Multivariate Data Assimilation Research Articles

Joint assimilation of satellite-based surface soil moisture and vegetation conditions into the Noah-MP land surface model

This study explores the potential of integrating satellite retrievals of surface soil moisture (SSM) and vegetation conditions into the Noah-MP land surface model. In total, five data assimilation (DA) experiments were carried out. One of the experiments only assimilates SSM retrievals from the Soil Moisture Active Passive mission, two experiments only assimilate retrievals of vegetation conditions: either optical retrievals of leaf area index (LAI) from the Copernicus Global Land Service, or X-band microwave-based retrievals of vegetation optical depth (VOD) from the Advanced Microwave Scanning Radiometer 2. Additionally, two joint DA experiments are performed, each incorporating SSM and one of the vegetation products. The DA experiments are compared with a model-only run, and all experiments are evaluated using independent ground reference data of soil moisture, evapotranspiration, net ecosystem exchange and gross primary production (GPP). Assimilating only SSM improves estimates of the soil moisture profile (median SSM anomaly correlation improves with 0.02 compared to a model-only run), whereas assimilating LAI predominantly improves GPP estimates (reduction in median RMSD of 0.024 gC m−2 day−1 compared to a model-only run). The joint assimilation of SSM and vegetation conditions captures both of these improvements in a single, physically consistent analysis product. The DA increments show that this combined setup allows one satellite product to compensate for potential degradations introduced into the system by the other product. Furthermore, the joint SSM and VOD DA experiment has the smallest ensemble spread in its estimates (21% reduction in SSM spread compared to a model-only run). Overall, our results underline the potential of multi-sensor and multivariate DA, in which information from different sources is combined to improve the estimates of several land surface states and fluxes simultaneously.

Improving modelled streamflow using time-varying multivariate assimilation of remotely sensed soil moisture and in-situ streamflow observations

Hydrological models are widely used to estimate and forecast streamflow for various applications. Given the inherent uncertainties in these models, there is a pressing need to enhance the current state of the modelling strategies. Historically, hydrological models showed significant improvement through soil moisture (SM) assimilation. Particularly, when SM observations are combined with streamflow observations from the interior catchment locations during multivariate assimilation, models showed better performance. Recent studies also emphasized the importance of updating the parameters due to the transient nature of the catchment during the assimilation period. Additionally, it is crucial to determine whether it is necessary to assimilate all available observations. To address these issues, this study introduces five data assimilation (DA) scenarios ingesting advanced scatterometer (ASCAT) SM and in-situ streamflow observations into a conceptual hydrological model, aiming to enhance its performance. The findings reveal that while univariate assimilation improves both SM and streamflow estimates, but a substantial enhancement is observed in multivariate data assimilation (MVDA). Time-varying MVDA (TV-MVDA) substantially improves the model's performance compared to the open-loop scenario case. However, this improvement is only marginal when compared to the TV-MVDA scenario. Finally, the sensitivity-based TV-MVDA scenario exhibits comparable performance in streamflow estimation, while assimilating just 30 % of observations. These results suggest that Sensitivity based TV-MVDA can improve the model efficiently with minimal observation requirements.

Multivariate Data Assimilation at a Partially Mixed Estuary

Abstract In 2013 and 2014, multiple field excursions of varying scope were concentrated on the Columbia River, a highly energetic, partially mixed estuary. These experiments included surface drifter and synthetic aperture radar (SAR) measurements during the ONR RIVET-II experiment, and a novel animal tracking effort that samples oceanographic data by employing cormorants tagged with biologging devices. In the present work, several different data types from these experiments were combined in order to test an iterative, ensemble-based inversion methodology at the mouth of the Columbia River (MCR). Results show that, despite inherent limitations of observation and model accuracy, it is possible to detect dynamically relevant bathymetric features such as large shoals and channels while originating from a linear, featureless prior bathymetry in a partially mixed estuary by inverting surface current and gravity wave observations with a 3D hydrostatic ocean model. Bathymetry estimation skill depends on two factors: location (i.e., differing estimation quality inside versus outside the MCR) and observation type (e.g., surface currents versus significant wave height). Despite not being inverted directly, temperature and salinity outputs in the hydrodynamic model improved agreement with observations after bathymetry inversion.

Flash Drought Onset and Development Mechanisms Captured With Soil Moisture and Vegetation Data Assimilation

AbstractFlash droughts evolve and intensify rapidly under the influence of anomalous atmospheric conditions. In this study, we investigate the role of assimilating remotely sensed soil moisture (SM) and vegetation properties in capturing the evolution and impacts of two flash droughts in the Northern Great Plains. We find that during 2016 drought triggered by anomalously high temperatures and excessive evaporative demands, multivariate data assimilation (DA) of MODIS‐derived leaf area index (LAI) and Soil Moisture Active Passive SM within Noah‐Multiparameterization model helps capture elevated transpiration at onset. Assimilation of LAI particularly helped model the resulting rapid decline in SM during onset with as high as 10.0% steeper rate of decline compared to the simulation without any assimilation. Modeled‐SM anomalies exhibit a 7.5% and 11.7% increase in similarity with Evaporative Stress Index (ESI) data and U.S. Drought Monitor (USDM) maps, respectively. In contrast, during 2017 flash drought driven by record‐low precipitation during summers, SM assimilation resulted in largest rates of decline in rootzone SM, as large as 48.4% compared to results from no assimilation. Multivariate DA of SM and LAI results in 6.7% and 14.3% higher spatial similarity with ESI and USDM, respectively, and is necessary to model rapid intensification caused by anomalous precipitation deficits. This study elucidates the need to incorporate multiple observational constraints from remote sensing to effectively capture rapid onset rates, intensification, and severity of flash drought following different propagation mechanisms. This is fundamental for drought early detection to provide a wider window of response and implement efficient mitigation strategies.

SWAT_DA: Sequential Multivariate Data Assimilation‐Oriented Modification of SWAT

AbstractMultivariate data assimilation (DA), a novel way to couple big data with land surface models, was extensively employed in forecasting‐reanalyzing systems (FRSs), for example, ECMWF and GLDAS. Meanwhile, most (distributed) hydrological models, like soil and water assessment tool (SWAT), have not been equipped with straightforward ways to link to DA algorithms. Therefore, it is one of the main barriers to utilizing such hydrological models in FRSs. This paper deals with multivariate DA into SWAT (DA‐SWAT), which is complicated since the original model does not provide full access to the models' initial conditions (ICs) at the hydrologic response unit (HRU) scale. The preceding DA‐SWAT works commonly used an integrated approach in which the DA and SWAT codes were implemented in the same programming environment. We discuss how this approach complicates and prevents the application of DA‐SWAT in multivariate, multimodel, and multisensor systems. Accordingly, we proposed a new approach for DA‐SWAT by which SWAT can be perfectly linked with any DA algorithm of interest coded in any desired programming environment. Our framework utilizes input/output text files to access ICs and to link DA with SWAT. Moreover, we designed some univariate and multivariate scenarios for assimilating in situ streamflow measurement and MODIS's snow cover fraction (SCF) data, which has not yet been focused on in the SWAT calibration context. Results show that compared to the univariate assimilation of streamflow (SCF), the multivariate assimilation mitigates the equifinality problem and more accurately estimates SCF (streamflow) by improving NS and PBIAS measures with the differences of 0.4 (0.86), 12% (64%), respectively.

A Quantile-Conserving Ensemble Filter Framework. Part I: Updating an Observed Variable

Abstract A general framework for deterministic univariate ensemble filtering is presented. The framework fits a continuous prior probability density function (PDF) to the prior ensemble. A functional representation for the observation likelihood is combined with the prior PDF to get a continuous analysis (posterior) PDF. Cumulative distribution functions for the prior and analysis are also required. The key innovation is that an analysis ensemble is computed so that the quantile of each ensemble member is the same as its prior quantile. Many choices for the prior PDF family and the likelihood function are described. A choice of normal prior with normal likelihood is equivalent to the ensemble adjustment Kalman filter. Some other choices for the prior include gamma, inverse gamma, beta, beta prime, lognormal, and exponential distributions. Both prior distributions and likelihoods can be defined over a set of intervals giving additional flexibility that can be used to implement methods like a Huber likelihood for observations with occasional outliers. Priors and likelihoods can also be defined as sums of distributions allowing choices like bivariate normals or kernel filters. Empirical distributions, for instance piecewise linear approximations to arbitrary PDFs and functions can be used. Another empirical choice leads to the rank histogram filter. Results here are univariate and can be used to compute increments for observed variables or marginal distributions for any variable for a reanalysis. Linear regression of increments can be used to update state variables in a serial filter to build a comprehensive data assimilation system. Part 2 will discuss other methods for extending the framework to multivariate data assimilation. Significance Statement Data assimilation is used to combine information from model forecasts with subsequent observations to obtain better estimates of the current state of the atmosphere or other parts of the Earth system. Ensemble data assimilation uses a number of forecasts to get more information about uncertainty. A new method allows much more flexibility in the assumptions that must be made when doing ensemble data assimilation. As an example, the method can be better for quantities that are bounded like the amount of an atmospheric trace pollutant.

How does precipitation data influence the land surface data assimilation for drought monitoring?

Droughts are among the costliest natural hazards that occur annually worldwide. Their socioeconomic impacts are significant and widespread, affecting the sustainable development of human societies. This study investigates the influence of different forcing precipitation data in driving Land Surface Models (LSMs) and characterizing drought conditions. Here, we utilize our recently developed LSM data assimilation system for probabilistically monitoring drought over the Contiguous United States (CONUS). The Noah-MP LSM model is forced with two widely used precipitation data including IMERG (Integrated Multi-satellitE Retrievals for GPM) and NLDAS (North American Land Data Assimilation System). Soil moisture and evapotranspiration are known to have a strong relationship in the land-atmospheric interaction processes. Unlike other studies that attempted the individual assimilation of these variables, here we propose a multivariate data assimilation framework. Therefore, in both modeling scenarios, the data assimilation approach is used to integrate remotely sensed MODIS (Moderate Resolution Imaging Spectroradiometer) evapotranspiration and SMAP (Soil Moisture Active Passive) soil moisture observations into the Noah-MP LSM. The results of this study indicate that the source of precipitation data has a significant impact on the performance of LSM data assimilation system for drought monitoring. The findings revealed that NLDAS and IMERG precipitation can result in a significant difference in identifying drought severity depending on the region and time of the year. Furthermore, our analysis indicates that regardless of the precipitation forcing data product used in the land surface data assimilation system, our modeling framework can effectively detect the drought impacts on crop yield. Additionally, we calculated the drought probability based on the ensemble of soil moisture percentiles and found that there exist temporal and spatial discrepancies in drought probability maps generated from the NLDAS and IMERG precipitation forcings.

Implications of Multivariate Non-Gaussian Data Assimilation for Multi-scale Weather Prediction

Abstract Weather prediction models currently operate within a probabilistic framework for generating forecasts conditioned on recent measurements of Earth’s atmosphere. This framework can be conceptualized as one that approximates parts of a Bayesian posterior density estimated under assumptions of Gaussian errors. Gaussian error approximations are appropriate for synoptic-scale atmospheric flow, which experiences quasi-linear error evolution over time scales depicted by measurements, but are often hypothesized to be inappropriate for highly nonlinear, sparsely-observed mesoscale processes. The current study adopts an experimental regional modeling system to examine the impact of Gaussian prior error approximations, which are adopted by ensemble Kalman filters (EnKFs) to generate probabilistic predictions. The analysis is aided by results obtained using recently-introduced particle filter (PF) methodology that relies on an implicit non-parametric representation of prior probability densities—but with added computational expense. The investigation focuses on EnKF and PF comparisons over month-long experiments performed using an extensive domain, which features the development and passage of numerous extratropical and tropical cyclones. The experiments reveal spurious small-scale corrections in EnKF members, which come about from inappropriate Gaussian approximations for priors dominated by alignment uncertainty in mesoscale weather systems. Similar behavior is found in PF members, owing to the use of a localization operator, but to a much lesser extent. This result is reproduced and studied using a low-dimensional model, which permits the use of large sample estimates of the Bayesian posterior distribution. Findings from this study motivate the use of data assimilation techniques that provide a more appropriate specification of multivariate non-Gaussian prior densities or a multi-scale treatment of alignment errors during data assimilation.

Assessing Performances of Multivariate Data Assimilation Algorithms with SMOS, SMAP, and GRACE Observations for Improved Soil Moisture and Groundwater Analyses

Multivariate data assimilation (DA) of satellite soil moisture (SM) and terrestrial water storage (TWS) observations has recently been used to improve SM and groundwater storage (GWS) simulations. Previous studies employed the ensemble Kalman approach in multivariate DA schemes, which assumes that model and observation errors have a Gaussian distribution. Despite the success of the Kalman approaches, SM and GWS estimates can be suboptimal when the Gaussian assumption is violated. Other DA approaches, such as particle smoother (PS), ensemble Gaussian particle smoother (EnGPS), and evolutionary smoother (EvS), do not rely on the Gaussian assumption and may be better suited to non-Gaussian error systems. The objective of this paper is to evaluate the performance of these four DA approaches (EnKS, PS, EnGPS, and EvS) in multivariate DA systems by assimilating satellite data from the Soil Moisture and Ocean Salinity (SMOS), Soil Moisture Active Passive (SMAP), and Gravity Recovery And Climate Experiment (GRACE) missions into the Community Atmosphere and Biosphere Land Exchange (CABLE) land surface model. The analyses are carried out in Australia’s Goulburn River catchment, where in situ SM and groundwater data are available to comprehensively validate the DA performance. Results show that all four DA approaches have outstanding performances and improve correlation coefficients of SM and GWS estimates by ~20% and 100%, respectively. The EvS outperforms the others, but its benefit is relatively marginal compared to Gaussian approaches (e.g., EnKS). This is due to the fact that SM and TWS error distributions in this study are close to Gaussian: a suitable condition for, e.g., EnKS, EnGPS. The robust performance of EvS appears to be the optimal approach for jointly assimilating multi-source hydrological observations to improve regional hydrological analyses.

The Joint Assimilation of Remotely Sensed Leaf Area Index and Surface Soil Moisture into a Land Surface Model

This work tests the hypothesis that jointly assimilating satellite observations of leaf area index and surface soil moisture into a land surface model improves the estimation of land vegetation and water variables. An Ensemble Kalman Filter is used to test this hypothesis across the Contiguous United States from April 2015 to December 2018. The performance of the proposed methodology is assessed for several modeled vegetation and water variables (evapotranspiration, net ecosystem exchange, and soil moisture) in terms of random errors and anomaly correlation coefficients against a set of independent validation datasets (i.e., Global Land Evaporation Amsterdam Model, FLUXCOM, and International Soil Moisture Network). The results show that the assimilation of the leaf area index mostly improves the estimation of evapotranspiration and net ecosystem exchange, whereas the assimilation of surface soil moisture alone improves surface soil moisture content, especially in the western US, in terms of both root mean squared error and anomaly correlation coefficient. The joint assimilation of vegetation and soil moisture information combines the results of individual vegetation and soil moisture assimilations and reduces errors (and increases correlations with the reference datasets) in evapotranspiration, net ecosystem exchange, and surface soil moisture simulated by the land surface model. However, because soil moisture satellite observations only provide information on the water content in the top 5 cm of the soil column, the impact of the proposed data assimilation technique on root zone soil moisture is limited. This work moves one step forward in the direction of improving our estimation and understanding of land surface interactions using a multivariate data assimilation approach, which can be particularly useful in regions of the world where ground observations are sparse or missing altogether.

The Regional Ice Ocean Prediction System v2: a pan-Canadian ocean analysis system using an online tidal harmonic analysis

Abstract. Canada has the longest coastline in the world and includes diverse ocean environments, from the frozen waters of the Canadian Arctic Archipelago to the confluence region of Labrador and Gulf Stream waters on the east coast. There is a strong need for a pan-Canadian operational regional ocean prediction capacity covering all Canadian coastal areas in support of marine activities including emergency response, search and rescue, and safe navigation in ice-infested waters. Here we present the first pan-Canadian operational regional ocean analysis system developed as part of the Regional Ice Ocean Prediction System version 2 (RIOPSv2) running in operations at the Canadian Centre for Meteorological and Environmental Prediction (CCMEP). The RIOPSv2 domain extends from 26∘ N in the Atlantic Ocean through the Arctic Ocean to 44∘ N in the Pacific Ocean, with a model grid resolution that varies between 3 and 8 km. RIOPSv2 includes a multivariate data assimilation system based on a reduced-order extended Kalman filter together with a 3D-Var bias correction system for water mass properties. The analysis system assimilates satellite observations of sea level anomaly and sea surface temperature, as well as in situ temperature and salinity measurements. Background model error is specified in terms of seasonally varying model anomalies from a 10-year forced model integration, allowing inhomogeneous anisotropic multivariate error covariances. A novel online tidal harmonic analysis method is introduced that uses a sliding-window approach to reduce numerical costs and allow for the time-varying harmonic constants necessary in seasonally ice-infested waters. Compared to the Global Ice Ocean Prediction System (GIOPS) running at CCMEP, RIOPSv2 also includes a spatial filtering of model fields as part of the observation operator for sea surface temperature (SST). In addition to the tidal harmonic analysis, the observation operator for sea level anomaly (SLA) is also modified to remove the inverse barometer effect due to the application of atmospheric pressure forcing fields. RIOPSv2 is compared to GIOPS and shown to provide similar innovation statistics over a 3-year evaluation period. Specific improvements are found near the Gulf Stream for all model fields due to the higher model grid resolution, with smaller root mean squared (rms) innovations for RIOPSv2 of about 5 cm for SLA and 0.5 ∘C for SST. Verification against along-track satellite observations demonstrates the improved representation of mesoscale features in RIOPSv2 compared to GIOPS, with increased correlations of SLA (0.83 compared to 0.73) and reduced rms differences (12 cm compared to 14 cm). While the RIOPSv2 grid resolution is 3 times higher than GIOPS, the power spectral density of surface kinetic energy provides an indication that the effective resolution of RIOPSv2 is roughly double that of the global system (35 km compared to 66 km). Observations made as part of the Year of Polar Prediction (2017–2019) provide a rare glimpse at errors in Arctic water mass properties and show average salinity biases over the upper 500 m of 0.3–0.4 psu in the eastern Beaufort Sea in RIOPSv2.

Multi-mission satellite remote sensing data for improving land hydrological models via data assimilation

Satellite remote sensing offers valuable tools to study Earth and hydrological processes and improve land surface models. This is essential to improve the quality of model predictions, which are affected by various factors such as erroneous input data, the uncertainty of model forcings, and parameter uncertainties. Abundant datasets from multi-mission satellite remote sensing during recent years have provided an opportunity to improve not only the model estimates but also model parameters through a parameter estimation process. This study utilises multiple datasets from satellite remote sensing including soil moisture from Soil Moisture and Ocean Salinity Mission and Advanced Microwave Scanning Radiometer Earth Observing System, terrestrial water storage from the Gravity Recovery And Climate Experiment, and leaf area index from Advanced Very-High-Resolution Radiometer to estimate model parameters. This is done using the recently proposed assimilation method, unsupervised weak constrained ensemble Kalman filter (UWCEnKF). UWCEnKF applies a dual scheme to separately update the state and parameters using two interactive EnKF filters followed by a water balance constraint enforcement. The performance of multivariate data assimilation is evaluated against various independent data over different time periods over two different basins including the Murray–Darling and Mississippi basins. Results indicate that simultaneous assimilation of multiple satellite products combined with parameter estimation strongly improves model predictions compared with single satellite products and/or state estimation alone. This improvement is achieved not only during the parameter estimation period (sim 32% groundwater RMSE reduction and soil moisture correlation increase from sim 0.66 to sim 0.85) but also during the forecast period (sim 14% groundwater RMSE reduction and soil moisture correlation increase from sim 0.69 to sim 0.78) due to the effective impacts of the approach on both state and parameters.

Multivariate Assimilation of Remotely Sensed Soil Moisture and Evapotranspiration for Drought Monitoring

AbstractSoil moisture (SM) and evapotranspiration (ET) are key variables of the terrestrial water cycle with a strong relationship. This study examines remotely sensed soil moisture and evapotranspiration data assimilation (DA) with the aim of improving drought monitoring. Although numerous efforts have gone into assimilating satellite soil moisture observations into land surface models to improve their predictive skills, little attention has been given to the combined use of soil moisture and evapotranspiration to better characterize hydrologic fluxes. In this study, we assimilate two remotely sensed datasets, namely, Soil Moisture Operational Product System (SMOPS) and MODIS evapotranspiration (MODIS16 ET), at 1-km spatial resolution, into the VIC land surface model by means of an evolutionary particle filter method. To achieve this, a fully parallelized framework based on model and domain decomposition using a parallel divide-and-conquer algorithm was implemented. The findings show improvement in soil moisture predictions by multivariate assimilation of both ET and SM as compared to univariate scenarios. In addition, monthly and weekly drought maps are produced using the updated root-zone soil moisture percentiles over the Apalachicola–Chattahoochee–Flint basin in the southeastern United States. The model-based estimates are then compared against the corresponding U.S. Drought Monitor (USDM) archive maps. The results are consistent with the USDM maps during the winter and spring season considering the drought extents; however, the drought severity was found to be slightly higher according to DA method. Comparing different assimilation scenarios showed that ET assimilation results in wetter conditions comparing to open-loop and univariate SM DA. The multivariate DA then combines the effects of the two variables and provides an in-between condition.

Variable update strategy to improve water quality forecast accuracy in multivariate data assimilation using the ensemble Kalman filter

Data assimilation in complex water quality modeling is inevitably multivariate because several water quality variables interact and correlate. In ensemble Kalman filter applications, determining which variables to include and the structure of the relationships among these variables is important to achieve accurate forecast results. In this study, various analysis methods with different combinations of variables and interaction structures were evaluated under two different simulation conditions: synthetic and real. In the former, a synthetic experimental setting was formulated to ensure that issues, including incorrect model error assumption problem, spurious correlation between variables, and observational data inconsistency, would not distort the analysis results. The latter did not have such considerations. Therefore, this process could demonstrate the undistorted effects of the different analysis methods on the assimilated outputs and how these effects might diminish in real applications. Under synthetic conditions, updating a single active variable was found to improve the accuracy of the other active variables, and updating multiple active variables in a multivariate manner mutually enhanced the accuracy of the variables if proper ensemble covariance and observation data consistency were ensured. The results of the real case indicated a weakened mutual enhancement effect, and the methods in which variable localization were applied yielded the best analysis results. However, the multivariate analysis methods produced more accurate forecasting results, indicating that these methods could be superior. Therefore, it is suggested that multivariate analysis methods be considered first for water quality modeling, and the application of variable localization should be considered if significant spurious correlations and data inconsistency are present.

Multivariate data assimilation of GRACE, SMOS, SMAP measurements for improved regional soil moisture and groundwater storage estimates

Assimilating remote sensing observations into land surface models has become common practice to improve the accuracy of terrestrial water storage (TWS) estimates such as soil moisture and groundwater, for understanding the land surface interaction with the climate system, as well as assessing regional and global water resources. Such remote sensing observations include soil moisture information from the L-band Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) missions, and TWS information from the Gravity Recovery And Climate Experiment (GRACE). This study evaluates the benefit of assimilating them into the Community Atmosphere and Biosphere Land Exchange (CABLE) land surface model. The evaluation is conducted in the Goulburn River catchment, South-East Australia, where various in situ soil moisture and groundwater level data are available for validating data assimilation (DA) approaches. It is found that the performance of DA mainly depends on the type of observations that are assimilated. The SMOS/SMAP-only assimilation (SM DA) improves the top soil moisture but degrades the groundwater storage estimates, whereas the GRACE-only assimilation (GRACE DA) improves only the groundwater component. Assimilating both observations (multivariate DA) results in increased accuracy of both soil moisture and groundwater storage estimates. These findings demonstrate the added value of multivariate DA for simultaneously improving different model states, thus leading to a more robust DA system.

NCA-LDAS Land Analysis: Development and Performance of a Multisensor, Multivariate Land Data Assimilation System for the National Climate Assessment

Abstract This article describes one of the first successful examples of multisensor, multivariate land data assimilation, encompassing a large suite of soil moisture, snow depth, snow cover, and irrigation intensity environmental data records (EDRs) from the Scanning Multichannel Microwave Radiometer (SMMR), Special Sensor Microwave Imager (SSM/I), Advanced Scatterometer (ASCAT), Moderate-Resolution Imaging Spectroradiometer (MODIS), Advanced Microwave Scanning Radiometer (AMSR-E and AMSR2), Soil Moisture Ocean Salinity (SMOS) mission, and Soil Moisture Active Passive (SMAP) mission. The analysis is performed using the NASA Land Information System (LIS) as an enabling tool for the U.S. National Climate Assessment (NCA). The performance of the NCA Land Data Assimilation System (NCA-LDAS) is evaluated by comparing it to a number of hydrological reference data products. Results indicate that multivariate assimilation provides systematic improvements in simulated soil moisture and snow depth, with marginal effects on the accuracy of simulated streamflow and evapotranspiration. An important conclusion is that across all evaluated variables, assimilation of data from increasingly more modern sensors (e.g., SMOS, SMAP, AMSR2, ASCAT) produces more skillful results than assimilation of data from older sensors (e.g., SMMR, SSM/I, AMSR-E). The evaluation also indicates the high skill of NCA-LDAS when compared with other LSM products. Further, drought indicators based on NCA-LDAS output suggest a trend of longer and more severe droughts over parts of the western United States during 1979–2015, particularly in the southwestern United States, consistent with the trends from the U.S. Drought Monitor, albeit for a shorter 2000–15 time period.

Multivariate EnOI-based data assimilation in the high resolution ocean model

The ensemble Optimal Interpolation (EnOI) method is considered for assimilation of observational data in the INMIO ocean model of high spatial resolution. Its parallel implementation is described in the form of the Data Assimilation Service (DAS) based on the Compact Modeling Framework (CMF3.0). The on EnOI based parallel algorithm is designed to assimilate data from various sources (drifters, satellites, etc) with the ability to construct cross-covariance matrices (between different model values) for adjusting all model fields and the model forecast as a whole, including those functions, for which there is no measurements. This method based on EnOI is planned to be used for all observational data available to date for the World Ocean.

Assimilation of Sea Surface Temperature in a Global Hybrid Coordinate Ocean Model

The Hybrid Coordinate Ocean Model (HYCOM) uses different vertical coordinate choices in different regions. In HYCOM, the prognostic variables include not only the seawater temperature, salinity and current fields, but also the layer thickness. All prognostic variables are usually adjusted in the assimilation when multivariate data assimilation methods are used to assimilate sea surface temperature (SST). This paper investigates the effects of SST assimilation in a global HYCOM model using the Ensemble Optimal Interpolation multivariate assimilation method. Three assimilation experiments are conducted from 2006-08. In the first experiment, all model variables are adjusted during the assimilation process. In the other two experiments, the temperature alone is adjusted in the entire water column and in the mixed layer. For comparison, a control experiment without assimilation is also conducted. The three assimilation experiments yield notable SST improvements over the results of the control experiment. Additionally, the experiments in which all variables are adjusted and the temperature alone in all model layers is adjusted, produce significant negative effects on the subsurface temperature. Also, they yield negative effects on the subsurface salinity because it is associated with temperature and layer thickness. The experiment adjusting the temperature alone in the mixed layer yields positive effects and outperforms the other experiments. The heat content in the upper 300 m and 300-700 m layers further suggests that it yields the best performance among the experiments.

A particle filter scheme for multivariate data assimilation into a point-scale snowpack model in an Alpine environment

Abstract. The accuracy of hydrological predictions in snow-dominated regions deeply depends on the quality of the snowpack simulations, with dynamics that strongly affect the local hydrological regime, especially during the melting period. With the aim of reducing the modelling uncertainty, data assimilation techniques are increasingly being implemented for operational purposes. This study aims to investigate the performance of a multivariate sequential importance resampling – particle filter scheme, designed to jointly assimilate several ground-based snow observations. The system, which relies on a multilayer energy-balance snow model, has been tested at three Alpine sites: Col de Porte (France), Torgnon (Italy), and Weissfluhjoch (Switzerland). The implementation of a multivariate data assimilation scheme faces several challenging issues, which are here addressed and extensively discussed: (1) the effectiveness of the perturbation of the meteorological forcing data in preventing the sample impoverishment; (2) the impact of the parameter perturbation on the filter updating of the snowpack state; the system sensitivity to (3) the frequency of the assimilated observations, and (4) the ensemble size.The perturbation of the meteorological forcing data generally turns out to be insufficient for preventing the sample impoverishment of the particle sample, which is highly limited when jointly perturbating key model parameters. However, the parameter perturbation sharpens the system sensitivity to the frequency of the assimilated observations, which can be successfully relaxed by introducing indirectly estimated information on snow-mass-related variables. The ensemble size is found not to greatly impact the filter performance in this point-scale application.

Combined assimilation of streamflow and snow water equivalent for mid-term ensemble streamflow forecasts in snow-dominated regions

Abstract. The potential of data assimilation for hydrologic predictions has been demonstrated in many research studies. Watersheds over which multiple observation types are available can potentially further benefit from data assimilation by having multiple updated states from which hydrologic predictions can be generated. However, the magnitude and time span of the impact of the assimilation of an observation varies according not only to its type, but also to the variables included in the state vector. This study examines the impact of multivariate synthetic data assimilation using the ensemble Kalman filter (EnKF) into the spatially distributed hydrologic model CEQUEAU for the mountainous Nechako River located in British Columbia, Canada. Synthetic data include daily snow cover area (SCA), daily measurements of snow water equivalent (SWE) at three different locations and daily streamflow data at the watershed outlet. Results show a large variability of the continuous rank probability skill score over a wide range of prediction horizons (days to weeks) depending on the state vector configuration and the type of observations assimilated. Overall, the variables most closely linearly linked to the observations are the ones worth considering adding to the state vector due to the limitations imposed by the EnKF. The performance of the assimilation of basin-wide SCA, which does not have a decent proxy among potential state variables, does not surpass the open loop for any of the simulated variables. However, the assimilation of streamflow offers major improvements steadily throughout the year, but mainly over the short-term (up to 5 days) forecast horizons, while the impact of the assimilation of SWE gains more importance during the snowmelt period over the mid-term (up to 50 days) forecast horizon compared with open loop. The combined assimilation of streamflow and SWE performs better than their individual counterparts, offering improvements over all forecast horizons considered and throughout the whole year, including the critical period of snowmelt. This highlights the potential benefit of using multivariate data assimilation for streamflow predictions in snow-dominated regions.