|M.Sc Student||Yardeni Anat|
|Subject||On-Line Biomonitoring for Pollution Detection in Potable|
Water Network with the FishToximeter -
A Video Based Early Warning System
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Eran Friedler|
|Dr. Ram Porat|
|Full Thesis text - in Hebrew|
Biological early warning systems (BEWS) are used for detecting toxic chemicals in water systems.
This work describes the implementation of the FishToximeter (FishTox; bbe, Germany), a Video-based BEWS, at the National Water Carrier of Mekorot Eshkol Site, and the modification of its operation, to enable alarm generation upon exposure to low concentrations of selected contaminants, while minimizing false alarms.
The FishTox consists of fish aquarium, video camera, and alarm evaluation software which analyzes the video image of the fish online, and continuously calculates changes in the fish behavioral characteristics and compares them to patterns of preceding hours or to pre-set limits.
A contaminant invasion can cause changes in the swimming behavior. Then, the software allocates "toxicity points", the accumulation of which - leads to alert or alarm.
Striped-Bass fingerlings, the selected test fish, were found to be adaptable to continuous alternating exposure to desalinated and Lake Kinneret water and mixtures of the two, but sensitive to the presence of CO2 in water, with evaluated LC50 concentrations of 31 to 98 mg-CO2/L for exposure times of 72 to 1.5 h, respectively. Experiments showed that 30 minutes exposure to chlorine concentrations above 0.3 mg/L was fatal to the fingerlings, while lower ones caused behavioral changes that could lead to onset of alarm.
The Bass fingerlings in the FishTox were continuously exposed to un-contaminated Eshkol reservoir water for three years. During this period thirty toxicity experiments were performed in the FishTox, with varying concentrations of four organic contaminants from different groups which were dosed via the inflow. In each experiment fish were exposed for 1 h, after which uncontaminated water was fed to the FishTox. Fish behavior was monitored for 48 h after the exposure or until the fish died.
Data from experiments and non-test periods, were run in a simulation program through which analysis of parameters that caused the most false alarms (in non-test periods) but the least real alarms (toxicity experiments) was performed, using Venn diagrams and ROC plots. Following this analysis weights of some coefficients of behavioral parameters in the software were modified. The adjusted coefficient enabled on-setting of alarm within 120 minutes after 1 h exposure to 0.38-1.7 mg/L of the four contaminants, and detection of 81% of the intentional contamination events (the maximum detection ability of the FishTox given the experiments performed). This set of coefficients is expected to cause no more than two false alarms per year.