Monitoring water pollution in the river Ganga with innovations in airborne remote sensing and drone technology

Transcription

1 Monitoring water pollution in the river Ganga with innovations in airborne remote sensing and drone technology RAJIV SINHA, DIPRO SARKAR DEPARTMENT OF EARTH SCIENCES, INDIAN INSTITUTE OF TECHNOLOGY KANPUR, INDIA PATRICE CARBONNEAU DEPARTMENT OF GEOGRAPHY, UNIVERSIY OF DURHAM, UK

2 AIRBORNE REMOTE SENSING Airships Aeroplanes UAVs Helicopters Fixed Wing UAV Advantage Efficient aerodynamics Greater coverage High speed, Natural gliding w/o power Mildly wind /rain resistant Greater payload Disadvantage Needs a runway/launcher Cannot stay stationary Rotary Wing UAV Advantage Vertical take-offs Agile manoeuvrings Payload configurable Can stay stationary Disadvantage Smaller coverage More maintenance due to more moving parts Susceptible to damage during wind /rain (Lavender, 2005).

3 WATER QUALITY ASSESSMENT Water quality refers to the chemical, physical, biological, and radiological characteristics of water. Limitations of prevailing Water Quality measurements methods: The techniques are labour intensive, time consuming, and costly. Involves long term commitment on local infrastructure. Low awareness of the social importance of water quality leads to lower priority in funding leading to operational budget cuts. Point data source seldom reveals the big picture when it comes to assessing large waterbodies. Site inaccessibility can hinder the acquisition of data. Field error during sample acquisition and laboratory error during processing affects the result. (El-Din et al, 2013; Ritchie, 2003; Chipman,2009.)

4 REMOTE SENSING METHOD OF WATER QUALITY ASSESSMENT Advantages of Remote Sensing method: Provides a synoptic view of large waterbodies which is more effective in monitoring the temporal and spatial changes Takes lesser time to get a comprehensive idea. Variation in concentration are more prominent due to continuous data. Optically active parameters - Chlorophyll a, Temperature, Coloured Dissolved Organic Matters, Total Suspended Matters, Dissolved Organic Carbon, Turbidity, Sea Surface Salinity. Reflectance vs wavelength Less costly. (Seyhan et al,1974, Kallio 2000)

5 GENERAL METHODOLOGY Empirical Approach: Where, Y =A+BX or Y = AB X Y = Reflectance value A & B = empirically derived factors determined statistically from the spectral value and in-situ measurements. (Seyhan et al,1986, Conrad et al, 1971)

6 MULTISPECTRAL VS HYPERSPECTRAL Multispectral In a single observation a multispectral sensor generate 3-10 spectral bands Covers large spectral bands Images are captured in snapshot Until a mechanical shutter is used, electronic shutter creates geometric distortion. Hyperspectral In a single observation a hyperspectral sensor can observe more than 100 spectral bands Covers narrow spectral bands Images are captured in single digital array ensuring larger instantaneous view. Lowest possible geometrical distortion Observations Water is optically active only in a small part of the spectrum. Due to very low concentration of the heavy metals, the bulk of the reflective spectra constitutes of those reflected by suspended sediments and chlorophyll. Spaceborne sensors provides multispectral images with large band gaps and large swath. (Siegmund, Menz, Jensen,2007. Ferrato, 2012)

7 RESEARCH OBJECTIVES AND TOOLS Water quality mapping of large rivers for airborne platforms Cessna Aircraft and UAVs Identifications of major hotspots of water pollution Characterisation of pollutants based on spectral characteristics Tools Cessna Aircraft UAVs (UX5, Phantom 4) Cameras and filters Cessna Aircraft UAV UX5 DJI Phanton 4 Single band cameras Multispectral Camera Hyperspectral Camera

8 DESIGN OF MOUNT FOR THE CESSNA AIRCRAFT Four cameras nearly co-axial Filters: 520, 650, 800, 852 nm Video capture

9 FLIGHT PLAN AND IMAGES Flight path: Bithur to Jajmau (25km) Height: 1100 ft Speed: Variable Image resolution: 1 mp GSD: ~ 1m Filters: 520,650,800,852 (nm) 520 nm 650 nm 800 nm 852 nm

10 Results: Spectral characteristics of different parts of channel Colour Band Red 520 Green 650 Blue 800 Id Class 1 Dense crops 2 Sparse crops 3 Inland Water bodies 4 Dry Sand Bar Id Class 5 Damp Sand Bar 6 Wet Sand Bar 7 Water Body 30 Brightness (DN)

11 Along the channel variation in water quality due to pollutants (394,520,650 nm) Major drain Frame 1 Frame 2 Frame 3 Upstream Flight path Downstream Frame 1 Frame 2 Frame 3 Red Green Blue 520 nm 650 nm 800 nm Three bands have distinct signals, respond to different parameters (Blue metals, greenchlorophyll, infraredturbidity) Similar to Frame 1, a little darker due to difference in illumination Blue band shift, sensitive to metallic ions Infrared signal- two peaks due to turbidity Effect of the pollutants added by the drain

12 Identifying a source of pollution through multispectral camera at Kanpur City Outlet Flight height: 1100 feet (335m) Image resolution: 20cm (Gopro), 40cm(FCC) Bit Depth: 32bit (GoPro 4), 8-bit (Multispectral Camera System) Bands Used: RGB (Go Pro 4), 520,652,800nm and 852 nm (MCS). The 852 nm band did not captured any reproducible image. Base image : Bing World Map Service City Outlet RGB GoPro 4 image FCC Multispectral Camera System (MCS)

13 Identifying source of pollution using a multispectral camera Histogram comparison of RGB and multispectral image (520, 650 & 800 nm band) City Outlet Fresh Water 3 No. of pixels 2. At outlet 3. D/s of outlet Blue band shifting to left, darker, not much metallic ions Green band most affected, chlorophyll concentration may be higher due higher flux of N and P Red band very sharp at the outlet higher particulate matter, and therefore very high turbidity Brightness (DN number) Legend: Red ~ 800nm Band Green ~ 650nm Band Blue ~ 520nm Band * All bands are have ±5nm tolerance

14 Spectral variation in a stretch of stagnant water Input from drain Zone 1 Zone 2 No. of pixels Zone 3 Zone 4 Zone 5 Brightness (DN number)

15 Drone flights: Imaging and Histogram analysis over an outlet in Ganga 1.No Pollution Images from left to right: Original image with inputs from drains; After removing sandbar and then applying histogram equalization on PC1 image; Zonation of the area. Total stretch: 4 km Flight height: 273 m Image capturing: 20 m Camera: RGB PCA analysis of three band data (PC1 image shown) No. of pixels 2.Light mixing 3 & 4. Medium Mixing 5. Highly Polluted Water 6A. Discharge from first drain Legend: Arrows 1.Yellow Primary drain; 2.Green Small drains; 3.Red Medium sized drain 6B. Discharge from multiple small drain 6D. Discharge from medium size drain 6C. Discharge from multiple small drain

16 Comparisons: Cessana and Drone flights 1.No Pollution (DJI) Shivrajpur Kanpur Jajmau Sultanpur 1.No Pollution (Cessna) 3 & 4. Medium Mixing (DJI) 6. Discharge from outlet (Cessna) Cessna images: Red ~ 800nm Band Green ~ 650nm Band Blue ~ 520nm Band DJI images: Standard RGB 3 & 4. Medium Mixing (Cessna) 6. Discharge from outlet (DJI) Histograms of RGB and multispectral image (520, 650 & 800 nm band) * All filters in multispectral camera is having ±5nm tolerance

17 NEXT STEPS Design and customization of the UAV for the Multispectral and hyperspectral imaging payload Multispectral imaging Trace pollution plume back to the source Relate spectal characteristics to specific pollutants Hyperspectral imaging Needs designing of a gimbal to carry the payload High resolution data in narrow spectral bands Should allow better differentiation of specral response of clean and polluted water Ground measurements and community participation Ground sampling and laboratory analysis for validating the UAV based measurements Training the community for water quality measurements using portable kits LALE UAS developed by IIT Kanpur