H23C-0972
A Probabilistic Model for Propagating Ungauged Basin Runoff Prediction Variability and Uncertainty Into Estuarine Water Quality Dynamics and Water Quality-Based Management Decisions
The latest official assessment of United States (US) surface water quality indicates that pathogens are a leading cause of coastal shoreline water quality standard violations. Rainfall-runoff and hydrodynamic water quality models are commonly used to predict fecal indicator bacteria (FIB) concentrations in these waters and to subsequently identify climate change, land use, and pollutant mitigation scenarios which might improve water quality and lead to reinstatement of a designated use. While decay, settling, and other loss kinetics dominate FIB fate and transport in freshwater systems, previous authors identify tidal advection as a dominant fate and transport process in coastal estuaries. As a result, acknowledging hydrodynamic model input (e.g. watershed runoff) variability and parameter (e.g tidal dynamics parameter) uncertainty is critical to building a robust coastal water quality model. Despite the widespread application of watershed models (and associated model calibration procedures), we find model inputs and parameters are commonly encoded as deterministic point estimates (as opposed to random variables), an approach which effectively ignores potential sources of variability and uncertainty. Here, we present an innovative approach to building, calibrating, and propagating uncertainty and variability through a coupled data-based mechanistic (DBM) rainfall-runoff and tidal prism water quality model. While we apply the model to an ungauged tributary of the Newport River Estuary (one of many currently impaired shellfish harvesting waters in Eastern North Carolina), our model can be used to evaluate water quality restoration scenarios for coastal waters with a wide range of designated uses. We begin by calibrating the DBM rainfall-runoff model, as implemented in the IHACRES software package, using a regionalized calibration approach. We then encode parameter estimates as random variables (in the rainfall-runoff component of our comprehensive model) via the probabilistic modeling software program Analytica. This approach not only reflects uncertainty in parameter estimates but, by modeling the predicted daily runoff rate as a random variable, propagates that variability into the tidal prism model as well. The tidal prism model has the advantage of having only one hydrodynamic calibration parameter, the tidal exchange ratio (the ratio between the volume of water returning to an estuary on an incoming tide and the volume of water which exited the estuary on the previous outgoing tide). We estimate the tidal exchange ratio by calibrating the tidal prism model to salinity data using a Bayesian Markov chain Monte Carlo (MCMC) procedure and, as with other parameters, encode it as a random variable in the comprehensive model. We compare our results to those of a purely deterministic model, and find that intrinsic sources of variability in ungauged basin runoff predictions, when ignored, lead to pollutant concentration forecasts with unnecessarily large prediction intervals, and to potentially over-conservative management decisions. By demonstrating an innovative approach to capturing and explicitly acknowledging uncertainty in runoff model parameter estimates, our modeling approach serves as an ideal building block for future comprehensive model-based pollutant mitigation planning efforts in ungauged coastal watersheds, including those implemented through the US Environmental Protection Agency total maximum daily load program.
H23C-0973
Development and assessment of a 59-year (1948-2006) global 0.5-degree near-surface atmospheric data
Daily precipitation, snowfall and specific humidity and 3-hourly temperature, shortwave radiation and longwave radiation data were developed for 59-years (1948-2006) with 0.5-degree resolution, which can be used to drive land surface models and global hydrological models; these data were created using parameters obtained from daily observations that are available in recent years. Global terrestrial snowfall was estimated by applying gauge undercatch correction for snowfall and rainfall based on daily meteorological data and gauge type. The estimated 59-year time series of snowfall amount shows a downward trend after the mid- 1980s that may be due to a warmer climate. One of the advantages of this data set is that the statistical characters of the created variables are more similar to observation than those of reanalysis data (e.g., precipitation, temperature and radiation). Other advantages are the availability of data for recent years and the expectation of future extensions. Because we estimated undercatch correction using daily atmospheric data, the estimated snowfall allows to determine long-term variations of snowfall more reliably than snowfall estimates that are derived using a climatology of correction factors. Together with its relatively high resolution (0.5-degree) and these advantages, the newly obtained data may be preferred to other forcing data sets in case of hydrological and climate change studies, in particular if the study results are sensitive to daily variations in atmospheric conditions.
H23C-0974
Internal validation of conceptual rainfall-runoff models using baseflow records
An important tool in the management of floods is the use of rainfall-runoff models to predict the arrival of discharge peaks. These models generally use rainfall and potential evapotranspiration rates as input, and relate these to the catchment discharge through a number of conceptual equations. The parameters of these equations are estimated through a comparison of the modeled discharge to the observations. Only one variable, catchment discharge, is thus generally used to calibrate and validate these models. The objective of this paper is to validate the internal model dynamics of two widely used rainfall-runoff models using baseflow records. These models, the Hydrologiska Byrans Vattenbalansavdelning (HBV) model and the Probability Distributed Model (PDM), were calibrated using one year of hourly discharge data. Two different parameter estimation algorithms were used for model calibration. The Shuffled Complex Evolution (SCE-UA) algorithm minimizes the Root Mean Square Error between the model simulations and the observations, while Multistart Weight-Adaptive Recursive Parameter Estimation (MWARPE) uses the Extended Kalman Filter equations in an iterative, Monte-Carlo framework. An assessment of the model performance and impact of the calibration algorithms with respect to both baseflow and total discharge will be made.
H23C-0975
Capacity Building for Disaster Management in Vulnerable Regions of Africa: Implementing an Operational Flood Warming System in Lake Victoria
NASA Applied Science program has partnered with USAID and The Regional Centre for Mapping of Resources for Development (RCMRD) in Africa to implement an operational flood warning system for East Africa, SERVIR-Africa project. The project seeks to take advantage of remote sensing information as an alternative and supplemental to ground-based observation in order to preserve the spatial extent of flood hazards. The recently available and virtually uninterrupted supply of satellite-based rainfall estimates is increasingly becoming a cost-effective data source for flood prediction in many under-gauged regions around the world. Our initial focus aims to provide an operational flood warning system for Lake Victoria, a flood-prone region home to 30 million people. The key datasets enabling the development of a distributed hydrological model in Africa include TRMM-based Multi-satellite Precipitation Analysis (TMPA), digital elevation data from the Shuttle Radar Topography Mission (SRTM) mission, HydroSHEDS hydrological products, MODIS Land cover, and soil parameters provided by FAO. This research focuses on evaluation and integration the TMPA Real- Time product into an online operational flood prediction system. We will also identify the optimal calibration strategy for satellite rainfall data into real-time hydrological modeling, one current knowledge gap that has remained relatively unexplored. Early results demonstrate this flood modeling system is useful decision- support tool for governmental officials and international aid organizations to better quantify flood impacts and extent of hazard risk, as well as more expediently respond to flood emergencies.
H23C-0976
Applicability of Partially Observed Runoff Data in Parameter Calibration
Conceptual water balance models are typically calibrated using continuously observed runoff data of a basin. Most African river basins, however, have limited hydrologic data especially runoff data. When selecting the years of continuously observed runoff data only, the length of calibration data may be significantly reduced. To avoid the loss of valuable hydrologic data embedded in partially observed streamflow data, this study proposes to use these partially observed runoff data in parameter calibration instead of using continuously observed data only. The proposed methodology is applied to the upper Blue Nile River Basin of Ethiopia where data are limited and hydrologic importance is high to the riparian Nile Basin system. This study evaluates the value of information of using partially observed hydrographs by comparing calibration and verification efficiencies obtained using an existing water balance model, and continuous and random calibration data sets. Continuous calibration data sets are sampled using 1 to 25 years of moving windows from the 30 year observed hydrograph while the corresponding lengths of random calibration data sets are randomly sampled from the same hydrograph. The results reveal that it requires at least 10 years (120 months) of continuous calibration data and 36 months of random calibration data for robust parameter calibration with low parameter uncertainty. It is likely that random calibration data need a shorter period to include various runoff dynamics of the basin than continuous calibration data. It is also found that parameter calibration becomes robust when including more than 30% of high flows. The proposed methodology and findings of the study are essential to estimate runoff of sparsely monitored sub-basins and to develop a hydrologic monitoring strategy for the upper Blue Nile River Basin.
H23C-0977
Combination of Radar Altimeter and In-Situ Measurements to deduce Rating-Curves at Some Virtual Stations in the Ungauged Amazon and Orinoco Basins
In the last two years, virtual gauged stations have been proposed to increase the density of hydrological network in ungauged or very poorly monitored basins (Leon, 2006). In spatial hydrology a virtual station is considered as any crossing of water body surface (i.e., large rivers) by radar altimeter satellite tracks. The main objective of this study is to review the usefulness of altimetric data presenting rating curves obtained for some virtual stations at the poorly gauged basins of Caqueta (Colombian Amazon basin), Uaupes and Upper Negro (Brazilian Amazon basin) and Upper Orinoco. Rating curve parameters at virtual stations are estimated by fitting with a power law distribution the temporal series of water surface altitude derived from ENVISAT satellite measurements and modeled discharges. The applied methodology (Leon et al. 2006a) allows the ellipsoidal height of effective zero flow to be estimated. This parameter is a good proxy of the mean water depth from which the river bed slope can be computed. These quantities combined with rating-curve parameters are highly valuable for understanding hydrological behaviour, especially at ungauged basins where hydrodynamical studies had always been prevented by the lack of in-situ data. The results obtained allow to propose a new insight into the hydrological behaviour of the region shared by Colombia, Brazil and Venezuela, which is very difficult to access, and then very poorly known.
H23C-0978
Simulating hydrological impacts of climate and land use changes in the North Saskatchewan River Watershed, Alberta, using the ACRU agro-hydrological modelling system
Throughout the province of Alberta there is increasing demand for water due to population growth, and an increasing demand from agriculture and industry. In contrast, availability of future water resources is uncertain due to climate change, and future land use practices. The upper North Saskatchewan River (UNSR) watershed is situated south-west of Edmonton, with a watershed area of slightly over 28,000km2. This on-going research looks to model the UNSR watershed to help predict future stream flows in the UNSR, based on potential future climate change and land use changes within the watershed, by setting up the ACRU agro-hydrological modeling system (Schulze et al., 2004). Contour, stream, lake, and elevation point data were collected and processed to created a DEM of the study area. From the DEM, sub watersheds were created from flow gauging stations to later run the model and verify results on smaller portions of the watershed. The DEM was then used to interpolate solar radiation data, slope and aspect data for the study area. Soil, land cover, climate, and stream data were all collected and processed using ArcGIS to create hydrological response units (HRUs). Each HRU will be set up individually, so that a minimum daily time series of 30 years is available to simulate all elements of the hydrological cycle for each of the HRUs, including snow pack development and snow melt, actual evaporation, transpiration, soil moisture storage, storm flow and groundwater contributions. Once all the spatial parameters have been processed in the GIS and inputted into ACRU, the data will be simulated and outputs verified against streamflow observations from gauging stations within the UNSR watershed. This will be done to generate confidence in the various scenarios of hydrological impacts within the UNSR watershed. Verification analysis is important to determine if outputs from the model are consistent with the behaviour of the hydrological system. The verification analyses are based on comprehensive statistical analyses. Once the model has been determined to be within physically meaningful ranges, different climatic and land use scenarios will be integrated into the model to simulate the impacts of expected environmental changes, such as a decreased snow pack, or land use change, such as forest fires. Results of risk analyses of available water in rivers and reservoirs will provide critical information for water managers, particularly those in the energy industry. References: Schulze, R.E.; Lorentz, S.; Kienzle, S.W.; Perks, L. 2004: Modelling the impacts of land-use and climate change on hydrological responses in the mixed underdeveloped / developed Mgeni catchment, South Africa. In: Kabat, P. et al. (Eds.): Vegetation, Water, Humans and the Climate A New Perspective on an Interactive System. BAHC-IGBP Publication, Springer, 17pp, with 14 Figures and 2 tables
H23C-0979
Spatio-Temporal Extrapolation of Unit Hydrograph Parameters Based on Catchment and Event Extrapolators for Predictions in Ungauged Basins
The scarcity of observed discharge limits the applicability of straight forward calibration of a hydrological model. Regionalization, extrapolation of model parameters from gauged catchments to hydrologic-similar ungauged catchments, is widely recognized as a potential approach to tackle the problem of predictions in ungauged catchments. But two major concerns regarding the currently established regionalization techniques have been pointed out. Firstly the problem of equifinality, due to non-uniqueness of the parameter sets that lead to a reasonable performance of a model the sub-sequent regional function fitting the parameters and regional extrapolators cannot be considered appropriate. Secondly, the hydrologic regimes in most of the inhabited regions are going through changes due to anthropogenic as well as natural influences, which bring the reliability of a hydrological model calibrated against past observed discharge under skepticism. The following investigation attempts to address the two concerns. During this study a semi-distributed Nash cascade unit hydrograph model was implemented to derive flood runoff hydrograph, regionalization was carried out for the Nash cascade parameters, number of reservoirs (N) and reservoir constant (K). For the regionalization the regional transfer functions for K and the coefficient (β) of the inter-parameter function were calibrated through a regional optimization procedure by using a regional objective function. To consider the influence of temporal changes, in addition to spatial extrapolators, time or event dependent extrapolators such as landuse, rainfall intensity and duration were also incorporated within the transfer functions. The investigation was carried out using a database of 210 rainfall-runoff events from 41 mesoscale catchments located in south-west Germany. The outcome of the investigations underlines the advantage of regionalizing inter-parameters relationship and the approach of regional optimization to tackle equifinality. Additionally it also sheds light on the sensitivity of the model parameters to the event dependent extrapolators.
H23C-0980
Development of a Single Parameter Variable-Source-Area Model of Stream Flow Generation in Ungauged Basins
This research developed a relation between distributed saturation deficit, highest saturation deficit and the threshold area of saturation with an exponential scaling formulation of the soil-moisture. With this formulation, a hydrological overland forecasting model is presented that attempts to combine the distributed soil-moisture deficit and dynamic contributing area with the advantage of the use of a single parameter that has the possibility of direct measurement, hence independent of the topographic index distribution and its scale effects. The possibility of direct measurement of the single parameter makes the model potentially applicable for predicting saturated excess runoff in ungauged catchments. This will help avoid regionalization and parameter transferability problems. The model is applied in test catchments and compared with TOPMODEL.
H23C-0981
Multiscale Parameter Regionalization of a Grid-based Hydrologic Model
Integrated water resources planning and management at the mesoscale requires, among other things of a parsimonious and distributed hydrologic model able to reproduce not only the discharge hydrograph at any gauged or ungauged location but also the spatio-temporal distribution of state variables such as soil moisture. Furthermore, this model should be able to take into account changes in land cover and climate as well as management practices. The state-of-the-art with respect to these issues, however, is not yet satisfactory. More specifically, existing models suffer from overparametrization; the lack of an effective technique to integrate the spatial heterogeneity of soils, vegetation, and topography; the non-transferability of model parameters to ungauged basins; and a considerably large execution time. The main goal of this study is to present and validate a multiscale regionalization technique integrated into a grid-based mesoscale hydrologic model (mHM) aiming to address these issues simultaneously. mHM is based on accepted hydrological conceptualizations and require three levels of spatial information: level-2 for the climatic information, level-1 for the state variables of the model, and level-0 for physiographic input data such as soil textures, land cover, elevation, and geological formations. Their spatial resolution varies from (1000-5000)m, (500-1000)m, and (50-100)m respectively. Model parameters at level-1 are location and time dependent. They are estimated through upscaling operators that link level-0 information with global transfer-function parameters, which in turn are found through optimization. The functional relationships constituting these operators are based on process understanding or empirical evidence. Results obtained for 34 basins located in Germany indicated that this regionalization technique contributed not only to reduce significatively the number of free parameters but also to ensure their transferability to ungauged basins. The Nash-Sutcliffe Efficiency (NSE) of mHM was on average 6% greater than that obtained with the standard HBV model but required 85% less effective parameters to calibrate. The HBV model was regionalized with the standard homogeneous response unit (HRU) concept. Moreover, uncertainty and leave-on-out crossvalidation tests showed that this technique produced acceptable streamflow predictions in target basins which were assumed ungauged during crossvalidation. The NSE in the calibration and crossvalidation phases ranged between 0.70 to 0.85 and between 0.55 to 0.75 respectively. Additionally, soil moisture patterns compared well against proxies derived from daily MODIS images (NASA).
H23C-0982
MODIS Vegetation Metrics as Indicators of Hydrological Response in Watersheds of California Mediterranean-type Climate Zones
Vegetation characteristics of a watershed can be important in determining hydrological response variables (HRVs) such as streamflow (Q), evapotranspiration (ET), and river yield (Q/P). Quantifying the relationship between satellite-derived vegetation metrics and hydrological response to precipitation (P) has the potential to aid in the prediction of streamflow and evapotranspiration for ungauged watersheds. The utility of Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data to estimate HRVs of watersheds at the regional scale (southern and central California) is tested in this study. An exhaustive statistical regression analysis was conducted, to quantify the relationship between MODIS vegetation metrics and HRVs. Both ordinary least squares and spatially varying parameter models were tested. Moderate linear relationships were found between several MODIS-derived vegetation metrics and both ET and Q. MODIS Leaf Area Index (LAI) regressed on ET exhibited the strongest relationship regression coefficients of 0.50 and 0.95 for the study period mean and strongest annual mean, respectively. Results indicate that the inclusion of spatial-varying parameters can improve the fit in the relationship between MODIS vegetation metrics and HRVs relative to that of the traditional fixed parameter model. The use of a relatively novel exhaustive statistical approach to testing multiple watersheds with varying land cover and climatic characteristics shows that drought conditions in the southern watersheds (2002 and 2004) yielded the strongest fits between MODIS vegetation metrics and HRVs. Alternatively, the relationship between MODIS vegetation metrics and HRVs was weaker when sampling watersheds from similar climatic zones (i.e. humid or semi-arid). Results suggest that direct region-wide estimates of annual streamflow and evapotranspiration for ungauged watersheds can be made reliably for drier years, and that MODIS-derived vegetation metrics may be useful for aiding identification of hydrologically similar watersheds for model-based regionalization purposes.
H23C-0983
Assessment of Terrestrial Water Storage Dynamics from Daily to Interannual Timescales Using Combined Atmosphere-land Water Balance Computation
Most of previous observation-based estimates on regional terrestrial water storage (TWS) from water balance computation or GRACE satellite data have focused on only long-term climatology or monthly- seasonal timescales. Here, regional TWS is estimated by using combining atmospheric-land water balance computation at daily, monthly, seasonal and interannual timescales for the period of 1980-2006. The result is compared with direct observations in Illinois (soil moisture, groundwater and snow), land surface model simulations and GRACE. The motivation in this study is to provide quantitative assessment on the dynamics controlling the propagation of hydroclimatic anomalies across the atmospheric and terrestrial branches of regional hydrology from daily to interannual timescales. In particular, two critical issues related to TWS are examined: (1) How does TWS changes balance with atmospheric vapor convergence and runoff at different timescales? (2) Are there identifiable relationships between regional TWS and basin discharges? if yes, at which timescales? The findings from this study are anticipated to enhance understanding on hydrological predictions in ungauge basins (PUB) and improve land surface model parameterizations.