Experiences from the use of sensors for assessing water quality in - PowerPoint PPT Presentation

Experiences from the use of sensors for assessing water quality in rivers in Finland HELCOM workshop on total uncertainties in the input estimates, 18-19.5.2015 Uppsala, Sweden Sirkka Tattari Finnish Environment Institute (SYKE) Helsinki, Finland With contributions to my colleagues: Jari Koskiaho, Elina Röman, Jarmo Linjama

Why to use automatic water quality monitoring? ● New sensor technology allows high precision observations of multiple water quality variables. It provides high frequency data in a cost-efficient way (considering the number of measurements) that allows covering most of the peak events. High frequency water quality data allow also more accurate load estimates if precise flow data is available. ● It also provides more accurate data for modeling and model calibration and contributes to a better understanding of in-stream processes, flow pathways, and how effective different management actions and mitigation measures are. 2

Automatic measurements include chlorophyll-a, turbidity, nitrate-nitrogen, pH, DOC, electrical conductivity… Turbidity: 20 stations Nitrate: 9 DOC: 2 pH: 7 EC: 6 Before … Chlorophyll-a: 2 After cleaning … Rivers, lakes and small catchments 3

Used water quality sensors Sensor type Manufacturer Finnish supplier 18.5.2015 OBS3+ Campbell Scientific, A-lab Oy Inc. (www.a-lab.fi) (www.campbellsci.co m) s::can scan Messtechnik Luode Consulting nitro::lyser GmbH Oy (www.s-can.at) (www.luode.net)

Differences of the sensor types 18.5.2015 Functioning principle • OBS3+ sensor works by emitting near-infrared light into the water, then measuring the light that bounces back from the suspended particles • Functioning of s::can nitro::lyser is based on a continuous optical spectrum reaching from low ultraviolet to visible light, which makes it possible to measure NO 3 -N concentration simultaneously with turbidity Cleaning of the sensor lenses • OBS3+ sensors were equipped with a battery-powered mechanical wiper brushes • s::can sensor lenses were cleaned by bursts of compressed air generated by either electric-powered compressor or exchangeable bottle of pressurized air

HYDROTEMPO database for real-time automatic monitoring: Usage at one's own risk!!! 6

Savijoki small Turbidity: 3 stations in Finland catchment Raw data , collected continuously into the Nitrate-N mg/L Hydrotempo database Agric. at 1 hour interval. basin Vantaanjoki 1.4.2015  5.5.2015 river Tot. Org. carbon mg/L Vantaanjoki river 7

Water quality sensors ● Savijoki small catchment, since 06/2007 ● Nitrate-N, turbidity 18.5.2015

Selection of location for monitoring is amongst the first things to be considered ● Sensors should be located deep enough in the water to prevent wrong measurements if too close to the bottom or due to damages by ice and intensive biofilm formation if too high. ● Formation of biofilm on the measuring sensors, especially in summer, is common as well as sedimentation to the sensors. ● Wrongly selected location of the device (e.g. in the middle of the stream) can make cleaning very challenging. 9

Pitkäkoski measurement station at the river Vantaanjoki 18.5.2015 Assembling the sensor in October 2010 Maintenance of sensor in winter 2011 s::can nitro::lyser sensor The hole in the ice remained unfrozen under the Finnfoam insulation plates

Good quality data can be produced only if proper maintenance procedures are followed. It includes periodic manual removal of organisms and sediments or automated cleaning of sensors/lenses with liquids, compressed air or mechanically by brush. 18.5.2015

Data flow and communication system with main data services of SoilWeather WSN Automatic alarm Database Supervision Quality for of quality control validated Control Control of Requests for maintenance data procedure Status information Data query Data query Maintenance User 1 Message database Raw Data WEB interface Database Maintenance Data query and control of a-Lab sensor nodes sensor User 2 nodes Data query Automatic warnings User 3… User n Requests for Data query maintenance User 1 WEB interface Database Data query Luode sensor User 2 nodes Quality Data query control and calibration User 3…

Savijoki 2011-2012 Calibration equations 8 800 y = 1.048x + 0.0311 y = 1.2098x - 43.963 7 700 R² = 0.9447 R² = 0.9764 Nitrate-N mg/L 6 600 5 500 Turbidity FNU 4 400 3 300 2 200 100 1 0 0 0 200 400 600 800 0 2 4 6 8 Turbidity [FTU]_autom. Nitrate-N [mg/l]_autom. Production of reliable data requires calibration of sensors that can be done by using samples from the studied water body. Each sensor is different and thus local calibration should be sensor specific. If the sensor is mounted into a new place or land use in the catchment area changes, local calibration has to be rearranged.

Savijoki 2011-2012 Conversion equations (n=43) 1000 9 y = 1,20x + 70,06 y = 1,12x + 0,76 900 8 R² = 0,940 R² = 0,950 800 Total nitrogen mg/l 7 700 6 Ptot, µg / l 600 5 500 4 400 3 300 200 2 100 1 0 0 0 100 200 300 400 500 600 700 800 900 1000 0 1 2 3 4 5 6 7 8 9 Nitrate-N mg/l Turbidity FNU

One example, how the different hydrological flow patterns are caughted

Load calculations, Aurajoki in south-western Finland Tot P load Tot P load Tot N load Tot N load Load calculated with (kg/a) (kg/ha/a) (kg/a) (kg/ha/a) Hertta, monthly mean concentrations multiplied 59 267 0.68 901 533 10.32 with monthly Q Sensor, monthly mean concentrations multiplied 59 887 0.69 602 374 6.89 with monthly Q Sensor, daily mean concentrations multiplied 76 354 0.87 763 532 8.74 with daily Q High frequency water quality data allow also more accurate load estimates if precise flow data is available.

RIVER BASIN SCALE Location of 5 turb. mesurement stations in the Karjaanjoki River basin Calibration eqs. (r 2 ) Billnäs: y =2,71* x (0,94) Väänteenjoki: y =2,8+2,12* x (0,86) Häntäjoki: y =0,5+5,23* x (0,90) Olkkalanjoki: y =2,51* x (0,97) Vanjoki: y =1,08* x (0,93) Difference in material flux (total SS) in the Karjaanjoki basin in 2009-2012 as calculated on the basis of (i) automatic monitoring and (ii) water sampling. Environ. Monit. Assess 187(2015) 17

Aerial image of Hovi wetland Wetland area: 0.6 ha (5% of the catchment) Locations of sensors

Turbidity (FTU) 1000 2000 3000 4000 Sensor-detected turbidity in the Hovi CW during 0 10.10.07 9.11.07 10.12.07 9.1.08 9.2.08 10.3.08 10.4.08 10.5.08 10.6.08 10.7.08 10.8.08 9.9.08 10.10.08 2007 – 2010 9.11.08 10.12.08 9.1.09 9.2.09 11.3.09 Time 11.4.09 11.5.09 11.6.09 11.7.09 11.8.09 Winter break during Jan.-Apr. 2010 10.9.09 11.10.09 10.11.09 11.12.09 10.1.10 5.5.10 4.6.10 5.7.10 4.8.10 Outflow Inflow 4.9.10 4.10.10 4.11.10 4.12.10

NO3-N (mg/l) Sensor-detected NO3-N concentration in the Hovi 10 15 20 25 30 35 40 0 5 10.10.07 9.11.07 10.12.07 9.1.08 9.2.08 10.3.08 10.4.08 10.5.08 wetland during 2007 – 2010 10.6.08 10.7.08 10.8.08 9.9.08 10.10.08 9.11.08 10.12.08 9.1.09 9.2.09 11.3.09 Time 11.4.09 11.5.09 11.6.09 11.7.09 11.8.09 Winter break during Jan.-Apr. 2010 10.9.09 11.10.09 10.11.09 11.12.09 10.1.10 5.5.10 4.6.10 5.7.10 4.8.10 4.9.10 Outflow Inflow 4.10.10 4.11.10 4.12.10

Model vs. measurements 6.0 year 2015 5.5 4 5.0 3.5 y = 1,1x + 0,39 Autom. Ntot, µg/l R² = 0,309 4.5 Total nitrogen mg/l 3 Automat. totN VEMALA model 4.0 2.5 3.5 2 3.0 1.5 2.5 1 1 1.5 2 2.5 3 3.5 4 2.0 VEMALA- Ntot, µg/l 1.5 1.0 25.2. 7.3. 17.3. 27.3. 6.4. 16.4. 26.4. 6.5. 21

Model vs. measurements totP totP 300 TotP VEMALA TotP, autom 300 250 250 Autom. Ptot, µg/l y = 1x + 16 Total phosphorus μ g/l 200 R² = 0,686 200 150 100 150 50 100 0 0 100 200 VEMALA- Ptot, µg/l 50 0 25.2. 7.3. 17.3. 27.3. 6.4. 16.4. 26.4. 6.5. 22

SUMMARY ● Quality of the collected data can vary a lot depending on selection of location for monitoring, maintenance of the devices and data handling. Therefore, reliability of the produced data series is not always sufficient. ● The existing measuring methods differ in their functioning principle, measurement range and accuracy. ● Only limited number of variables can be measured with the presently available sensors. Nitrates, which are often the major N fraction in agricultural runoff, can be measured directly. Turbidity is often highly correlated with suspended solids and total P concentrations, enabling load calculations of these substances 23

Kiitos! Thank you! 24