A. James Clark School of Engineering

Permanent URI for this communityhttp://hdl.handle.net/1903/1654

The collections in this community comprise faculty research works, as well as graduate theses and dissertations.

Browse

Subject: search.filters.subject.data assimilation

Search Results

Now showing 1 - 7 of 7
  • Thumbnail Image
    Item
    Exploring the Utility of Machine Learning-Based Passive Microwave Brightness Temperature Data Assimilation over Terrestrial Snow in High Mountain Asia
    (MDPI, 2019-09-28) Kwon, Yonghwan; Forman, Barton A.; Ahmad, Jawairia A.; Kumar, Sujay V.; Yoon, Yeosang
    This study explores the use of a support vector machine (SVM) as the observation operator within a passive microwave brightness temperature data assimilation framework (herein SVM-DA) to enhance the characterization of snow water equivalent (SWE) over High Mountain Asia (HMA). A series of synthetic twin experiments were conducted with the NASA Land Information System (LIS) at a number of locations across HMA. Overall, the SVM-DA framework is effective at improving SWE estimates (~70% reduction in RMSE relative to the Open Loop) for SWE depths less than 200 mm during dry snowpack conditions. The SVM-DA framework also improves SWE estimates in deep, wet snow (~45% reduction in RMSE) when snow liquid water is well estimated by the land surface model, but can lead to model degradation when snow liquid water estimates diverge from values used during SVM training. In particular, two key challenges of using the SVM-DA framework were observed over deep, wet snowpacks. First, variations in snow liquid water content dominate the brightness temperature spectral difference (ΔTB) signal associated with emission from a wet snowpack, which can lead to abrupt changes in SWE during the analysis update. Second, the ensemble of SVM-based predictions can collapse (i.e., yield a near-zero standard deviation across the ensemble) when prior estimates of snow are outside the range of snow inputs used during the SVM training procedure. Such a scenario can lead to the presence of spurious error correlations between SWE and ΔTB, and as a consequence, can result in degraded SWE estimates from the analysis update. These degraded analysis updates can be largely mitigated by applying rule-based approaches. For example, restricting the SWE update when the standard deviation of the predicted ΔTB is greater than 0.05 K helps prevent the occurrence of filter divergence. Similarly, adding a thin layer (i.e., 5 mm) of SWE when the synthetic ΔTB is larger than 5 K can improve SVM-DA performance in the presence of a precipitation dry bias. The study demonstrates that a carefully constructed SVM-DA framework cognizant of the inherent limitations of passive microwave-based SWE estimation holds promise for snow mass data assimilation.
  • Thumbnail Image
    Item
    SMAP soil moisture assimilated Noah-MP model output
    (2021) Ahmad, Jawairia; Forman, Bart; Kumar, Sujay
    The data archived here includes the NASA Noah-MP (version 4.0.1) land surface model output used in the investigation of the impact of passive microwave-based soil moisture retrieval assimilation on soil moisture estimation in South Asia (Ahmad et al., 2021). SMAP soil moisture retrievals are assimilated into the Noah-MP land surface model to improve the estimation of soil moisture and other related states. The open loop (OL) represents Noah-MP’s modeling capabilities using MERRA2 and IMERG precipitation. Two different types of data assimilation runs were executed using the MERRA2 and IMERG precipitation boundary conditions, i.e., with CDF-matching (DA-CDF) and without CDF matching (DA-NoCDF). The key findings in this paper include: 1) assimilation results without any CDF-matching yielded the lowest error in estimated soil moisture, 2) the best goodness-of-fit statistics were achieved for the IMERG-forced DA-NoCDF soil moisture experiment, 3) biases associated with unmodeled hydrologic processes such as irrigation were corrected via assimilation, and 4) the highest influence of assimilation was observed across croplands.
  • Thumbnail Image
    Item
    Estimating terrestrial water budget components across high mountain Asia using remote sensing, data assimilation, and machine learning
    (2021) Ahmad, Jawairia; Forman, Barton A.; Civil Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Contemporary studies have predicted a vulnerable future for key water budget components across high mountain Asia (HMA) and the adjoining areas. Considering the regional population and its dependence on agrarian economies, it is imperative that efforts be channelized towards improving the estimation of the hydrologic cycle across HMA. In this study, data assimilation methods were employed to assimilate remotely-sensed observations into land surface models to improve snow mass, soil moisture, and runoff estimates. The NASA Land Information System was used to simulate the hydrologic cycle across HMA and the adjoining areas using the Noah-MP land surface model. In an effort to improve snow mass estimation, passive microwave brightness temperature spectral differences (∆Tb) from the Advanced Microwave Scanning Radiometer-2 (AMSR2) were assimilated into Noah-MP snow mass estimates. Support vector machine regression, a supervised machine learning technique, was used as the observation operator to map the geophysical states into the observed ∆Tb space. Evaluation of the assimilation routine highlighted the decrease in domain-wide snow mass bias. The assimilation framework proved to be more effective during the (dry) snow accumulation season resulting in decreased snow mass bias and RMSE at 76% and 58% of the comparative locations, respectively. Diagnostic metrics such as the innovation sequence were studied to assess the snow-related observation error characteristics of AMSR2 ∆Tb. To improve the spatiotemporal variability of modeled soil moisture estimates, Soil Moisture Active Passive (SMAP) soil moisture retrievals were assimilated into Noah-MP. Assimilation was carried out using bias corrected (via CDF-matching) and raw (without CDF-matching) SMAP retrievals. Comparison against in-situ soil moisture measurements across the Tibetan Plateau highlighted the improvement in modeled soil moisture with reductions in mean bias and RMSE by 8.4% and 9.4%, respectively, even though assimilation occurred during <10% of the total study period across the Tibetan Plateau. More importantly, SMAP retrieval assimilation corrected biases that were generated due to unmodeled hydrologic phenomenon (i.e., surface irrigation associated with agricultural production). Improvements in soil moisture translated into changes in the modeled evapotranspiration. Further, the improvement in fine-scale (0.05 degree) modeled soil moisture estimates by assimilating coarse-scale soil moisture retrievals (36 km) indicated the potential of the described methodology for soil moisture estimation over data scarce regions. Soil moisture assimilation also increased the gridded total runoff (particularly baseflow) and volumetric streamflow across irrigated areas; however, limited impact was noted in terms of volumetric streamflow along high-flow river tributaries. In this study, data assimilation was leveraged to advance contemporary land surface modeling of the terrestrial water budget components across HMA. The study objectives explored how assimilation systems could be used to improve critical geophysical state estimation for a better informed future of regional water resources.
  • Thumbnail Image
    Item
    DATA-DRIVEN SIMULATIONS OF WILDFIRE SPREAD AT REGIONAL SCALES
    (2018) Zhang, Cong; Trouve, Arnaud; Mechanical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Current wildfire spread simulators lack the ability to provide accurate prediction of the active flame burning areas at regional scales due to two main challenges: a modeling challenge associated with providing accurate mathematical representations of the multi-physics multi-scale processes that induce the fire dynamics, and a data challenge associated with providing accurate estimates of the initial fire position and the physical parameters that are required by the fire spread models. A promising approach to overcome these limitations is data assimilation: data assimilation aims at integrating available observations into the fire spread simulator, while accounting for their respective uncertainties, in order to infer a more accurate estimate of the fire front position and to produce a more reliable forecast of the wildfire behavior. The main objective of the present study is to design and evaluate suitable algorithms for regional-scale wildfire spread simulations, which are able to properly handle the variations in wildfire spread due to the significant spatial heterogeneity in the model inputs and to the temporal changes in the wildfire behavior. First we developed a grid-based spatialized parameter estimation approach where the estimation targets are the spatially-varying input model parameters. Then we proposed an efficient and robust method to compute the discrepancy between the observed and simulated fire fronts, which is based on a front shape similarity measure inspired from image processing theory. The new method is demonstrated in the context of Luenberger observer-based state estimation strategy. Finally we developed a dual state-parameter estimation method where we estimate both model state and model parameters simultaneously in order to retrieve more accurate physical values of model parameters and achieve a better forecast performance in terms of fire front positions. All these efforts aim at designing algorithmic solutions to overcome the difficulties associated with spatially-varying environmental conditions and potentially complex fireline shapes and topologies. It paves the way towards real-time monitoring and forecasting of wildfire dynamics at regional scales.
  • No Thumbnail Available
    Item
    Estimating snow mass in North America through assimilation of AMSR-E brightness temperature observations using the Catchment land surface model and support vector machines
    (2018-04-16) Xue, Yuan; Forman, Barton; Reichle, Rolf; Forman, Barton
    To estimate snow mass across North America, multi-frequency brightness temperature (Tb) observations collected by the Advanced Microwave Scanning Radiometer - Earth Observing System (AMSR-E) from 2002 to 2011 were assimilated into the Catchment land surface model using a support vector machine (SVM) as the observation operator as part of a one-dimensional ensemble Kalman filter. The performance of the assimilation system is evaluated through comparisons against ground-based measurements and publicly-available reference SWE and snow depth products. Assimilation estimates agree better with ground-based snow depth measurements than model-only (“open loop”, or OL) estimates in approximately 82% (56 out of 62) of pixels that are colocated with at least two ground-based stations. In addition, assimilation estimates tend to agree better with all snow products over tundra snow, alpine snow, maritime snow, as well as sparsely-vegetated snow-covered pixels. Improvements in snow mass via assimilation translate into improvements in cumulative runoff estimates when compared against discharge measurements in 11 out of 13 major snow-dominated basins in Alaska. These results prove that a SVM can serve as an efficient and effective observation operator for snow mass estimation within a radiance assimilation system.
  • Thumbnail Image
    Item
    ASSIMILATION OF PASSIVE MICROWAVE BRIGHTNESS TEMPERATURES FOR SNOW WATER EQUIVALENT ESTIMATION USING THE NASA CATCHMENT LAND SURFACE MODEL AND MACHINE LEARNING ALGORITHMS IN NORTH AMERICA
    (2017) Xue, Yuan; Forman, Barton A.; Civil Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    Snow is a critical component in the global energy and hydrologic cycle. It is important to know the mass of snow because it serves as the dominant source of drinking water for more than one billion people worldwide. To accurately estimate the depth of snow and mass of water within a snow pack across regional or continental scales is a challenge, especially in the presence of dense vegetations since direct quantification of SWE is complicated by spatial and temporal variability. To overcome some of the limitations encountered by traditional SWE retrieval algorithms or radiative transfer-based snow emission models, this study explores the use of a well-trained support vector machine to merge an advanced land surface model within a variant of radiance emission (i.e., brightness temperature) assimilation experiments. In general, modest improvements in snow depth, and SWE predictability were witnessed as a result of the assimilation procedure over snow-covered terrain in North America when compared against available snow products as well as ground-based observations. These preliminary findings are encouraging and suggest the potential for global-scale snow estimation via the proposed assimilation procedure.
  • Thumbnail Image
    Item
    Data-Driven Wildfire Propagation Modeling with FARSITE-EnKF
    (2016) Theodori, Maria Faye; Trouve, Arnaud; Fire Protection Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)
    The goal of this study is to provide a framework for future researchers to understand and use the FARSITE wildfire-forecasting model with data assimilation. Current wildfire models lack the ability to provide accurate prediction of fire front position faster than real-time. When FARSITE is coupled with a recursive ensemble filter, the data assimilation forecast method improves. The scope includes an explanation of the standalone FARSITE application, technical details on FARSITE integration with a parallel program coupler called OpenPALM, and a model demonstration of the FARSITE-Ensemble Kalman Filter software using the FireFlux I experiment by Craig Clements. The results show that the fire front forecast is improved with the proposed data-driven methodology than with the standalone FARSITE model.