Bias Correction Scheme Research Articles

Recursive estimators of mean-areal and local bias in precipitation products that account for conditional bias

This paper presents novel formulations of Mean field bias (MFB) and local bias (LB) correction schemes that incorporate conditional bias (CB) penalty. These schemes are based on the operational MFB and LB algorithms in the National Weather Service (NWS) Multisensor Precipitation Estimator (MPE). By incorporating CB penalty in the cost function of exponential smoothers, we are able to derive augmented versions of recursive estimators of MFB and LB. Two extended versions of MFB algorithms are presented, one incorporating spatial variation of gauge locations only (MFB-L), and the second integrating both gauge locations and CB penalty (MFB-X). These two MFB schemes and the extended LB scheme (LB-X) are assessed relative to the original MFB and LB algorithms (referred to as MFB-O and LB-O, respectively) through a retrospective experiment over a radar domain in north-central Texas, and through a synthetic experiment over the Mid-Atlantic region. The outcome of the former experiment indicates that introducing the CB penalty to the MFB formulation leads to small, but consistent improvements in bias and CB, while its impacts on hourly correlation and Root Mean Square Error (RMSE) are mixed. Incorporating CB penalty in LB formulation tends to improve the RMSE at high rainfall thresholds, but its impacts on bias are also mixed. The synthetic experiment suggests that beneficial impacts are more conspicuous at low gauge density (9 per 58,000 km2), and tend to diminish at higher gauge density. The improvement at high rainfall intensity is partly an outcome of the conservativeness of the extended LB scheme. This conservativeness arises in part from the more frequent presence of negative eigenvalues in the extended covariance matrix which leads to no, or smaller incremental changes to the smoothed rainfall amounts.

C-GLORSv5: an improved multipurpose global ocean eddy-permitting physical reanalysis

Abstract. Global ocean reanalyses combine in situ and satellite ocean observations with a general circulation ocean model to estimate the time-evolving state of the ocean, and they represent a valuable tool for a variety of applications, ranging from climate monitoring and process studies to downstream applications, initialization of long-range forecasts and regional studies. The purpose of this paper is to document the recent upgrade of C-GLORS (version 5), the latest ocean reanalysis produced at the Centro Euro-Mediterraneo per i Cambiamenti Climatici (CMCC) that covers the meteorological satellite era (1980–present) and it is being updated in delayed time mode. The reanalysis is run at eddy-permitting resolution (1∕4° horizontal resolution and 50 vertical levels) and consists of a three-dimensional variational data assimilation system, a surface nudging and a bias correction scheme. With respect to the previous version (v4), C-GLORSv5 contains a number of improvements. In particular, background- and observation-error covariances have been retuned, allowing a flow-dependent inflation in the globally averaged background-error variance. An additional constraint on the Arctic sea-ice thickness was introduced, leading to a realistic ice volume evolution. Finally, the bias correction scheme and the initialization strategy were retuned. Results document that the new reanalysis outperforms the previous version in many aspects, especially in representing the variability of global heat content and associated steric sea level in the last decade, the top 80 m ocean temperature biases and root mean square errors, and the Atlantic Ocean meridional overturning circulation; slight worsening in the high-latitude salinity and deep ocean temperature emerge though, providing the motivation for further tuning of the reanalysis system. The dataset is available in NetCDF format at doi:10.1594/PANGAEA.857995.

A Long-Term Wind Speed Ensemble Forecasting System with Weather Adapted Correction

Wind forecasting is critical in the wind power industry, yet forecasting errors often exist. In order to effectively correct the forecasting error, this study develops a weather adapted bias correction scheme on the basis of an average bias-correction method, which considers the deviation of estimated biases associated with the difference in weather type within each unit of the statistical sample. This method is tested by an ensemble forecasting system based on the Weather Research and Forecasting (WRF) model. This system provides high resolution wind speed deterministic forecasts using 40 members generated by initial perturbations and multi-physical schemes. The forecasting system outputs 28–52 h predictions with a temporal resolution of 15 min, and is evaluated against collocated anemometer towers observations at six wind fields located on the east coast of China. Results show that the information contained in weather types produces an improvement in the forecast bias correction.

Observation bias correction schemes in data assimilation systems: a theoretical study of some of their properties

Observation bias correction schemes are important components of the data assimilation (DA) systems used in operational numerical weather prediction (NWP). They are used at present mainly to correct for biases in satellite radiance observations and their observation operators. These schemes attempt to remove biases in observations relative to the NWP model background or analysis field. The presence of bias in the NWP model itself can substantially complicate this process. Using a very simple scalar forecast model and DA system, this study explores the extent to which model bias leads to biases in the background and analysis states. Theory is presented for systems that attempt to remove observation bias relative to the background or relative to the analysis. It is shown that background and analysis biases are functions of three parameters: the weights given within the DA system to the observations to be bias‐corrected and to the other ‘anchoring’ observations, and the rate at which the NWP model state relaxes towards its own climatology. These effects are quantified for ‘baseline’ values of these three parameters intended to be representative of the current Met Office global NWP system, and for large variations around these baseline values. If the baseline values are indeed representative, then background and analysis biases in the range 0.21–0.33 of the model bias are expected. However, substantial variations are expected within the model domain, depending on variations in observation density, in the balance between bias‐corrected and anchoring observations, and in the rate at which the NWP model state relaxes towards its own climatology. Moreover, the effect of model bias on background and analysis biases will increase as more observations are bias‐corrected and a smaller proportion are used as ‘anchor’ observations. This has important implications for observation bias correction strategies used in NWP and in reanalyses.

Statistical bias correction for creating coherent total ozone record from OMI and OMPS observations

Abstract A long-term coherent total column ozone (TCO) record is essential to ozone layer variability assessment, especially the detection of early signs of ozone recovery after years of depletion. Because of differences in satellite platforms and instruments design, calibration, and retrieval algorithms, however, significant cross-mission biases are observed between multiple sensor TCO observations in the common time–space domain. To attain a coherent TCO record, observed cross-mission biases should be accurately addressed prior to the data-merging scheme. In this study, a modified statistical bias correction method was proposed based on the quantile-quantile adjustment to remove apparent cross-mission TCO biases between the Ozone Monitoring Instrument (OMI) and Ozone Mapping and Profiler Suite (OMPS). To evaluate the effectiveness of this modified algorithm, the overall inconsistency (OI), a unique time-series similarity measure, was proposed to quantify the improvements of consistency (or similarity) between cross-mission TCO time series data before and after bias correction. Common observations during the overlapped time period of 2012–2015 were used to characterize the systematic bias between OMPS and OMI through the modified bias correction method. TCO observations from OMI during 2004–2015 were then projected to the OMPS level by removing associated cross-mission biases. This modified bias correction scheme significantly improved the overall consistency, with an average improvement of 90% during the overlapped time period at the global scale. In addition to the evaluation of consistency improvements before and after bias correction, impacts of cross-mission biases on long-term trend estimations were also investigated. Comparisons of derived trends from the merged TCO time series before and after bias correction across 38 ground-based stations indicate that cross-mission biases not only affect magnitudes of estimated trends, but also result in different phases of trends. Further comparisons of estimated seasonal TCO trends before and after bias correction at the global scale suggest that trends derived from the bias-corrected time series are more accurate than those without bias correction. Overall, the bias correction scheme developed in this study is essential for preparing an accurate long-term TCO record representative of trend analysis to support future assessment of ozone recovery at the global scale.

Precipitation ensembles conforming to natural variations derived from a regional climate model using a new bias correction scheme

Abstract. This study presents a novel bias correction scheme for regional climate model (RCM) precipitation ensembles. A primary advantage of using model ensembles for climate change impact studies is that the uncertainties associated with the systematic error can be quantified through the ensemble spread. Currently, however, most of the conventional bias correction methods adjust all the ensemble members to one reference observation. As a result, the ensemble spread is degraded during bias correction. Since the observation is only one case of many possible realizations due to the climate natural variability, a successful bias correction scheme should preserve the ensemble spread within the bounds of its natural variability (i.e. sampling uncertainty). To demonstrate a new bias correction scheme conforming to RCM precipitation ensembles, an application to the Thorverton catchment in the south-west of England is presented. For the ensemble, 11 members from the Hadley Centre Regional Climate Model (HadRM3-PPE) data are used and monthly bias correction has been done for the baseline time period from 1961 to 1990. In the typical conventional method, monthly mean precipitation of each of the ensemble members is nearly identical to the observation, i.e. the ensemble spread is removed. In contrast, the proposed method corrects the bias while maintaining the ensemble spread within the natural variability of the observations.

Estimation and correction of model bias in the NASA/GMAO GEOS5 data assimilation system: Sequential implementation

This study presents a simplified multivariate bias correction scheme that is sequentially implemented in the GEOS5 data assimilation system and compared against a control experiment without model bias correction. The results show considerable improvement in terms of the mean biases of rawinsonde observation-minus-background (OmB) residuals for observed water vapor, wind and temperature variables. The time series spectral analysis shows whitening of bias-corrected OmB residuals, and mean biases for rawinsonde observation-minus-analysis (OmA) are also improved. Some wind and temperature biases in the control experiment near the equatorial tropopause nearly vanish from the bias-corrected experiment. Despite the analysis improvement, the bias correction scheme has only a moderate impact on forecast skill. Significant interaction is also found among quality-control, satellite observation bias correction, and background bias correction, and the latter positively impacts satellite bias correction.

Sea Surface Temperature Climate Data Record for the North Sea and Baltic Sea

Abstract A 30-yr climate data record (CDR) of sea surface temperature (SST) has been produced with daily gap-free analysis fields for the North Sea and the Baltic Sea region from 1982 to 2012 by combining the Pathfinder AVHRR satellite data record with the Along-Track Scanning Radiometer (ATSR) Reprocessing for Climate (ARC) dataset and with in situ observations. A dynamical bias correction scheme adjusts the Pathfinder observations toward the ARC and in situ observations. Largest Pathfinder–ARC differences are found in the summer months, when the Pathfinder observations are up to 0.4°C colder than the ARC observations on average. Validation against independent in situ observations shows a very stable performance of the data record, with a mean difference of −0.06°C compared to moored buoys and a 0.46°C standard deviation of the differences. The mean annual biases of the SST CDR are small for all years, with a negligible temporal trend when compared against drifting and moored buoys. Analysis of the SST CDR reveals that the monthly anomalies for the North Sea, the Danish straits, and the central Baltic Sea regions show a high degree of correlation for interannual and decadal time scales, whereas the monthly variability differs from one region to another. The linear trends of the 1982–2012 SST anomaly time series range from 0.037°C yr−1 for the North Sea to 0.041°C yr−1 for the Baltic Sea.

A novel bias correction methodology for climate impact simulations

Abstract. Understanding, quantifying and attributing the impacts of extreme weather and climate events in the terrestrial biosphere is crucial for societal adaptation in a changing climate. However, climate model simulations generated for this purpose typically exhibit biases in their output that hinder any straightforward assessment of impacts. To overcome this issue, various bias correction strategies are routinely used to alleviate climate model deficiencies, most of which have been criticized for physical inconsistency and the nonpreservation of the multivariate correlation structure. In this study, we introduce a novel, resampling-based bias correction scheme that fully preserves the physical consistency and multivariate correlation structure of the model output. This procedure strongly improves the representation of climatic extremes and variability in a large regional climate model ensemble (HadRM3P, climateprediction.net/weatherathome), which is illustrated for summer extremes in temperature and rainfall over Central Europe. Moreover, we simulate biosphere–atmosphere fluxes of carbon and water using a terrestrial ecosystem model (LPJmL) driven by the bias-corrected climate forcing. The resampling-based bias correction yields strongly improved statistical distributions of carbon and water fluxes, including the extremes. Our results thus highlight the importance of carefully considering statistical moments beyond the mean for climate impact simulations. In conclusion, the present study introduces an approach to alleviate climate model biases in a physically consistent way and demonstrates that this yields strongly improved simulations of climate extremes and associated impacts in the terrestrial biosphere. A wider uptake of our methodology by the climate and impact modelling community therefore seems desirable for accurately quantifying changes in past, current and future extremes.

Bias and observation error characterization of the Fengyun-2F Stretched Visible and Infrared Spin Scan Radiometer with the aim of its radiance assimilation

ABSTRACTRecent studies start to explore the potential benefits of directly assimilating the radiance from the infrared imagers onboard geostationary satellites to the short-range numerical weather prediction both on the synoptic- and meso-scales, where accurate quantification of the bias and observation error arising from the numerical model and the instrument is a key component in such experiments. This paper investigates both the bias and the observation errors of the Stretched Visible and Infrared Spin Scan Radiometer (S-VISSR) onboard Chinese operational geostationary satellite Fengyun-2F, as the first step towards its clear-sky radiance assimilation. In this study, 774,530 observations that passed a five-step quality control procedure during the period of 1–7 July in 2014 are collected to investigate the mentioned two quantities while the synthetic simulated clear radiance is calculated through a numerical weather prediction model and a radiative transfer model. Persistent warm biases are found for FY-2 F S-VISSR channel 3, 4 and 5 during day and night, and they are possibly due to the uncertainty from the instrument calibration process. The observation errors of S-VISSR infrared channels vary with the surface types, generally with larger values over land than over ocean, which could partially be explained by the variation of the land types over land. Besides, whether these biases change with viewing geometry or evolve with time is also examined in our study. Biases from the three surface channels exhibit complex dependence on the scan position and evolve with time, which encourages incorporating both the scan position and temporal corrections into the S-VISSR bias correction scheme in the data assimilation system.

Estimating forecast model bias in coupled global and limited-area models

Idealised perfect model experiments suggest that performing data assimilation on a ‘composite’ state vector, constructed from global and limited-area model states, can be beneficial to both model states. Here, an illustrative scheme is implemented to account for systematic forecast errors attributed to the imperfect model dynamics. Results from numerical experiments suggest that even simple bias correction schemes can correct forecast errors in composite states.

Bias correction methods for regional climate model simulations considering the distributional parametric uncertainty underlying the observations

In this paper, we present a comparative study of bias correction methods for regional climate model simulations considering the distributional parametric uncertainty underlying the observations/models. In traditional bias correction schemes, the statistics of the simulated model outputs are adjusted to those of the observation data. However, the model output and the observation data are only one case (i.e., realization) out of many possibilities, rather than being sampled from the entire population of a certain distribution due to internal climate variability. This issue has not been considered in the bias correction schemes of the existing climate change studies. Here, three approaches are employed to explore this issue, with the intention of providing a practical tool for bias correction of daily rainfall for use in hydrologic models ((1) conventional method, (2) non-informative Bayesian method, and (3) informative Bayesian method using a Weather Generator (WG) data). The results show some plausible uncertainty ranges of precipitation after correcting for the bias of RCM precipitation. The informative Bayesian approach shows a narrower uncertainty range by approximately 25–45% than the non-informative Bayesian method after bias correction for the baseline period. This indicates that the prior distribution derived from WG may assist in reducing the uncertainty associated with parameters. The implications of our results are of great importance in hydrological impact assessments of climate change because they are related to actions for mitigation and adaptation to climate change. Since this is a proof of concept study that mainly illustrates the logic of the analysis for uncertainty-based bias correction, future research exploring the impacts of uncertainty on climate impact assessments and how to utilize uncertainty while planning mitigation and adaptation strategies is still needed.

Impact of the assimilation of INSAT‐3D radiances on short‐range weather forecasts

The present study concentrates on the assimilation of clear‐sky radiances from the recently launched INSAT‐3D satellite. The imager and sounder are two primary meteorological instruments aboard the INSAT‐3D satellite. Pre‐assimilation monitoring of the radiances has been carried out from April to July 2014. A double‐difference technique using the High Resolution Infrared Sounder (HIRS) on MetOp‐A is employed to remove the effect of deficiencies in the model‐analysed profiles which can contribute to the biases. Observed radiances of both instruments appear to be colder (∼1.5–2 K) with respect to radiative transfer model (RTM) simulated radiances, except water vapour channel radiances, which are warmer (0.5–1 K). The standard deviations of observed minus RTM‐simulated radiances are between 0.5 and 1.5 K. Prior to data assimilation, these biases are corrected using a variational bias correction scheme. The largest impacts on the analyses, when assimilating imager radiances, are found in the mid‐ and upper‐tropospheric moisture, while assimilation of sounder radiances impacted both moisture and temperature throughout the troposphere. The more accurate analyses with the INSAT‐3D radiance assimilation lead to improved moisture, wind, temperature and precipitation forecasts compared to the control case in which only conventional observations were assimilated. The results demonstrate the ability of temperature‐ and water vapour‐sensitive radiances to improve not only the temperature and moisture fields, but also the wind fields, enhancing their importance, particularly over tropical regions where wind observations are more important.

Field‐aligned neutral wind bias correction scheme for global ionospheric modeling at midlatitudes by assimilating FORMOSAT‐3/COSMIC hmF2 data under geomagnetically quiet conditions

AbstractThis study demonstrates the usage of a data assimilation procedure, which ingests the FORMOSAT‐3/COSMIC (F3/C) hmF2 observations to correct the model wind biases to enhance the capability of the new global Ionosphere Plasmasphere Electrodynamics (IPE) model under geomagnetically quiet conditions. The IPE model is built upon the field line interhemispheric plasma model with a realistic geomagnetic field model and empirical model drivers. The hmF2 observed by the F3/C radio occultation technique is utilized to adjust global thermospheric field‐aligned neutral winds (i.e., a component of the thermospheric neutral wind parallel to the magnetic field) at midlatitudes according to a linear relationship between time differentials of the field‐aligned wind and hmF2. The adjusted winds are further applied to drive the IPE model. The comparison of the modeled electron density with the observations of F3/C and ground‐based GPS receivers at the 2012 March equinox suggests that the modeled electron density can be significantly improved in the midlatitude regions of the Southern Hemisphere, if the wind correction scheme is applied. Moreover, the F3/C observation, the IPE model, and the wind bias correction scheme are applied to study the 2012 Southern Hemisphere Midlatitude Summer Nighttime Anomaly (southern MSNA)/Weddell Sea Anomaly (WSA) event at December solstice for examining the role of the neutral winds in controlling the longitudinal variation of the southern MSNA/WSA behavior. With the help of the wind bias correction scheme, the IPE model better tracks the F3/C‐observed eastward movement of the southern MSNA/WSA feature. The apparent eastward movement of the southern MSNA/WSA features in the local time coordinate is primarily caused by the longitudinal variation in the declination angle of the geomagnetic field that controls the field‐aligned projection of both geographic meridional and zonal components of the neutral wind. Both the IPE simulations and the F3/C observations show the significant longitudinal variation in the speed of the eastward movement of the southern MSNA/WSA.

Estimation of the Added Value of the Absolute Calibration of GPS Radio Occultation Data for Numerical Weather Prediction

Abstract An analysis of the impact of GPS radio occultation observations on Environment Canada’s global deterministic weather prediction system is presented. Radio occultation data, as any other source of weather observations, have a direct impact on the analyses. Since they are assimilated assuming that they are well calibrated, they also impact the bias correction scheme employed for other data, such as satellite radiances. The authors estimate the relative impact of occultation data obtained from, first, their assimilation as atmospheric measurements and, second, their influence on the bias correction for radiance data. This assessment is performed using several implementations of the thermodynamic relationships involved, and also allowing or blocking this influence to the radiance bias correction scheme. The current implementation of occultation operators at Environment Canada is presented, collecting upgrades that have been detailed elsewhere, such as the equation of state of air and the expression of refractivity. The performance of the system with and without assimilation of occultations is reviewed under conditions representative of current operations. Several denial runs are prepared, withdrawing only the occultation data from the assimilation, but keeping their influence on the radiance bias correction, or assimilating occultations but denying their impact on the bias correction procedure, and a complete denial. It is shown that the impact of occultations on the analysis is significant through both paths—assimilation and radiance bias correction—albeit the first is larger. The authors conclude that the traceability link of the ensemble of occultations has an added value, beyond the value of each datum as an atmospheric measurement.

WRF Dynamical Downscaling and Bias Correction Schemes for NCEP Estimated Hydro-Meteorological Variables

Rainfall and Reference Evapotranspiration (ETo) are the most fundamental and significant variables in hydrological modelling. However, these variables are generally not available over ungauged catchments. ETo estimation usually needs measurements of weather variables such as wind speed, air temperature, solar radiation and dew point. After the development of reanalysis global datasets such as the National Centre for Environmental Prediction (NCEP) and high performance modelling framework Weather Research and Forecasting (WRF) model, it is now possible to estimate the rainfall and ETo for any coordinates. In this study, the WRF modelling system was employed to downscale the global NCEP reanalysis datasets over the Brue catchment, England, U.K. After downscaling, two statistical bias correction schemes were used, the first was based on sophisticated computing algorithms i.e., Relevance Vector Machine (RVM), while the second was based on the more simple Generalized Linear Model (GLM). The statistical performance indices for bias correction such as %Bias, index of agreement (d), Root Mean Square Error (RMSE), and Correlation (r) indicated that the RVM model, on the whole, displayed a more accomplished bias correction of the variability of rainfall and ETo in comparison to the GLM. The study provides important information on the performance of WRF derived hydro-meteorological variables using NCEP global reanalysis datasets and statistical bias correction schemes which can be used in numerous hydro-meteorological applications.

Multi-scale analysis of bias correction of soil moisture

Abstract. Remote sensing, in situ networks and models are now providing unprecedented information for environmental monitoring. To conjunctively use multi-source data nominally representing an identical variable, one must resolve biases existing between these disparate sources, and the characteristics of the biases can be non-trivial due to spatio-temporal variability of the target variable, inter-sensor differences with variable measurement supports. One such example is of soil moisture (SM) monitoring. Triple collocation (TC) based bias correction is a powerful statistical method that is increasingly being used to address this issue, but is only applicable to the linear regime, whereas the non-linear method of statistical moment matching is susceptible to unintended biases originating from measurement error. Since different physical processes that influence SM dynamics may be distinguishable by their characteristic spatio-temporal scales, we propose a multi-timescale linear bias model in the framework of a wavelet-based multi-resolution analysis (MRA). The joint MRA-TC analysis was applied to demonstrate scale-dependent biases between in situ, remotely sensed and modelled SM, the influence of various prospective bias correction schemes on these biases, and lastly to enable multi-scale bias correction and data-adaptive, non-linear de-noising via wavelet thresholding.

Adjustment of wind-drift effect for real-time systematic error correction in radar rainfall data

An effective bias correction procedure using gauge measurement is a significant step for radar data processing to reduce the systematic error in hydrological applications. In these bias correction methods, the spatial matching of precipitation patterns between radar and gauge networks is an important premise. However, the wind-drift effect on radar measurement induces an inconsistent spatial relationship between radar and gauge measurements as the raindrops observed by radar do not fall vertically to the ground. Consequently, a rain gauge does not correspond to the radar pixel based on the projected location of the radar beam. In this study, we introduce an adjustment method to incorporate the wind-drift effect into a bias correlation scheme. We first simulate the trajectory of raindrops in the air using downscaled three-dimensional wind data from the weather research and forecasting model (WRF) and calculate the final location of raindrops on the ground. The displacement of rainfall is then estimated and a radar–gauge spatial relationship is reconstructed. Based on this, the local real-time biases of the bin-average radar data were estimated for 12 selected events. Then, the reference mean local gauge rainfall, mean local bias, and adjusted radar rainfall calculated with and without consideration of the wind-drift effect are compared for different events and locations. There are considerable differences for three estimators, indicating that wind drift has a considerable impact on the real-time radar bias correction. Based on these facts, we suggest bias correction schemes based on the spatial correlation between radar and gauge measurements should consider the adjustment of the wind-drift effect and the proposed adjustment method is a promising solution to achieve this.

An improved bias correction scheme based on comparative precipitation characteristics

AbstractBias correction is a necessary post‐processing procedure in order to use regional climate model (RCM) simulated local climate variables as the input data for hydrological models owing to systematic errors of RCMs. Most of present bias correction methods adjust statistical properties between observed and simulated data on the basis of calendar periods, e.g. month or season. However, this matching statistic is only a necessary condition, not a sufficient condition, as temporal distribution of the precipitation between observed and simulated data is ignored. This study suggests an improved bias correction scheme that considers not only statistical properties but also the temporal distribution between the time series of observed and modelled data. The ratio of the observed precipitation to simulated precipitation is used to compare the behaviour between the observed and modelled precipitation data, and three criteria are proposed when dividing bias correction periods: (1) underestimation of precipitation, (2) stability of /underestimation of precipitation, (2) stability of precipitation ratio and (3) oscillation of precipitation ratio. The results show that the output of the proposed bias correction method follows the trend of the observed precipitation better than that of the conventional bias correction method. This study indicates that temporal distribution should not be ignored when choosing a comparison period for bias correction. However, the study is only a preliminary attempt to address this important issue, and we hope it will stimulate more research activities to improve the methodology. Future efforts on several unsolved problems have been suggested such as how to find out the optimal group number to avoid the overfitting and underfitting conditions. Copyright © 2014 John Wiley & Sons, Ltd.

Dual Forcing and State Correction via Soil Moisture Assimilation for Improved Rainfall–Runoff Modeling

AbstractUncertainties in precipitation forcing and prestorm soil moisture states represent important sources of error in streamflow predictions obtained from a hydrologic model. An earlier synthetic twin experiment has demonstrated that error in both antecedent soil moisture states and rainfall forcing can be filtered by assimilating remotely sensed surface soil moisture retrievals. This opens up the possibility of applying satellite soil moisture estimates to address both key sources of error in hydrologic model predictions. Here, in an attempt to extend the synthetic analysis into a real-data environment, two satellite-based surface soil moisture products—based on both passive and active microwave remote sensing—are assimilated using the same dual forcing/state correction approach. A bias correction scheme is implemented to remove bias in background forecasts caused by synthetic perturbations in the ensemble filtering routines, and a triple collocation–based technique is adopted to derive rescaled observations and observation error variances. Results are largely in agreement with the earlier synthetic analysis. That is, the correction of satellite-derived rainfall forcing is able to improve streamflow prediction, especially during relatively high-flow periods. In contrast, prestorm soil moisture state correction is more efficient in improving the base flow component of streamflow. When rainfall and soil moisture state corrections are combined, the RMSE of both the high- and low-flow components of streamflow can be reduced by ~40% and ~30%, respectively. However, an unresolved issue is that soil moisture data assimilation also leads to underprediction of very intense precipitation/high-flow events.