N. Kouwen

Towards closing the vertical water balance in Canadian atmospheric models: Coupling of the land surface scheme class with the distributed hydrological model watflood

Abstract Second generation land surface schemes are the subject of much development activity among atmospheric modellers. This work is aimed at, among other things, improving the representation of the soil water balance in order to simulate, more properly, exchanges with the atmosphere and to permit the use of model output to generate streamflow for model validation. The Canadian development program is centred on CLASS, the Canadian Land Surface Scheme, developed at Environment Canada. This paper focuses on the improvement of hydrology in CLASS. This was accomplished by designing a two‐way interface to WATFLOOD, a distributed hydrologic model developed at the University of Waterloo. The two models share many features, which facilitated the coupling procedure. The interface retains the three‐layer vertical moisture budget representation in CLASS but adds three horizontal runoff possibilities. Runoff from the surface water follows Mannings equation for overland flow. Interflow is generated from the near‐surface soil layer using a parametrization of Richards equation and base flow is produced by Darcian flow from the bottom of layer 3. An approximation of the internal topography of grid elements is used to supply horizontal gradients for the runoff components. Tests are in progress in four Canadian study areas. Initial results are presented for the summer of 1993 for the Saugeen River in southwestern Ontario. The new scheme produces realistic hydrographs, whereas the old scheme did not. Bare ground evaporation is reduced by about 17% as a consequence of reduced water availability in layer 1. Evapotranspiration is not affected because the rooting depth extends into layer 3, in which soil moisture does not change appreciably with the new scheme. These results suggest that the new scheme improves the representation of streamflow in WATFLOOD/CLASS and of the soil moisture budget in CLASS. Work is in progress to validate this result over basins, such as the BOREAS study watersheds, where both runoff and evapotranspiration measurements are available.

Watershed modeling using land classifications

A comparison was made between using a lumped hydrological model and using a version of the same model applied successively to different land uses within subbasins. A watershed in the Rocky Mountains of British Columbia was divided into three contributing subbasins, and each of these was further subdivided by land cover classification using Landsat images. A hydrological model was applied separately to each land cover class in each subbasin, and the resulting hydrographs were routed to the subbasin outlet and then through lower subbasins. The final hydrographs were compared to those obtained using the model on the basin as a whole. It was found that using a semidistributed model gives goodness of fit statistics that are better than the lumped basin approach. The land class dependent parameter values found through optimization confirm the physical variations in storages and infiltration rates that would be expected in a mountain basin. The advantage of the semidistributed model is that relating the parameter values to land cover characteristics provides a method of investigating land use changes and allows the model to be more easily transferred to other basins.

A Land Cover-Based Snow Cover Representation for Distributed Hydrologic Models

A snow cover depletion curve (SDC) summarizes the relationship between snow cover distribution and an average snow cover property, such as depth or water equivalent, for a given area. Snow cover depletion curves have been developed for, and applied in, hydrological models on a watershed or elevation zone basis. However, land cover-;based SDCs are not prominent in the literature. For this study the areal distribution of snow cover for dominant land cover units was measured during the winters of 1991 and 1992 in the Laurel Creek watershed in southern Ontario, Canada. On the basis of these data a general model for land cover-;based SDCs is developed for these land cover units, namely, short grass, ploughed fields, and deciduous forests. This model is derived from the three-parameter lognormal distribution, which is shown to characterize the areal depletion curves of the land cover units studied. The SDCs based on this new model provide a formal distributed snow cover representation that can be used in vegetation-based distributed hydrological models requiring accurate spatial representations of snow cover attributes.

On the Use of Coupled Atmospheric and Hydrologic Models at Regional Scale

Abstract. The purpose of this study is to present the possibilities offered by coupled atmospheric and hydrologic models as a new tool to validate and interpret results produced by atmospheric models. The advantages offered by streamflow observations are different from those offered by conventional precipitation observations. The dependence between basins and sub-basins can be very useful, and the integrating effect of the large basins facilitates the evaluation of stateof-the-art atmospheric models by filtering out some of the spatial and temporal variability that complicate the point-by-point verifications that are more commonly used. The streamflow predicted by the coupled atmospheric-hydrologic model versus the measured streamflow is sufficiently sensitive to clearly assess atmospheric model improvements resulting from increasing horizontal resolution and altering the treatment of precipitation processes in the model. A case study for several southern Ontario river basins is presented with the Watflood hydrologic model developed at the University of Waterloo. It is passively coupled to a nonhydrostatic mesoscale atmospheric model (mc2) that is integrated 318 HIGH PERFORMANCE COMPUTING SYSTEMS AND APPLICATIONS at horizontal resolutions of 35, 10 and 3 km. The Watflood model is also driven by radar derived precipitation amounts from King City Radar observations. It is demonstrated that the hydrological model is sufficiently sensitive and accurate to diagnose model and radar errors. This tool brings an additional degree of verification that will be very important in the improvement of technologies associated with atmospheric models, radar observations and the water resources management.

Application of a grouped response unit hydrological model to a northern Wetland region

Applications of hydrological models to northern wetland-dominated regions have been limited in the past to a few case studies on small basins employing lumped models. Only recently have there been attempts to apply the grouped response unit (GRU) distributed modelling approach using terrain classifications to these same basins. This study summarizes recent efforts in applying such a model. For the purposes of implementing the GRU approach, terrain types that are hydrologically significant and characteristic to the wetland-dominated regime were successfully discriminated using a principal component analysis and a hybrid unsupervised/supervised classification technique on Landsat-Thematic Mapper imagery. The terrain classifications were then used as input into a distributed hydrological model for calibration and validation using recorded spring runoff events. Preliminary model applications and results are described. Calibration to a historic spring runoff event yielded an γ 2 value of 0.86. Model validation, however, yielded much poorer results. The problems of model applicability to this region and limitations of sparse data networks are highlighted. The need for more field research in this type of hydrological regime, and associated improvements to the model parameter set are also identified.

MODELING THE RAINFALL-RUNOFF RESPONSE OF A HEADWATER WETLAND

In the eastern temperate region of North America, treed headwater swamps are a familiar watershed feature. These low-gradient wetlands commonly exist at groundwater discharge sites and represent a link between the underlying groundwater system and the surface drainage system. In contrast to the extensive literature pertaining to the hydrologic modeling of agricultural and forest land classes, little attention has been focused on the development and testing of numerical simulation models for predicting the hourly stormflow response from headwater wetland sites. If required to predict the rate of outflow from a wetland-dominated catchment, the hydrologist or engineer has few numerical tools and little data available to assist in the prediction. The objective of this research was to investigate the feasibility of applying a numerical model to simulate the rainfall-runoff response from a treed headwater wetland site. The wetland model utilizes a hydrology model coupled to a hydraulic stream-routing model. A depth-averaged laminar flow model is used to simulate the horizontal movement of stormwater both through and over the wetland sediments. The development and testing of the wetland model were completed in conjunction with a data collection program in which hydrometric and meteorologic data were obtained at a 400-ha first-order headwater swamp located within the Teeswater River watershed in southern Ontario, Canada. An analysis revealed that the simulated wetland streamflows were sensitive to the antecedent saturation of the wetland sediments, the storage and flow transport characteristics of the wetland sediments, and the conveyance capabilities of the wetland channel system.

Application of indexed snowmelt algorithms in a northern wetland regime

Applications of snowmelt models in Canadas north have been limited largely to energy balance models concentrating on micro scale studies or macroscale applications. The latter rely solely on basin-wide, optimized parameters for hydrological simulation and often neglect some of the major physical processes controlling melt production. Although physically realistic models can be implemented at the micrometeorological scale, the data requirements are often too numerous to make these types of models practical to apply at the meso- or macroscales. On the other hand, the standard lumped model approach often oversimplifies the physical processes and fails to reveal subtle differences between land cover types and their specific response to meteorological inputs. This paper focuses on the use of indexed snowmelt algorithms derived for individual land cover component characteristics of the wetland dominated region of the lower Liard River Valley. NWT, Canada. These algorithms use an hourly temperature index and a combination radiation-temperature index approach to estimate snowmelt within the different land cover types. The algorithms developed are incorporated into a fully distributed hydrological model (SPL7) that uses the grouped response unit (GRU) method for basin discretization. Snowmelt indices are estimated for both approaches using snow cover depletion data obtained during an extensive field campaign. The indices are then validated using historical data from complementary studies. Results show that the radiation-temperature algorithm provided slightly improved calibration results: however, both algorithms validated equally well.

The specific surface area of fresh dendritic snow crystals

The surface area to mass ratio or specific surface area (SSA) is an often neglected characteristic of the snowpack that varies substantially with time, and with the shape of the individual snow crystal for fresh snow. The SSA for the dendritic shape of snow crystals was computed using a series of images photographed by W. A. Bentley. The specific images were dendritic crystals (P1d, P1e, P1f) and crystals that take a partial dendritic form and have ends or extensions (P2a, P2b, P2d, P2e, P2f, P2g) according to the Magono and Lee snow crystal classification. Image analysis, using known geometric relationships between length and width, and particle size distributions, examined the spatial properties of 50 sample snow crystals. Probability distribution functions were derived for SSA and these compared well with measured and other computed estimates of fresh snow SSA. For the non-rimed condition, the average SSA was 0.182 m 2 /g with a range from 0.09 to 0.33 m 2 /g. The presence of rime is discussed, and depending on the shape of the rime particles and the degree of surface coverage, the SSA can be doubled (20% coverage for needle or plate shaped rime). Fractal analysis was performed to determine various geometric relationships that characterize the dendritic form of snow crystal.

Soil Moisture in Pasture Fields Using ERS-1 SAR Data: Preliminary Results

RESUMELinformation sur lhumidite du sol est utilisee dans letablissement de modeles hydrologiques physiographiques distribues spatialement en vue de determiner les niveaux dinfiltration et devapotranspiration. De meme, cette information est utile dans le domaine de lagriculture. Les images obtenues par teledetection dans les hyperfrequences actives constituent une source relativement nouvelle de donnees susceptibles de permettre aux hydrologues et aux pedologues de cartographier avec precision lhumidite du sol dans les bassins-versants. Dans la documentation scientifique, la plupart des etudes portent sur les sols denudes et les sols cultives. Dans la presente etude, les auteurs se sont interesses aux pâturages. On y presente les resultats preliminaires dune correlation entre les mesures dhumidite du sol realisees dans des pâturages et la retrodiffusion radar sur des images ERS-1 (bande C, polarisation VV) dun secteur dune superficie de 1500 km2 du bassin-versant de Grand River, en Ontario. Une...

Surface temperature adjustments to improve weather radar representation of multi-temporal winter precipitation accumulations

Hydrologists and water resources managers who work in areas that receive a significant portion of the annual precipitation in the form of snowfall rely on good approximates of snow accumulation in order to assess snowpack volumes for snowmelt streamflow estimation. Weather radar rainfall estimation has been used for hydrological modelling and radar has been used for the estimation of snowfall from individual events, yet radar has rarely been used to measure snowfall accumulation over time periods longer. Snowfall estimates for weekly, monthly, and seasonal accumulation periods have been compared to measured Nipher-shielded Belfort precipitation gauge quantities. A local scaling issue that caused overestimates is discussed. To enhance the accumulation estimates, the conventional scan radar images were adjusted using the near surface air temperatures. The adjustment for mixed precipitation improved the accumulation estimates, while the subsequent particle shape adjustment for snow crystal shape did not further enhance the radar estimates.