Sandis Dejus

ONLINE MEASUREMENT USAGE FOR PREDICTING WATER AGE FROM TRACER TESTS TO VALIDATE A HYDRAULIC MODEL

The tracer tests were performed using two approaches in the drinking water distribution network (1374 km) of Riga city, Latvia with the goal to verify and validate a hydraulic model build up in Epanet 2.0. The water is supplied for 700 000 inhabitants with average daily water demand of 1500 l/s from six different water sources. The each water source has a different electric conductivity (EC) value with the lowest numbers for deep groundwater 272±25 µS/cm (25 O C) and the highest for surface and artificially recharged groundwater 468±100 and 784±18 µS/cm, respectively. The significant difference for EC in sources was used as a natural tracer. The first measurement approach is based on field works during shut-down or start-up of WTP due to monthly working regime or yearly drainage of clear water tanks at stations located in the north-east of city. In total more than 20 locations in distribution network were used for measurements of EC from hydrants and/or at taps in buildings applying online conductivity meters 3400sc™ (Dr. Bruno Lange GmbH&Co, Germany) and condu::lyser™ (s::can Mestechnik GmbH, Austria) in predictable time interval. The second approach of tracer measurements was based on EC measurements in network within close distance to track minor changes of EC due to production/consumption ratio. The hydraulic model of Riga city was validated in 8 nodes with the maximum age of 25:40 hours, while the distal nodes in model accounts for 60 hours. Results showed that in order to measure correct changes of trace element it is necessary to do it very close to transient pipes with high flow. The measurements of tracer near reservoirs in network are unpredictable due to proper mixing of water. The approach of measurements with two online instruments in close distance (up to 5 km) to track minor changes of EC can be useful for validation of hydraulic models. However, the statistical analysis (z-test) of this method showed significant difference between mean values of measurements which can be explained with too many uncertainties (cross-connections of parallel lines, loops) for water mixing in city with population of 700 000 inhabitants.