C33B-01
Limitations and future of global snow observations
Comparisons between the climatology derived from global models and from around 30 years of passive microwave observations indicate large discrepancies in snow water equivalent, particularly in Siberia, so there is a need to understand the errors in both observations and models in order to account for the difference. The algorithm used to derive the measurements depends on the difference in brightness temperature measured at two different frequencies, and assumes constant snow grain size and density temporally and spatially. A sensitivity study of the Helsinki University of Technology snow microwave emission model has showed that Mie scattering theory, on which the retrieval algorithm is based, is strongly sensitive to grain size but is much less sensitive to snow depth and mass. Consequently, retrievals of snow mass are generally inaccurate at a point, but the error is reduced when spatial heterogeneity of the snow properties is taken into account, which indicates that retrievals are more accurate over larger areas. Given the sensitivity of the model to snow grain size, a new retrieval system that incorporates information about the grain size should lead to more accurate snow retrievals. Grain size may be derived from other sources such as visible radiation, or could be modelled. It may even be possible to determine grain size directly from the passive microwave measurements from multiple frequencies and dual polarizations offered by current instruments, although the relation between the physical and conceptual grain size must be understood. A data assimilation framework offers the potential to combine the models and measurements synergistically provided that the errors in both are quantified. This technique also allows data from different sources to be combined and collectively improve the snow model. Although significant challenges remain, a new retrieval method is planned that will combine snow observations from multiple instruments with a physically-based snow model. This system will ultimately be used to reprocess historical passive microwave datasets in order to produce a new snow climatology suitable for comparison with global models.
C33B-02
Finland Validation of the New AFWA-NASA Blended Snow Products
As part of an ongoing effort to validate satellite remote sensing snow products for the recently-developed U.S. Air Force Weather Agency (AFWA) - NASA blended snow product, satellite and in-situ data for snow extent and snow water equivalent (SWE) are evaluated in Finland for the 2006-2007 snow season Finnish Meteorological Institute (FMI) daily weather station data and Finnish Environment Institute (SYKE) bi-monthly snow course data are used as ground truth. Initial comparison results display positive agreement between the AFWA NASA Snow Algorithm (ANSA) snow extent and SWE maps and in situ data, with discrepancies in accordance with known AMSR-E and MODIS snow mapping limitations. Future ANSA product improvement plans include additional validation and inclusion of fractional snow cover in the ANSA data product. Furthermore, the AMSR-E 19 GHz (horizontal channel) with the difference between ascending and descending satellite passes (Diurnal Amplitude Variations, DAV) will be used to detect the onset of melt, and QuikSCAT scatterometer data (14 GHz) will be used to map areas of actively melting snow.
C33B-03 INVITED
Assessing global snow water equivalent estimates from space using AMSR-E satellite- based observations: are we any closer to characterizing northern hemisphere seasonal snow accumulation?
For nearly 30 years, satellite passive microwave instruments have been used to observe the Earth's surface with an aim of gaining insight into the water mass and energy status of the planet. Snow accumulation is one of the most seasonally-dynamic hydrologic water stores and its accurate assessment is critical for effective water resource management and for assessing the nature of changes to water cycle dynamics. Yet, satellite passive microwave observations still struggle to accurately quantify this important global water stock. A fundamental question that continues to drive research in this field is "to what extent can we use passive microwave observations to estimate seasonal snow accumulation, particularly snow water equivalent?". To answer this question, two key elements need to be considered: first, what independent data are appropriate to test satellite passive microwave estimates, and second, what are the most promising passive microwave algorithm approaches for estimating global SWE? Both of these elements require an understanding of the process-scale spatial variability of SWE, which is often terrain-dependent, and a cognizance of the microwave emission properties of snow. In this paper the efficacy of selected historical and state-of-the-art algorithms applied to Advanced Microwave Scanning Radiometer - EOS observations is assessed for northern hemisphere winter seasons during the 2002-2008 period. They are tested using several common field data sets acquired between 2002-2008 and illustrate the relative merits of each approach. The assessment is also used to determine whether we are any closer to characterizing northern hemisphere SWE from passive microwave observations.
C33B-04 INVITED
Analyzing snow properties using global models and observations: The current status and future perspectives
Snow properties have been monitored for decades through a number of vastly different observation systems and models. Today, approximately 2000 in-situ measurements of snow depth are available each day in near real time. Numerical weather prediction models provide estimates of snow water equivalent and snow coverage and numerous satellite-derived snow products are available routinely. It is somewhat surprising that only very few operational data assimilation systems for the integration of measurements from different observation systems and model data have been developed. In this presentation, the operational snow analysis from the European Centre for Medium-range Weather Forecasts will be introduced. The system makes use of the modelled first guess, in-situ observations and the operational NOAA NESDIS snow cover product. Global snow coverage has been verified against independent observations from MODIS; snow water equivalent has been compared against the standard AMSR-E retrieval products. It has been found that the NOAA NESDIS satellite product considerably improves the operational analysis. For the domain of the United States the ECMWF analyses have been compared against output from the SNODAS analysis system. Again, the impact of the satellite data on the ECMWF analyses has been positive. However, significant changes to the numerical models and data assimilation systems are needed to make optimal use of current and future satellite-derived data sets. In the second half of the presentation, ESA's current and planned activities related to operational snow products will be outlined and discussed with respect to future assimilation schemes.
C33B-05 INVITED
Validation and Further Development of HUT Snow Emission Model for Satellite Microwave Radiometer Data Inversion and Assimilation
Global mapping of terrestrial snow cover and sea ice is an important application area for space-borne microwave radiometry. The algorithms applied e.g. for estimating snow water equivalent (SWE) oare typically empirical formulas obtained by data delineation or by fitting a regression model to brightness temperatures simulated by a physical forward model. On the other hand, algorithms based on the inversion of a forward brightness temperature model have been also applied. In all cases, an essential factor determining the performance of inversion algorithms or giving information on the validity of empirical approaches is the accuracy of forward brightness temperature modelling. The techniques used for modeling the brightness temperature of snow pack are typically based on the radiative transfer equation. In semi-empirical approaches, some simplifications are made to avoid numerical integration, which enables the use of forward models e.g. in iterative inversion algorithms. Relevant semi- empirical models include the HUT snow emission model and the MEMLS model. They consider snow pack as a single layer (HUT model) or as a multi-layer structure (MEMLS). In practice, the use of any model with space-borne or airborne radiometer data requires that snow ground-interactions as well as influence of vegetation and atmosphere have to be considered. This significantly complicates the modeling task. The HUT snow emission model was developed in 1998 to describe the microwave brightness temperature of snow covered forested terrain. Revisions to improve the consideration of snow extinction coefficient have been made in recent years by different research groups. The model treats different emission contributions and their interactions with simple analytical formulas that are derived either through empirical or semi- theoretical considerations. As the model is relatively simple, it can be used for the inversion of space-borne data in the estimation of quantitative snow pack characteristics. However, the limitations of the model include that it describes a snow pack as a single layer with a certain average density and a certain (effective) grain size. Thus, the applicability of model to simulate the brightness temperature of a vertically strongly stratified snow pack is limited. Here, a new multi-layer version of the HUT snow emission model is presented in order to better describe the emission behaviour of layered snow pack and that of snow covered lake/sea ice. This enables model predictions for the case of depth hoar layer, which is essential in higher latitudes with severe winter conditions. Additionally, it enables model simulations for snow covered lake ice. Recent investigations have shown that the influence of lakes is a major problem concerning the performance of SWE estimation from space-borne microwave radiometer data. The comparison of model predictions with experimental indicate that reasonable predictions for the brightness temperature of snow covered lake ice can be obtained with the modified HUT model. Moreover, model predictions using detailed in situ soil-snow-atmospheric parameter information as input indicate fairly good agreement for a period of two years. In the former case, airborne multi-channel HUTRAD observations were used as experimental radiometer data. They were obtained in years 2005 and 2006 from extensive transects across Finland. In the latter case, AMSR-E observations from a single region presenting seasonal cycles were applied.
C33B-06
Assimilating MODIS Snow Cover Fraction into CLM to Improve Snow Water Equivalent Estimates for Complex Terrain Areas
Much interest exists in improving hydrological predictions in complex terrain regions using high spatial
resolution land surface observations, such as snow measurements. Assimilation of snow observations in
land surface models is a promising approach to enhance runoff timing and discharges, mainly during critical
melt periods. The study area involves much of Washington state with a focus on the Yakima River Basin,
which is located on the eastern side of the Cascade Mountain Range in Washington. Remotely sensed snow
cover products are assimilated into the Community Land Model using the Land Information System (LIS) to
obtain more accurate snow water equivalent (SWE) analyses. The Moderate Resolution Imaging
Spectroradiometer (MODIS) snow cover area (SCA) and fractional (SCF) products are the main observations
used in the assimilation procedures. One of the challenges in using SCA is that it only provides a partial
source of information needed to estimate the SWE state. This study explores two different alternatives in
utilizing these observations for state optimization, through the use of different assimilation techniques with
disparate observation operators. Two of the assimilation methods applied include a form of direct insertion
(DI), which takes in to account a priori knowledge of the LSM temperature biases, and an Ensemble Kalman
Filter (EnKF) approach, which will use derived snow depletion curve estimates for the area. In-situ
measurements of SWE, snow depth and runoff are used for validation. Comparisons of these assimilation
approaches using MODIS snow cover products will be presented.
http://www.iges.org
C33B-07 INVITED
Radiance assimilation shows promise for snowpack characterization: a 1-d case study
We demonstrate an ensemble-based radiometric data assimilation (DA) methodology for estimating snow depth and snow grain size using ground-based passive microwave (PM) observations at 18.7 and 36.5 GHz collected during the NASA CLPX-1, March 2003, Colorado, USA. A land surface model was used to develop a prior estimate of the snowpack states, and a radiative transfer model was used to relate the modeled states to the observations. Snow depth bias was –53.3 cm prior to the assimilation, and –7.3 cm after the assimilation. Snow depth estimated by a non-DA-based retrieval algorithm using the same PM data had a bias of –18.3 cm. The sensitivity of the assimilation scheme to the grain size uncertainty was evaluated; over the range of grain size uncertainty tested, the posterior snow depth estimate bias ranges from -2.99 cm to - 9.85 cm, which is uniformly better than both the prior and retrieval estimates. This study demonstrates the potential applicability of radiometric DA at larger scales.