Nossana Karst Spring Behaviour under future Climate Change Scenarios

Nossana spring, located in the province of Bergamo, is one of the most important Northern Italy water resources for domestic supply. It sustains a large water distribution system managed by Uniacque S.p.A.. The aim of this work is to forecast the behaviour of this spring, and its hydrogeological parameters, under future climate change scenarios (IPCC, 2014). The study of the response of Nossana spring may give important guidelines to enforce managing tools in case of future severe climatic trends. Nossana spring behaviour of discharge has been modelled using the lumped-parameter model RRAWFLOW (Long and Mahler, 2013; Long, 2015). RRAWFLOW uses daily rainfall and daily minimum and maximum temperature time series as input to calculate discharge. Uniacque S.p.A. provided discharge and meteorological data for the period 1998-2018. Thirteen years (1998-2010) were taken as calibration period while the remaining eight years (2011-2018) were used to validate the model. The quality of calibration and validation was assessed through two indices, Nash-Sutcliffe Efficiency (NSE) and Root Mean Squared Error (RMSE), as suggested by Legates and McCabe (1999). For future projections of spring discharge, the meteorological variable time series needed as input in RRAWFLOW were derived by statistical downscaling of climate models, performed using change factors and a rainfall-temperature generator (Burton et al., 2010; Richardson, 1981). Change factors were calculated for five twenty-year intervals, from 2021 to 2100, for a total of 40 different regional climate model runs. These model runs, refers to three different Representative Concentration Pathway (RCP) scenarios with increasing severity (2.6; 4.5; 8.5) according to IPCC, 2014. Future simulations of spring discharge are still in course. Eventually, they will allow the preparation of probabilistic scenarios of discharge amounts under each RCP condition. Statistical significance of discharge changes in comparison to observations will be evaluated, too. RRAWFLOW simulation shows a NSE index of 0.45 and a RMSE index equal to 2.18, which indicate a satisfactory correlation, even if further improvement is required. The software has also calculated almost similar Efficient coefficients of calibration and validation (0.46 and 0.44) period, which indicates a high predictivity of the model. Results suggest that there is a strong correlation between recharge rate -both in cases of more frequent single precipitation event (summer – storm) and less frequent continuous event- and Nossana spring discharge rate. In fact, hydrograph indicates that discharge variation occurs mainly within 1-2 days from the beginning of a rainfall event. This is a typical behaviour of a karst spring in temperate regions,as in the case of the Valseriana. Preliminary results on climate change projections are still being processed; once available they will be used to simulate projected discharge behaviour. At the end of the analysis, 40 projected behaviours of Nossana spring discharge will be presented: 3 for RCP 2.6, 23 for RCP 4.5, and 14 for RCP 8.5. Water supply is one of the most important aspect that leads the general principles of water resources management: water suppliers should always guarantee an adequate quantity of water of suitable quality to meet public needs, given that the demand is reasonable and wastes are avoided (Kresic, 2013). To meet public water needs, Nossana spring, which is characterized by a variable discharge from 0.4 to 18 m3/s, must supply 0.5 m3/s of water; below this warning threshold, the water distribution system needs to resort to other sources, which are usually more expensive and difficult to manage. Simulations of extreme meteorological conditions are useful for waterworks managers to identify exceptional situations when it is necessary to integrate the water supply (i.e., activating supplementary water wells). This preliminary work could be used as an implementing tool to support the policy of the public water company in future climatic change scenarios In addition, this study demonstrated that lumped–parameter models can lead to good discharge predictions in karst environment. For this reason, if a complete simulation of the processes is not needed, they can be a better option for discharge simulation in karst environments than 3D models, which usually need a large amount of data and longer computational times.