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?

2

• New sensor technology allows high precision
• bservations 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.

3

Automatic measurements include chlorophyll-a, turbidity, nitrate-nitrogen, pH, DOC, electrical conductivity…

Before… After cleaning… Turbidity: 20 stations Nitrate: 9 DOC: 2 pH: 7 EC: 6 Chlorophyll-a: 2 Rivers, lakes and small catchments

18.5.2015

Used water quality sensors

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

18.5.2015

Differences of the sensor types

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 NO3-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

6

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

7

Turbidity: 3 stations in Finland

Raw data, collected continuously into the Hydrotempo database at 1 hour interval. 1.4.2015  5.5.2015 Nitrate-N mg/L

Vantaanjoki river Vantaanjoki river

Savijoki small catchment

• Tot. Org. carbon mg/L

Agric. basin

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

9

• 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
• f the stream) can make cleaning very challenging.

18.5.2015

Pitkäkoski measurement station at the river Vantaanjoki

Assembling the sensor in October 2010 Maintenance of sensor in winter 2011 The hole in the ice remained unfrozen under the Finnfoam insulation plates s::can nitro::lyser sensor

18.5.2015

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.

Database Raw Data Database WEB interface WEB interface Quality control Supervision

• f quality

Control User 1 Maintenance Quality control and calibration User 3… User 2 Luode sensor nodes a-Lab sensor nodes User 1 User 3… User 2 User n Automatic warnings Requests for maintenance Requests for maintenance Status information Maintenance and control of sensor nodes Data query Data query Data query Data query Automatic alarm Control of procedure Data query Data query Data query Database for validated data Message database

Data flow and communication system with main data services of SoilWeather WSN

Savijoki 2011-2012 Calibration equations

y = 1.2098x - 43.963 R² = 0.9764 100 200 300 400 500 600 700 800 200 400 600 800 Turbidity FNU Turbidity [FTU]_autom. y = 1.048x + 0.0311 R² = 0.9447 1 2 3 4 5 6 7 8 2 4 6 8

Nitrate-N mg/L

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)

y = 1,12x + 0,76 R² = 0,950

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9

Total nitrogen mg/l Nitrate-N mg/l

y = 1,20x + 70,06 R² = 0,940

100 200 300 400 500 600 700 800 900 1000 100 200 300 400 500 600 700 800 900 1000

Ptot, µg / l Turbidity FNU

One example, how the different hydrological flow patterns are caughted

Load calculations, Aurajoki in south-western Finland

Load calculated with Tot P load (kg/a) Tot P load (kg/ha/a) Tot N load (kg/a) Tot N load (kg/ha/a) Hertta, monthly mean concentrations multiplied with monthly Q

59 267 0.68 901 533 10.32

Sensor, monthly mean concentrations multiplied with monthly Q

59 887 0.69 602 374 6.89

Sensor, daily mean concentrations multiplied with daily Q

76 354 0.87 763 532 8.74 High frequency water quality data allow also more accurate load estimates if precise flow data is available.

17

Location of 5 turb. mesurement stations in the Karjaanjoki River basin

Calibration eqs. (r2) 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.

RIVER BASIN SCALE

• Environ. Monit. Assess 187(2015)

Aerial image of Hovi wetland

Locations of sensors Wetland area: 0.6 ha (5% of the catchment)

Sensor-detected turbidity in the Hovi CW during 2007–2010

1000 2000 3000 4000

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 9.11.08 10.12.08 9.1.09 9.2.09 11.3.09 11.4.09 11.5.09 11.6.09 11.7.09 11.8.09 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 4.10.10 4.11.10 4.12.10 Turbidity (FTU) Time

Inflow Outflow

Winter break during Jan.-Apr. 2010

Sensor-detected NO3-N concentration in the Hovi wetland during 2007–2010

5 10 15 20 25 30 35 40

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 9.11.08 10.12.08 9.1.09 9.2.09 11.3.09 11.4.09 11.5.09 11.6.09 11.7.09 11.8.09 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 4.10.10 4.11.10 4.12.10 NO3-N (mg/l) Time

Inflow Outflow

Winter break during Jan.-Apr. 2010

Model vs. measurements

21

1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 25.2. 7.3. 17.3. 27.3. 6.4. 16.4. 26.4. 6.5.

Total nitrogen mg/l

year 2015

• Automat. totN

VEMALA model

y = 1,1x + 0,39 R² = 0,309

1 1.5 2 2.5 3 3.5 4 1 1.5 2 2.5 3 3.5 4

• Autom. Ntot, µg/l

VEMALA- Ntot, µg/l

Model vs. measurements totP totP

22

50 100 150 200 250 300 25.2. 7.3. 17.3. 27.3. 6.4. 16.4. 26.4. 6.5.

Total phosphorus μg/l

TotP VEMALA TotP, autom

y = 1x + 16 R² = 0,686

50 100 150 200 250 300 100 200

• Autom. Ptot, µg/l

VEMALA- Ptot, µg/l

23

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

• f these substances

24

Kiitos! Thank you!