O. Fovet

Process consistency in models: The importance of system signatures, expert knowledge, and process complexity

Hydrological models frequently suffer from limited predictive power despite adequate calibration performances. This can indicate insufficient representations of the underlying processes. Thus, ways are sought to increase model consistency while satisfying the contrasting priorities of increased model complexity and limited equifinality. In this study, the value of a systematic use of hydrological signatures and expert knowledge for increasing model consistency was tested. It was found that a simple conceptual model, constrained by four calibration objective functions, was able to adequately reproduce the hydrograph in the calibration period. The model, however, could not reproduce a suite of hydrological signatures, indicating a lack of model consistency. Subsequently, testing 11 models, model complexity was increased in a stepwise way and counter-balanced by “prior constraints,” inferred from expert knowledge to ensure a model which behaves well with respect to the modelers perception of the system. We showed that, in spite of unchanged calibration performance, the most complex model setup exhibited increased performance in the independent test period and skill to better reproduce all tested signatures, indicating a better system representation. The results suggest that a model may be inadequate despite good performance with respect to multiple calibration objectives and that increasing model complexity, if counter-balanced by prior constraints, can significantly increase predictive performance of a model and its skill to reproduce hydrological signatures. The results strongly illustrate the need to balance automated model calibration with a more expert-knowledge-driven strategy of constraining models.

Identifying seasonal patterns of phosphorus storm dynamics with dynamic time warping

Phosphorus (P) transfer during storm events represents a significant part of annual P loads in streams and contributes to eutrophication in downstream water bodies. To improve understanding of P storm dynamics, automated or semiautomated methods are needed to extract meaningful information from ever-growing water quality measurement data sets. In this paper, seasonal patterns of P storm dynamics are identified in two contrasting watersheds (arable and grassland) through Dynamic Time Warping (DTW) combined with k-means clustering. DTW was used to align discharge time series of different lengths and with differences in phase, which allowed robust application of a k-means clustering algorithm on rescaled P time series. In the arable watershed, the main storm pattern identified from autumn to winter displayed distinct export dynamics for particulate and dissolved P, which suggests independent transport mechanisms for both P forms. Conversely, the main storm pattern identified in spring displayed synchronized export of particulate and dissolved P. In the grassland watershed, the occurrence of synchronized export of dissolved and particulate P forms was not related to the season, but rather to the amplitude of storm events. Differences between the seasonal distributions of the patterns identified for the two watersheds were interpreted in terms of P sources and transport pathways. The DTW-based clustering algorithm used in this study proved useful for identifying common patterns in water quality time series and for isolating unusual events. It will open new possibilities for interpreting the high-frequency and multiparameter water quality time series that are currently acquired worldwide.

Release of dissolved phosphorus from riparian wetlands: Evidence for complex interactions among hydroclimate variability, topography and soil properties

In agricultural landscapes, establishment of vegetated buffer zones in riparian wetlands (RWs) is promoted to decrease phosphorus (P) emissions because RWs can trap particulate P from upslope fields. However, long-term accumulation of P risks the release of dissolved P, since the unstable hydrological conditions in these zones may mobilize accumulated particulate P by transforming it into a mobile dissolved P species. This study evaluates how hydroclimate variability, topography and soil properties interact and influence this mobilization, using a three-year dataset of molybdate-reactive dissolved P (MRDP) and total dissolved P (TDP) concentrations in soil water from two RWs located in an agricultural catchment in western France (Kervidy-Naizin), along with stream P concentrations. Two main drivers of seasonal dissolved P release were identified: i) soil rewetting during water-table rise after dry periods and ii) reductive dissolution of soil Fe (hydr)oxides during prolonged water saturation periods. These mechanisms were shown to vary greatly in space (according to topography) and time (according to intra- and interannual hydroclimate variability). The concentration and speciation of the released dissolved P also varied spatially depending on soil chemistry and local topography. Comparison of sites revealed a similar correlation between soil P speciation (percentage of organic P ranging from 35-70%) and the concentration and speciation of the released P (MRDP from <0.10 to 0.40mgl-1; percentage of MRDP in TDP from 25-70%). These differences propagated to stream water, suggesting that the two RWs investigated were the main sources of dissolved P to streams. RWs can be critical areas due to their ability to biogeochemically transform the accumulated P in these zones into highly mobile and highly bioavailable dissolved P forms. Hydroclimate variability, local topography and soil chemistry must be considered to decrease the risk of remobilizing legacy soil P when establishing riparian buffer zones in agricultural landscapes.

Adaptive control of algae detachment in regulated canal networks

Open-channels networks used for water distribution are subject to algal developments that can induce major disturbances such as clogging issues on hydraulic devices (pipes, weirs, filters, …). We already studied the use of flushes to manage these algae developments. The flush is carried out by increasing the hydraulic shear conditions using the hydraulic structures of the canal network. In response to the shear stress increase, a part of the fixed algae is detached, then re-suspended into the water column, and finally transported into the canal network. This leads to a peak of turbidity. We develop in this paper a distributed linear model that is used for real-time adaptive control of the flushes. Simulations show the effectiveness of the adaptive controller, that can at the same time estimate the gain of the system, linked to the amount of fixed algae, and perform a flush without overtopping the turbidity limit.

Managing invasive algae with flushing flows: new perspectives for canal control

Algae may cause various disturbances in irrigation canals, such as conveyance decrease, clogging and water quality depreciation. Strategies of regular flushes were designed for two open-channels in southern France. The objective of flushes is to maintain the algal biomass by increasing the shear stress in the canal, thanks to hydraulic control. During a flush, turbidity is increased, possibly causing problems for end-users or treatment plants. While automation largely improved the hydraulic performance of canals, considering water quality control is a new challenge for canal control. In this paper, open-loop and closed-loop control strategies of turbidity are described. These strategies rely on transfer functions established between discharge (command input) and turbidity (output) at different locations of the canal. Simulation models allow testing easily various control methods. Flushes were also implemented in the Canal de Provence routinely in 2011. The analysis of the field data allows improving the discharge-turbidity transfer functions. Then, real-time closed-loop control was also implemented. These tests bring a new insight to sensor positioning and canal control, applied here to non-hydraulic variables.