FLOODING DYNAMICS AND NUTRIENT RETENTION IN THE MIDDLE EBRO FLOODPLAIN: EXPERIMENTAL ASSESMENT AND NUMERICAL MODELING
ABSTRACT
The present work highlights the numerical simulation as a tool capable of reproducing and predict the main processes that produces and maintains the floodplain ecosystems. To that end, floodplain flow dynamics, geomorphic activity, sediment deposition and nutrient uptake are evaluated through field experimentation. Then, the experimental data are included in a numerical model to perform a complete simulation tool that predict flow dynamics, geomorphic activity, sediment deposition, river nutrient contribution and nutrient uptake. A 2 km river segment representative of the meandering Middle Ebro (NE Spain) reach is selected for the study. 
The first task is to find the best representation of the floodplain hydraulics. We select a two-dimensional (2D) finite volume numerical model based on the 2D transient shallow water equations as the best option to perform the hydrodynamic simulation. Then, the importance of the correct characterization of the roughness coefficient and the topography is emphasized in the study. The former is estimated from a previous classification of structurally homogeneous habitats and the latter is defined by merging Digital Terrain Model data with a hydraulic river bed elevation reconstruction algorithm. The calibration of the full model resulting from the roughness, bed river and flow simulation models is based on field measurements of flooded area for two steady discharges of 50 and 500 m3/s. The validation is performed by comparing the numerical results with the water levels measured during five flooding events at certain times, with the flooded area and with time series of continuous point measurements of water-depth during different situations along the year $2007$. The validation results of the flooded area and water level were 79±13 % and 0.27±0.05 m, respectively.
Since the model provide accurate predictions of the floodplain flooding, for both, low and high flow discharges, the simulation results are used to analyze the current floodplain flooding dynamics and its geomorphic potential. As a result, we obtain that although the current flow regime appears to be enough to induce the floodplain river nutrient contribution, it does not produce enough morphological activity to maintain the shifting mosaic of habitats (SMH) characteristic of the floodplain ecosystems. On the basis of this analysis, five possible restoration scenarios, based on the terrain or flow modification, are simulated. The results enhance the role of the constructed defenses in the morphological activity decreasing.
A predictive nutrient uptake model is included into the 2D hydraulic model. For that purpose, retention of dissolved phosphorus and nitrogen is examined in several controlled experiments in an irrigation canal, located within the study site and whose flow comes from the Ebro river, and in a laboratory channel. The laboratory channel is used to estimate the sediment nutrient uptake capacity, whilst the irrigation canal is used to estimate the sediment and water column nutrient uptake capacity. Retention efficiency is measured using nutrient short-term nutrient and tracer injections to estimate nutrient uptake coefficient (k) during the irrigation period. Hydraulic tracer (Br-) and soluble reactive phosphorus (SRP) were co-injected into both canals. Nitrogen uptake is measured using the ambient concentration. The results show an SRP load reduction of 20.7±2.8 % of the net mass balance by means of the irrigation channel ecosystem, while the nitrogen concentration remained nearly unaltered. Hence, only the SRP is considered in the nutrient uptake formulation.
The new SRP uptake formulation is performed using the experimental data. First, main nutrient uptake agents are determined by means of an statistical analysis of the nutrient uptake coefficient and the physical, chemical and biological irrigation canal and laboratory channel characteristics. As a result, the sorption appears to be the main SRP uptake process, for both the sediment and the water column. Then, a non linear regression between the main SRP uptake agents and the uptake coefficient is performed obtaining an SRP uptake function. The equation is included into the 2D hydraulic model as a decay term and validated in both the irrigation canal and the laboratory channel. The comparison between measured and simulated data provided always a significant linear regression (p?0.05) whose r2?0.83 in 19 out of 20 experiments, where 3 of them were carried out using a different spatial-temporal scale. Finally, the new formulation is validated in the selected river reach by means of three new experimentations. The model predictions show a good accuracy, where the linear regressions between measured and calculated SRP concentration of the three experiments were significant r2?0.62; p?0.05).
The particulate solute transport and decay (sedimentation model) is also included into the hydraulic model, and is validated using field data collected during 2 real flooding events. The comparison between calculated and measured sediment deposition shows a significant (p?0.05) linear regression, whose r2=0.97 with a slope not significantly different from the unit (p?0.05).
The complete model that includes the erosive potential, the solute transport and SRP uptake is used to simulate and analyze floodplain sediment deposition, river nutrient contribution and SRP uptake. According to this analysis, the main SRP uptake process appears to be the sediment sorption. The analysis also reveals disequilibrium between erosion and sedimentation, where sediment deposition prevails over erosive processes. Finally, simulation results suggest the presence of a lateral gradient of hydrological connectivity that decreases as distance to the river increases and control the floodplain river matter contribution. According to this gradient, remote floodplain zones would receive very low or null suspended and dissolved nutrients, whilst the adjacent riparian areas would receive the highest concentrations. Hence, this lateral gradient could cause a lack of nutrient plant availability at the remote riparian forest.
Finally, the complete model was used to propose and simulate floodplain restoration scenarios based on terrain modification.