During recent years a hybrid model has been set up for the operational forecasting of flood discharges in the 6750km ² Tyrolean part of the River Inn catchment in Austria. The catchment can be characterized as a typical alpine area with large variations in altitude. The paper is focused on the error analysis of discharge forecasts of four main tributary catchments simulated with hydrological water balance models. The selected catchments cover an area of 2230km ², where the non-glaciated and glaciated parts are modeled using the semi-distributed HQsim and the distributed model SES, respectively.The forecast errors are evaluated as a function of forecast lead time and forecasted discharge magnitude using 14 events from 2007 to 2010. The observed and forecasted precipitation inputs were obtained under operational conditions. The mean relative bias of the forecasted discharges revealed to be constant with regard to the forecast lead time, varying between 0.2 and 0.25 for the different catchments. The errors as a function of the forecasted discharge magnitude showed large errors at lower values of the forecast hydrographs, where errors decreased significantly at larger discharges being relevant in flood forecasting

Stormwater biofilters have the ability to remove nutrients from stormwater. Reliable pollutant removal during the cold season is particularly important due to the comparably high contamination levels. However, the removal performance might be negatively affected by low temperatures. A biofilter column study was conducted in thermostat-controlled climate rooms (at 2, 7 and 20 °C) to investigate the effect of low temperatures on nutrient removal. Phosphorus and suspended solids removal were significantly correlated and consistently very high (typically in excess of 90 and 95%, respectively, at all temperatures). This is important for the successful implementation of biofilters in cold climates since phosphorus is commonly of principal concern, often being the limiting factor for eutrophication in freshwater ecosystems. Unfortunately, nitrogen removal was poor and leaching was shown, which increased with temperature. The increasing nitrate-nitrogen production rates with temperature were well described by the Arrhenius relationship with temperature coefficients Q10 in the range which is typically used to describe temperature effects on nitrification. Thus, temperature effects have to be considered when nitrogen removal is targeted and the biofilter might be exposed to cold temperatures.

SummaryFour indicator bacteria were measured in association with physico-chemical constituents and selected inorganics during rainfall, baseflow and snowmelt periods in storm sewers of four urban catchments in a northern Swedish city. The variation patterns of coliforms, E. coli, enterococci and C. perfringens concentrations were assessed in manually collected grab samples together with those of phosphorus, nitrogen, solids, and readings of pH, turbidity, water conductivity, temperature and flow rates to examine whether these constituents with variation patterns similar to those of indicator bacteria, and to exclude the constituents with less similarity. In the reduced data set, the similarities were quantified by the clustering correlation analysis. Finally, the positive/negative relationships found between indicator bacteria and the identified associated constituent groups were described by multilinear regressions. In the order of decreasing concentrations, coliforms, E. coli and enterococci were found in the highest mean concentrations during both rainfall and snowmelt generated runoff. Compared to dry weather baseflow, concentrations of these three indicators in stormwater were 10 (snowmelt runoff) to 102 (rain runoff) times higher. C. perfringens mean concentrations were practically constant regardless of the season and catchment. The type and number of variables associated with bacteria depended on the degree of catchment development and the inherent complexity of bacteria sources. The list of variables associated with bacteria included the flow rate, solids with associated inorganics (Fe and Al) and phosphorus, indicating similar sources of constituents regardless of the season. On the other hand, bacteria were associated with water temperature only during rain periods, and somewhat important associations of bacteria with nitrogen and pH were found during the periods of snowmelt. Most of the associated constituents were positively correlated with bacteria responses, but conductivity, with two associated inorganics (Si and Sr), was mostly negatively correlated in all the catchments. Although the study findings do not indicate any distinct surrogates to indicator bacteria, the inclusion of the above identified constituents (flow rate, solids and total phosphorus for all seasons, water temperature for rainfall runoff, and total nitrogen and pH for snowmelt only) in sanitary surveys of northern climate urban catchments would provide additional insight into indicator bacteria sources and their modelling.

Grass swales are important elements of the urban green infrastructure that convey and attenuate urban runoff and improve its quality mostly through stormwater infiltration into, and retention of conveyed pollutants by, swale soils. The retention of metals by grass swales was addressed in this study investigating the enrichment of swale soils by three common traffic-related metals: Cu, Pb and Zn. Three swales of various characteristics (L1, L2, L3) were selected for study and their soils were sampled by coring the top 30 cm and dividing the cores into 5 cm thick layers. Cumulative metal burdens were compared to those modelled by the proprietary StormTac Web model, which estimates annual loads of specific constituents for the given land uses and stormwater treatment. The comparisons of measured (MBm) and simulated (MBs) metal burdens retained by swales showed that the measured values exceed the simulated ones, as described by average ratios MBs/MBm = 0.64, 0.50 and 0.59, for swales L1, L2 and L3, respectively. The measured burdens were calculated after subtracting the native soil metal concentrations, assumed equal to those found in the deepest sampled layer, 25–30 cm below the surface. The results suggest the feasibility of assessing performance of grass swales by modelling metal (Cu, Pb, Zn) retention by swales, however for older facilities considered for rehabilitation, the simulated results should be supplemented by soil chemistry sampling.

To improve the understanding of interactions between the boreal forest and the climate system as a key issue for global climate change, the water budget of a mixed pine and spruce forest in central Sweden was estimated by measurements of the water flux components and the total evaporation flux during the period 16 May-31 October 1995. Total evaporation was measured using eddy correlation and the components were obtained using measurements of precipitation, throughfall, tree transpiration, and forest floor evaporation. On a daily basis, tree transpiration was the dominant evaporation component during the vegetation period. However, it could be efficiently blocked by a wet canopy associated with large interception evaporation. The accumulated total evaporation was 399 mm, transpiration was 243 mm, forest floor evaporation was 56 mm and interception evaporation was 74 mm. The accumulated sum of interception, transpiration, and floor evaporation was 51 mm larger than the actual measured total evaporation. This difference was mainly attributed to the fact that transpiration was measured in a rather dense 50-year-old stand while total evaporation represented the average conditions of older, roughly 100-year-old stands. To compare eddy-correlation measurements with small-scale measurements of evaporation components, a source area analysis was made to select the flux data that give the best representation of the investigated stand. Especially under stable atmospheric conditions the requirements for surface homogeneity were very high and extreme care had to be taken to be aware of the flux source areas. Canopy water storage was determined by two methods: by the water balance of the canopy, which gave a result of 3.3 mm; and by the so-called minimum method based on plots of throughfall versus precipitation, which gave a much lower value of 1.5 mm. Seasonal interception evaporation constituted 30% of the precipitation.

Suspended sediment is of importance in river and dam engineering. While, due to its high nonlinearity and stochasticity, sediment prediction by conventional methods is a challenging task. Consequently, this paper establishes a new hybrid model for an improved forecast of suspended sediment concentration (SSC). It is a nonlinear autoregressive network with exogenous inputs (NARX) integrated with a data pre-processing framework (denoted as INARX). In this model, wavelet transformation (WT) is used for time series decomposition and multigene genetic programing (MGGP) for details scaling. The two incorporated modules improve time and frequency domain analysis, allowing the network to unveil the embedded characteristics and capture its non-stationarity. At a hydrological station on the upper reaches of the Yangtze River, the records of daily water stage, flow discharge and suspended sediment are collected and refer to a nine-year period during 2004-2012. The data are used to evaluate the models. Several wavelets are explored, showing that the Coif3 leads to the most accurate prediction. Compared to the sediment rating curve (SRC), the conventional MGGP, multilayer perceptron neural network (MLPNN) and NARX, the INARX demonstrates the best forecast performance. Its mean coefficient of determination (CD) increases by 7.7%-38.6% and the root mean squared error (RMSE) reduces by 15.1%-54.5%. The INARX with the Coif3 wavelet is further evaluated for flood events and multistep forecast. Under flood conditions, the model generates satisfactory results, with CD > 0.83 and 84.7% of the simulated data falling within the ±0.1 kg/m3 error. For the multistep forecast, at a one-week lead time, the network also yields predictions with acceptable accuracy (mean CD = 0.78). The model performance deteriorates if the lead time becomes larger. The established framework is robust and reliable for real-time and multistep SSC forecast and provides reference for time series modeling, e.g. streamflow, river temperature and salinity.

Knowledge of the annual water balance within a catchment is still hampered by an insufficient understanding of winter conditions. The least known term of the winter water balance is evaporation - particularly that from intercepted snow. Methods to measure interception and evaporation are reviewed. Undisturbed meteorological conditions, high time resolution and continuous measurements of intercepted mass were taken as criteria for an ideal method. The ideal method should work during periods of melt and sublimation and above rough forest surfaces. None of the existing methods fulfils all the criteria. Most traditional hydrological methods to estimate winter-time evaporation, such as runoff and snow-course studies, non-weighing lysimeters, and precipitation and throughfall measurements, produce too little information to distinguish the causes of the observed changes. Among micrometeorological methods, only the eddy-correlation technique is suitable, and should be used in combination with interception measurements. Of the methods primarily designed to quantify interception, i.e. collection of intercepted snow, visual or photographic observations, displacement transducers, weighing of branches and trees and gamma-ray attenuation, only the last two yield continuous interception records. In combination with continuous measurements of throughfall and drip, these methods can be used to study the evaporation process. The gamma-ray method can be used to measure whole-canopy losses, whereas the other two methods deal only with losses from a branch or a single tree. The weighing-lysimeter technique produces the same information as the method of weighing a cut tree but employs a more complex set-up. The weighing-tree method combined with measurement of drip is one of the current methods that is most suitable for process studies. A new device for this purpose is presented, which differs from earlier designs by allowing a continuous weighing of drip in a tray of 2m width placed around the tree. The tray was fastened to the tree with wires and was automatically raised and lowered, thus allowing continuous measurements of both intercepted mass (with the tray raised) and evaported mass (with the tray lowered). A maximum evaporation rate of 0.3 mm h-1 was recorded with this device on 22-23 March 1990 at a site close to Luleå in northern Sweden. This maximum value was measured when wet snow was present, and confirmed earlier reports of high evaporation rates. The maximum evaporated mass during 24 h was 3.3 mm.

Snow storage on a coniferous forest canopy was measured using γ-ray attenuation and tree weighing systems, along with measurements of throughfall, using two plastic sheet net rainfall gauges. Meteorological parameters were measured with an automatic weather station. Estimates of evaporation of intercepted snow show an average rate of 0.24 mm h-1 and a maximum cumulative total of 3.9 mm in 7 h. Comparison with evaporation determined by a combination method with two different estimates of aerodynamic resistance (the "standard" rain aerodynamic resistance raL and a snow aerodynamic resistance raS-an order of magnitude larger than raL) showed that raL overestimated the evaporation by a factor of 2.6, whereas raS gave fair agreement with the measured evaporation. A multilayer model may be needed to take into account the variations of latent heat source area. Using the long-term measurements of the weight of snow on a single tree the total interception evaporation was estimated to be of the order 200 mm year-1.

Green roofs have emerged as effective stormwater management systems, but understanding the contribution of their various components to hydrological performance is crucial for optimizing their design and implementation. More empirically measured data on the hydrological function of green roof vegetation is needed, especially under realistic low-maintenance, non-irrigated scenarios. Further, targeted, evidence-based plant selection based on ecological theories may improve green roof hydrological performance. Previous research has suggested that, in contrast to monocultures, mixtures of species with complementary traits could optimize provisioning of various ecosystem services. Thus, species mixtures based on their adaptive life strategy using the CSR theory (Competitor, Stress tolerator, and Ruderal) were hypothesized to have better hydrological performance than a Sedum monoculture or bare substrate under natural conditions over multiple seasons. To test this hypothesis, the runoff from thirty 2 m2 green roof modules was measured. The retention and detention performance of different green roof treatments were evaluated for 84 precipitation events of varying rain depth and intensity during snow-free periods. Differences in retention as well as detention between the vegetation treatments varied, but generally increased with increasing rain event volume and the Stress-tolerant treatment generally performed better than bare substrate. On a mean event basis, the mixture of stress-tolerator species demonstrated a 74 % retention rate, while the Bare substrate retained 72 % of the rainfalls. Overall, the green roofs, including bare substrate and vegetated treatments, effectively retained >50 % of the cumulative precipitation depth. In line with previous studies, the Sedum monoculture generally showed worse hydrological performance than other non-succulent vegetation mixtures, despite its relatively high cover and survival. The vegetated treatment with the highest species richness and diversity in life strategies (Mix) did not provide the best vegetation cover, or hydrological performance. Instead, the Stress-tolerant treatment, characterized by the high survival rate of a single graminoid species, consistently demonstrated superior event-based stormwater retention and peak attenuation capabilities.

Reduction of runoff flow peaks and volumes is one of the performance objectives of grass swales in the context of Green Stormwater Infrastructure (GSI). Towards this end, a study of the feasibility of using a retrofitted swale outlet control weir (SOCW) to reduce runoff volume and peak flow, by enhancing swale runoff storage and infiltration into swale soils, was conducted in Luleå, Northern Sweden. Experimental field work consisted of 43 irrigation-driven runoff experiments, mimicking rainfall events with return periods between 1- to 50-years, with a constant intensity and duration of 30 min, in a 30-m long grass swale section. Experimental results confirmed that, under the tested conditions, swales with the retrofitted outflow control, reduced runoff volumes and peak flows. Such reductions ranged from 32 percentage points (for 2-year) to 1 and 4 percentage points (for 50-year return period) for runoff volumes and peak flows, respectively. Outcomes of scenarios with outflow controls clearly indicated a decreasing performance with increasing flow rates (and irrigation event return periods). Furthermore, the retrofitted swale controlled the outflow release during less frequent 20 to 50-year events, which would contribute to reducing flood risks in downstream urban areas.

Almost all drinking water in Gotland is groundwater and is mainly found in karst limestone. However, the unpredictable location and geometry of the karst cracks and caverns makes it very difficult to estimate groundwater storage and movement, as well as contaminant transport. The aim of this study was to test the performance of different geophysical techniques like Magnetic Resonance Sounding (MRS), Radiomagnetotelluric (RMT), Vertical Electrical Sounding (VES) and Ground Penetrating Radar (GPR) to characterize aquifers in Gotland, in respect to geometry and storage as well as connectivity over a wider area. The investigated area is located on the south-eastern part of Gotland. The geology here is dominated by 50 to 60 m thick successions of limestone that gradually turn into marlstone. The use of multiple techniques has shown to give a more coherent interpretation. However, the shallow penetration depth of GPR and the lack of soil cover in some places of the investigated area limit the use of geoelectrical methods and GPR. With MRS, water are found down to 60 m in depth, with a maximum water content at depths of 20 to 30 m. This coincides with the most resistive sections of the limestone. The water content varies between 0 and 3%, with a relaxation time (T1MRS) less than 400 ms suggesting that the aquifer is hosted in small fractures, molds and vugs rather than larger karst fractures and caverns. Two potential aquifers were identified with MRS, possibly separated by marlstone. From modelling it can be seen that such boundary separating two aquifer apart can be more easily discriminated in the N/S-, than in the E/W direction. In summary, MRS is therefore the only method in this survey that can detect and determine the vertical and lateral distribution of water within the aquifer together with the total volume of free water. The RMT method has shown to be effective in characterizing the limestone/marl interface, but also to locate anomalous low resistive zones, possibly associated with salt water. RMT also helps to constrain the final MRS model by choosing a suitable regularization for the MRS 3D inversion. All together, the combination of MRS and RMT seems most efficient of the tested methods and therefore most promising for future groundwater explorations in geological environments like in eastern Gotland.

The feasibility of simulating the hydrological response of a grass swale to runoff inflows was examined using the hydrological model Mike SHE and the available input data from 12 irrigation events mimicking runoff from block rainfalls. The test swale channel had a trapezoidal cross-section, bottom slope of 1.5%, length of 30 m, and was built in loamy fine sand. The irrigation events consisted in releasing two equal constant inflows to the swale: a concentrated longitudinal flow at the upstream end and a distributed lateral inflow along the swale side slope adjacent to the contributing drainage area. The total inflows approximated runoff from two events with return periods of 2 months and 3 years, respectively, for durations of 30 min. Irrigation experiments were done for two states of the initial soil moisture, dry or wet antecedent moisture conditions (AMC). Mike SHE has been extensively used on catchments of various sizes, but rarely for small stormwater management facilities and their detailed topography investigated in this study. The latter application required high spatial and temporal resolutions, with computational cells of 0.2 × 0.2 m and time steps as short as 0.6 s to avoid computational instabilities. For dominant hydrological processes, the following computational options in Mike SHE were chosen: Soil infiltration – the van Genuchten equation, unsaturated zone flow – the one-dimensional Richards equation, and overland flow – the diffusive wave approximation of the St. Venant equations. For study purposes, the model was calibrated for single events representing one of four combinations of low and high inflows, and dry and wet AMC, and then applied to the remaining 11 events. This was complemented by calibration for two events, representing high inflow on wet AMC and low inflow in dry AMC. The goodness of fit was statistically assessed for observed and simulated peak flows, hydrograph volumes, Nash-Sutcliffe model efficiencies (NSE), and soil water content (SWC) in swale soil layers. The best fit (NSE > 0.8) was obtained for high inflows and wet AMC (i.e., when the primary swale function is flow conveyance); the least fit was noted for low inflows and dry AMC, when the primary swale function is flow attenuation. Furthermore, this observation indicates the overall importance of correct modelling of the soil infiltration. The effects of spatial variation of SWC on the swale discharge hydrograph could not be confirmed from simulation results, but high topographical accuracy was beneficial for reproducing well the locations of the observed water ponding. No significant increases in simulated SWC at 0.3 m or greater depths were noted, which agreed with field observations. Overall, the results indicated that Mike SHE was effective in process-oriented small-scale modelling of grass swale flow hydrographs.

In applying digital filtering techniques for hydrological forecasting by means of cascade models a discrete model of the continuous processes involved is necessary. The paper deals with the discretization of the continuous Nash cascade. The principles of state space modelling are used. The continuous state variable model of the Nash cascade together with the computation of the state transition matrix and the impulse response is given. A discrete state variable model is derived whose state and input transition matrices are in a dual relationship. Any different discrete state representations of the continuous Nash cascade are related to each other via a linear transformation. In the limit case the discrete state models are identical with the continuous Nash cascade. The discrete state models are discretely coincident with the continuous model but at the same time consider the dynamic changes in between the sampling points. A numerical example is presented

Turbulent structures generated by vegetation patches play a dominant role in the dispersion of suspended sediment, which in turn is of great significance for ecosystem cycling and river geomorphology development. High fidelity Large Eddy Simulations (LES) coupled with the Discrete Phase Method (DPM) were used to explore the particle distribution and its variance (the non-uniformity in temporal and spatial space) in a partially vegetated straight channel. The novel findings and conclusions are outlined here. Firstly, the contour of the vertical vorticity component coincides well with particle preferential gatherings in the outer edge of the mixing layer in the near-bed region. Large-scale turbulent structures grow in mixing layer along the side of a vegetation patch (VP), which deplete particles away from the mixing layer into the neighbouring region. Also, higher vegetation densities (Dn) promote this depletion trend. Secondly, the Probability Density Function (PDF) and its variance were defined to quantify these phenomena, illustrating that the VP continuously interrupts the flow condition and promotes higher non-uniformity of particle distribution among the vegetated and non-vegetated regions. The variance of the PDF in the non-vegetated region is significantly higher than that in the neighbouring vegetated region located in the same streamwise location. The particle parcels are highly unevenly located along the periphery of the large eddies and are exchanged by the mixing flow between the non-vegetated and vegetated regions. Finally, the vertical entrainment of particles occurs in the vegetated region of the present cases. This is because the horseshoe structures provide an upwards velocity for the current Dn conditions (Dn<0.1) and an increase of Dn (Dn<0.1) accelerates the upward suspension. These findings complete our understanding of particles’ transportation in both spanwise and vertical directions.