, respectively. Results indicate that 3DVAR with a 5d assimilation frequency generally outperforms the other scenarios. The assimilation of the blended SWE product mitigates the underestimation of SWE evident in the open-loop simulations for both basins and its impacts are more pronounced for UCRB than for SRB since snowfall is the main source of precipitation in the former. Assimilation leads to improved streamflow over a majority of SRB subbasins, but over a minority of UCRB subbasins. The degradations in streamflow for UCRB subbasins are mainly caused by the overestimated SWE. In addition, the open-loop simulation often produces an earlier streamflow peak in UCRB, and this error is mitigated to a limited extent by assimilation. These findings in aggregate suggest that the efficacy of snow assimilation is strongly dependent upon the types of snowpack and differential assimilation methods and frequencies.

Understanding the key factors that influence soil moisture estimation using the unscented weighted ensemble Kalman filter

Accurate quantification of soil moisture contributes significantly to an understanding of land surface processes. In-situ observable soil moisture data are often sparsely distributed, and model performance is influenced by many factors. In this study, 14 numerical experimental schemes about the effects of uncertainties in multiple factors (soil property, time step, assimilation interval, precipitation, soil layer thickness and initial value) on soil moisture estimation were evaluated based on the unscented weighted ensemble Kalman filter (UWEnKF) and a one-dimensional vertical water flow model at the ELBARA field site in the Maqu monitoring network in the upper reaches of the Yellow River, China. The experiments showed that soil properties had little effect on model parameters (e.g., saturated soil moisture content, saturated soil hydraulic conductivity, saturated soil matric potential) in either the horizontal or vertical direction using the model numerical solving scheme adopted, and thus had little effect on soil moisture estimation. Using only the observed Ksatmay lead to better soil moisture predictions. Reducing the simulation time step has limited impact on soil moisture estimation. The effects of precipitation on soil moisture estimations varied due to overestimation or underestimation of soil moisture content in different soil layers, and differences in soil layer thicknesses led to uncertainty in soil moisture estimation. The model accurately predicted the change trend of soil moisture if the initial values were reasonable. UWEnKF performed well in terms of improving soil moisture estimations despite the uncertainty of many factors in data assimilation system, and performed better with high assimilation frequency (i.e., small assimilation interval). Thus, UWEnKF is an effective and practical technique for soil moisture assimilation whatever the uncertainty of multiple factors is.

Assimilation of X‐Band Phased‐Array Radar Data With EnKF for the Analysis and Warning Forecast of a Tornadic Storm

AbstractThe impact of assimilating China's operational X‐band Phased‐Array radar's (X‐PAR) data on the analysis and warning forecast of the vortex structure and intensity of the June 8, 2018 Foshan, Guangdong province, tornadic storm was investigated for the first time using an Ensemble Kalman Filter (EnKF) data assimilation system. Both radar radial velocity (Vr) and reflectivity (Z) from two S‐band operational radars and one X‐PAR were assimilated. Deterministic forecasts were launched every 6 min from 05:42 UTC (20 min before the tornado touched down) to 06:00 UTC from the EnKF mean analysis field. Five experiments were conducted to examine the added capability of Z assimilation of the EnKF system, and to investigate the impact of assimilating X‐PAR data on the analysis and prediction of the tornadic storm. Compared to the experiment without Z assimilation, the assimilation of Z reduced the analysis error and greatly reduced the forecast error of Z. The assimilation of X‐PAR data greatly improved the vortex structure of the tornadic storm at low levels, and improved the intensity of the rear inflow of the tornadic storm, especially with a higher assimilation frequency. Compared to the experiments without X‐PAR data assimilation, assimilating X‐PAR data improved the predictability of tornadic storm.

Effect of the Assimilation Frequency of Radar Reflectivity on Rain Storm Prediction by Using WRF-3DVAR

An attempt was made to evaluate the impact of assimilating Doppler Weather Radar (DWR) reflectivity together with Global Telecommunication System (GTS) data in the three-dimensional variational data assimilation (3DVAR) system of the Weather Research Forecast (WRF) model on rain storm prediction in Daqinghe basin of northern China. The aim of this study was to explore the potential effects of data assimilation frequency and to evaluate the outputs from different domain resolutions in improving the meso-scale NWP rainfall products. In this study, four numerical experiments (no assimilation, 1 and 6 h assimilation time interval with DWR and GTS at 1 km horizontal resolution, 6 h assimilation time interval with radar reflectivity, and GTS data at 3 km horizontal resolution) are carried out to evaluate the impact of data assimilation on prediction of convective rain storms. The results show that the assimilation of radar reflectivity and GTS data collectively enhanced the performance of the WRF-3DVAR system over the Beijing-Tianjin-Hebei region of northern China. It is indicated by the experimental results that the rapid update assimilation has a positive impact on the prediction of the location, tendency, and development of rain storms associated with the study area. In order to explore the influence of data assimilation in the outer domain on the output of the inner domain, the rainfall outputs of 3 and 1 km resolution are compared. The results show that the data assimilation in the outer domain has a positive effect on the output of the inner domain. Since the 3DVAR system is able to analyze certain small-scale and convective-scale features through the incorporation of radar observations, hourly assimilation time interval does not always significantly improve precipitation forecasts because of the inaccurate radar reflectivity observations. Therefore, before data assimilation, the validity of assimilation data should be judged as far as possible in advance, which can not only improve the prediction accuracy, but also improve the assimilation efficiency.

Improved Modeling of Spatiotemporal Variations of Fine Particulate Matter Using a Three‐Dimensional Variational Data Fusion Method

AbstractThe spatiotemporal concentration of multiple pollutants is crucial information for pollution control strategies to safeguard public health. Despite considerable efforts, however, significant uncertainty remains. In this study, a three‐dimensional variational model is coupled with a data assimilation (DA) system to analyze the spatiotemporal variation of PM2.5 for the whole of China. Monthly simulations of six sensitivity scenarios in different seasons, including different assimilation cycles, are carried out to assess the impact of the assimilation frequency on the PM2.5 simulations and the model simulation accuracy afforded by DA. The results show that the coupled system provides more reliable initial fields to substantially improve the model performance for PM2.5, PM10, and O3. Higher assimilation frequency improves the simulation in all geographic areas. Two statistical indicators—the root mean square error and the correlation coefficient of PM2.5 mass concentrations in the analysis field—are improved by 12.19 µg/m3 (33%) and 0.21 (48%), respectively. Although the 24‐h assimilation cycle considerably improves the model, assimilation at a 6‐h cycle raises the performance for PM2.5 to the performance goal level. The analysis shows that assimilating at a 24‐h cycle diminishes over time, whereas the positive impact of the 6‐h cycle persists. One pivotal finding is that the assimilation of PM2.5 in the outermost domain results in a substantial improvement in PM2.5 prediction for the innermost domain, which is a potential alternative method to the existing domain‐wide data fusion algorithm. The effect of assimilation varies among topographies, a finding that provides essential support for further model development.

Impacts of the radar data assimilation frequency and large-scale constraint on the short-term precipitation forecast of a severe convection case

Radar data assimilation (DA) can improve short-term precipitation forecasts. However, assimilating radar observations into a numerical prediction model may produce spurious precipitation and large errors in the position and magnitude of precipitation, and high-frequency cyclic assimilation may exacerbate this problem. In this paper, a severe convection case that occurred on 27 June 2018 over eastern China was used to investigate the impacts of the radar DA frequency and large-scale constraint on precipitation forecasts. The Weather Research and Forecasting (WRF) three-dimensional variational (3D-Var) DA system was used to assimilate radial velocity and reflectivity data. The sensitivity of the DA frequency using 15-min and 1-h assimilation intervals was examined first, revealing that the 15-min DA interval produced a greater overprediction bias of precipitation. A diagnosis of the analysis fields showed that higher-frequency DA cycling produced excessive wind speed and water vapor convergence at low levels and exaggerated upward movement. Then, the large-scale constraint was imposed on the regional model by the spectral nudging (SN) technique, which assimilated Global Forecast System (GFS) forecast field data during the forecast periods. The results demonstrated that the application of SN could significantly reduce the positional deviation of the precipitation forecast and the magnitude of overpredicted precipitation. SN could effectively adjust large-scale circulation fields, thereby improving the conditions of water vapor convergence. Moreover, SN also improved the forecasts of surface variables such as wind, temperature, and humidity.

Assimilation of Satellite-Retrieved Sea Ice Concentration and Prospects for September Predictions of Arctic Sea Ice

AbstractThe current GFDL seasonal prediction system achieved retrospective sea ice extent (SIE) skill without direct sea ice data assimilation. Here we develop sea ice data assimilation, shown to be a key source of skill for seasonal sea ice predictions, in GFDL’s next-generation prediction system, the Seamless System for Prediction and Earth System Research (SPEAR). Satellite sea ice concentration (SIC) observations are assimilated into the GFDL Sea Ice Simulator version 2 (SIS2) using the ensemble adjustment Kalman filter (EAKF). Sea ice physics is perturbed to form an ensemble of ice–ocean members with atmospheric forcing from the JRA-55 reanalysis. Assimilation is performed every 5 days from 1982 to 2017 and the evaluation is conducted at pan-Arctic and regional scales over the same period. To mitigate an assimilation overshoot problem and improve the analysis, sea surface temperatures (SSTs) are restored to the daily Optimum Interpolation Sea Surface Temperature version 2 (OISSTv2). The combination of SIC assimilation and SST restoring reduces analysis errors to the observational error level (~10%) from up to 3 times larger than this (~30%) in the free-running model. Sensitivity experiments show that the choice of assimilation localization half-width (190 km) is near optimal and that SIC analysis errors can be further reduced slightly either by reducing the observational error or by increasing the assimilation frequency from every 5 days to daily. A lagged-correlation analysis suggests substantial prediction skill improvements from SIC initialization at lead times of less than 2 months.

Evaluating Forecast Performance and Sensitivity to the GSI EnKF Data Assimilation Configuration for the 28–29 May 2017 Mesoscale Convective System Case

AbstractIn an effort to improve radar data assimilation configurations for potential operational implementation, GSI EnKF data assimilation experiments based on the operational system employed by the Center for Analysis and Prediction of Storms (CAPS) real-time Spring Forecast Experiments are performed. These experiments are followed by 6-h forecasts for an MCS on 28–29 May 2017. Configurations examined include data thinning, covariance localization radii and inflation, observation error settings, and data assimilation frequency for radar observations. The results show experiments that assimilate radar observations more frequently (i.e., 5–10 min) are initially better at suppressing spurious convection. However, assimilating observations every 5 min causes spurious convection to become more widespread with time, and modestly degrades forecast skill through the remainder of the forecast window. Ensembles that assimilate more observations with less thinning of data or use a larger horizontal covariance localization radius for radar data predict fewer spurious storms and better predict the location of observed storms. Optimized data thinning and horizontal covariance localization radii have positive impacts on forecast skill during the first forecast hour that are quickly lost due to the growth of forecast error. Forecast skill is less sensitive to the ensemble spread inflation factors and observation errors tested during this study. These results provide guidance toward optimizing the configuration of the GSI EnKF system. Among the DA configurations tested, the one employed by the CAPS Spring Forecast Experiment produces the most skilled forecasts while remaining computationally efficient for real-time use.

Impacts of Assimilation Frequency on Ensemble Kalman Filter Data Assimilation and Imbalances

AbstractThe ensemble Kalman filter (EnKF) has been widely used in atmosphere, ocean, and land applications. The observing network has been significantly developed, and thus, observations with highly dense temporal resolutions have become available. To better extract information from dense temporal observations, one straightforward strategy is to increase the assimilation frequency. However, more frequent assimilation may exacerbate the model imbalance and result in degraded forecasts. To combat the imbalance caused by ensemble‐based data assimilation due to sampling error and covariance localization, three‐ and four‐dimensional incremental analysis update (IAU) were proposed, which gradually introduce the analysis increments into model rather than intermittently updating the state. The trade‐off between the assimilation frequency and imbalance is systematically explored here by using an idealized two‐layer model and the NOAA GFS. Results from the idealized two‐layer model show that increasing assimilation frequency can reduce errors for state variables that are not sensitive to imbalances. For state variable that carries the signal of the external gravity mode and is sensitive to imbalances, increasing assimilation frequency without (with) IAU reduces (increases) errors. Without IAU, more frequent updates result in smaller increments and less insertion noise, while the initialization of IAU cannot effectively mitigate the imbalances with increased assimilation frequency. Results with a low‐resolution version of the NOAA GFS demonstrate that increasing assimilation frequency from 6 to 2 h improves the errors and biases of forecasts verified with conventional and radiance observations, although gravity wave noise in the forecast is increased.

Using an ensemble Kalman filter method to calibrate parameters of a prediction model for chemical transport from soil to surface runoff.

Water pollution from surface runoff is an important non-point pollution source, which has been a great threat to our environment. The model proposed by Gao et al. (2004) is of great significance to solve the non-point source pollution problem, which is a numerical advection-diffusion equation (ADE) model for chemical transport from soil to surface runoff. The ensemble Kalman filter (EnKF), the data assimilation (DA) method, is easy to be implemented and widely used in hydrology field. In this study, we use the EnKF method to update model state variables such as chemical concentrations in surface runoff and calibrate model parameters such as water transfer rate in Gao et al. (2004) under different study cases, while other model parameters are assumed to be known. The observations are generated from the simulation results based on synthetic real parameters. The objective of this study was to extend the application of the EnKF to the ADE-based prediction model of chemical transport from soil to surface runoff. The results of the predicted chemical concentration in the surface runoff with EnKF are greatly improved than those without EnKF in comparison with the observations, and the updated parameters are close to the real parameters. We explored feasibility of the EnKF method from six factors, including the initial parameter estimate, the ensemble size, the influence of multi-parameters, the assimilation time interval, the infiltration boundary conditions, and the relationship between the standard deviations of the observation error and initial parameter. Different study strategies are proposed for different factors. For assimilation time interval, the key observation can reduce the assimilation frequency. With the situation of much larger observation error covariance than the prediction covariance, we analyzed influences of the standard deviation of the observation error and initial parameter on the feasibility of the EnKF method. According to the study results, it is concluded that the EnKF is efficient to update the parameter for the ADE-based prediction model of chemical transport from soil to surface runoff.

On evaluating the effect of assimilating glider-observed T/S profiles with different horizontal resolutions and assimilation frequencies

Underwater gliders can provide real-time and spatially flexible temperature/salinity (T/S) observations for improving the marine forecast by data assimilation. By conducting Observing System Simulation Experiments (OSSEs), this study aims to investigate the effect of assimilating glider-observed T/S profiles regarding the horizontal resolution of glider deployment and assimilation frequency, as well as the combination of assimilating satellite-derived sea level anomaly (SLA), on the forecast skill for an extreme warm eddy in the Northwestern South China Sea (SCS) in 2010. The results of OSSEs show that assimilating either glider-observed T/S profiles or satellite-derived SLA is able to improve the forecast skill, and assimilating both of them gains the largest improvement. Under the premise of a full coverage of the eddy, it is found that the higher horizontal resolution of glider deployment is, the better forecast skill will be obtained. Meanwhile, the assimilation of the glider-observed T/S profiles with a 12-h interval achieves the best forecast skill among the intervals of 4 h, 8 h, 12 h, and 24 h. These results provide valuable reference for the deployment of underwater gliders as well as the assimilation strategy of glider observations for improving the real-time marine forecast in the Northwestern SCS in the future.