Identifying the factors influencing the total external hydraulic loads to the dese-zero watershed

Some watersheds are heavily influenced by human factors that affect their overall water balance and dynamics. This is the case of the Dese-Zero watershed, part of the Venice Lagoon watershed - VLW, located in North-East Italy. Such watershed is characterised by a highly modified environment, with the presence of several hydraulic works and devices, which, ultimately, ensure optimal flow and water availability conditions for different water needs. In addition to this artificial-hydraulic influence, the VLW is also heavily affected by groundwater contributions coming from a mainly external aquifer located in the Venetian high plains. Hydrological modelling under these circumstances is particularly challenging and requires detailed quantified information regarding the total external contributions (i.e. groundwater and deviated surface water from/to bordering watersheds). In order to cope with such complex dynamics, this study proposes a framework contemplating a coupled mechanistic-empirical modelling approach. Under such framework, the physically-based model simulates the internal processes of the studied watershed while the empirical model accounts for the total external hydraulic influences. Data pre-processing is performed with Principal Component Analysis - PCA aiming at the identification of the main factors contributing to the total external hydraulic contribution. The SWAT model is used as the mechanistic component while two alternative modelling techniques are tested for the empirical counterpart, namely: i) Multiple Linear Regression - MLR, and ii) Artificial Neural Networks - ANN. The results suggest that, among the studied weather variables, precipitation plays a major role in the estimation of the total external hydraulic loads. Moreover, both temporal (e.g. season of the year) and stream flow (e.g. SWAT simulated output) information is also relevant for the estimation of the studied process. Finally, it is concluded that the studied coupled mechanistic-empirical model is capable of simulating the hydrology of the Dese-Zero watershed while the non-liner neural networks model is the best option for estimating the total external hydraulic loads.