Public Lab Research note


Water Quality Measurements with Optical Spectroscopy

by philippg | July 10, 2013 20:15 10 Jul 20:15 | #8809 | #8809

What I want to do

I'm a PhD student working on water remote sensing. In water remote sensing, or 'ocean color', people extract water quality parameters from satellite imagery. Probably the most interesting parameter is chlorophyll-a concentration, which can be related to the phytoplankton (algae, cyanobacteria) biomass in the water. Applications are quite diverse, ranging from harmful algal blooms early warning systems, eutrophication assessment to climate change research - to name a few.

The same algorithms are used on spectra taken directly in the field to calibrate/validate remotely sensed concentrations. However, spectrometers used for these field measurements are quite expensive (>20k USD) and often not trivial to operate. This hampers the use of 'water color' by a wider audience, e.g. water managers, fisheries and especially the public that might want to know what the water quality in their pond is.

Spectra from cheap, open source spectrometers can potentially bridge this gap.

Challenges

From my experience with the foldable spectrometer and the desktop kit I see three main issues: sensitivity, spectral resolution and calibration.

  • Sensitivity Water appears often very dark, especially if only little scattering substance is present. In contrast, the reflection of the direct sun light on the water surface is extremely bright. Thus, to resolve the 'true' water color, the instrument has to be relatively sensitive.

  • Spectral Calibration and Calibration Pigment absorption, other optically active substances and the water itself cause the water's color. Pigments in phytoplankton, such as the main photosynthetic pigment chlorophyll-a, have very sharp absorption features. In order to resolve those peaks, the spectral resolution has has to be sufficiently high and the calibration sufficiently accurate.

First Experiments

These days I had the opportunity to take the desktop kit out on a fieldwork campaign on Lake Peipsi and Lake Vortsjarv in Estonia. Fortunately, this little Baltic country has one of the best 3G-networks worldwide and so I could use the spectral workbench to upload my spectra straight from the boats - pretty cool! However, an offline version of the workbench is essential (couldn't get the local webserver running so far). The spectra are available here: Lake Peipsi (R_sky, R_water), Lake Vortsjarv (R_sky, R_water).

First Conclusions

  • Preparations / Calibration The initial wavelength calibration with fluorescent light line peaks is pretty brilliant. Still, at least a rudimentary intensity calibration is necessary to interpret the shape of the spectra. I can use our calibration lab for this purpose and give some feedback to the community on how the spectral response curve of at least my desktop kit's webcam looks like. For public use, I'm thinking of using daylight spectra as a reference, as the shape is rather stable (not the intensity, due to atmospheric conditions). Maybe we can find a better solution for that.

  • Measurements We need at least two measurements in order to run a spectral unmixing algorithm: upwelling radiance (light from the water) and downwelling irradiance (complete skylight). As we most likely won't be able to build a cheap spectrometer that has even vaguely defined entry optics (e.g. 9deg field of view for the radiance and a perfect cosine response for the irradiance), some improvisation is needed. For the irradiance measurement I'm thinking of using a white table-tennis ball on top of the slit as a diffusor. The radiance measurement is mainly hampered by water surface reflections. To avoid those, I'd like to measure either just underneath the water surface (--> how to make the spectrometer water tight) or to use a sun shade (such as for camera lenses). For now, a black bucket with a hole in the bottom should do the job.

  • Postprocessing The current procedure to extract a spectrum from the webcam-video is pretty smart and straightforward but probably not optimal if sensitivity is a priority. Currently, as I understand, only one row of the 'stitched spectral image' is used to extract the spectrum ('set sample row' in the workbench). Skylight, as well as the water leaving radiance are stable on the timescales of a measurement. Therefore I'd suggest to average over all rows to improve the signal to noise ratio. If that is not enough, one could think about extracting not only one line from the webcam's video stream but e.g. ten and save the average in the stitched spectral image. In a last desperate step, one could use the whole webcam image and correct for the curvature caused by the DVD, however, I don't think this will be necessary.


I did this Help out by offering feedback! Browse other activities for "spectrometry"


People who did this (0)

None yet. Be the first to post one!


11 Comments

Hi Philip, This is very interesting and timely. California is now experiencing a huge boom in Hydraulic Fracturing, a practice that employs water and chemicals injected into oil or gas wells to release trapped oil and gas in shale rock layers. Based on much publicity about horror stories of contaminations in ground waters and air at thousands of fracking sites with longer history, such as Wyoming, Pennsylvania, New Jersey, and Texas. People are truly concerned. Wonder if your method can be used to monitor water quality of ground water, such as in wells, or coming out of faucets? My startup, WattminderInstruments.com is building a very low-cost platform to facilitate area, monitoring of contamination of groundwater, as well as air where fracking operations exist. We'd love to collaborate with you and any organization with similar interest.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


hey wattminder,

what kind of contaminants can be expected? if they have a specific optical signature, then they can be detected with optical spectroscopy. technically, analyzing samples in transmission is quite different to acquiring reflectance spectra. you could either use a known, artificial light source or daylight to illuminate your sample; the later would keep the device simple and cheap. it really boils down to the type and concentration of pollutant you are looking for. from my side, i think it's definitely worth a try!

cheers, philipp

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


this is cool. I'm in the Gulf Coast of the United States. We have a large algal bloom that happens every year, as well as a lot of nitrate waste problems. I'm looking forward to reading what you discover!

Reply to this comment...


wow, what a great sounding project Philipp. I like your thinking. Taking an average over 10 or more rows of the camera image is really good.

The issue of calibration is the area we are trying to resolve, with many suggestions on the plots-spectrometry forum. I'm guessing here that the normal technique we would use, of getting a baseline measurement from an uncontaminated source, such as distiilled water in a cuvette, would need to be extended.

How much water depth (or column depth) is typically analysed when looking for the chlorophyll-A using absorbance ? Is it in the order of centimetres or metres ?

Would a calibrated filter work to narrow down the problem of the source calibration ?

Are you looking at a specific wavelength for the absorbance or a range ? I'm guessing you are looking at absorbance at some other colour than green.

Forgive the questions, but water quality monitoring around these parts is rather important, the school I teach in runs a small aquaculture set up and getting a handle on the build up of algae would be excellent. I would also be interested in seeing if it could be tied in to pre-emptive chlorination of swimming pools, picking the (exponential ?) rise in algae as an indicator it is time to chlorinate, rather than rely on the measured value of the chlorine which is wasteful.

great research note, thanks heaps !

stu

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


I volunteer with LMVP (Lakes of Missouri Volunteer Program), and they conduct water quality monitoring. Their algae sampling is quite formal, where filtered, freeze-dried samples have the chlorophyll extracted and fluorometry is used to measure the chlorophyll. While the collection and freeze-drying is done by the volunteers, the analysis is done at a lab. I am interested in seeing what can be done on a local basis, with simple equipment. Looking at commercial probes, I do not see that they are that far different from the desktop spectrometer. LEDs in the 420-470 nm wavelength are easy to obtain. Have you looked at anything in this direction?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


@straylight: 'How much water depth (or column depth) is typically analysed when looking for the chlorophyll-A using absorbance ? Is it in the order of centimetres or metres ?' This really depends on the specific water body. During heavy blooms, the first optical depth can be as low as centimeters. Usually we have sth. on the order of meters in the blue/green. We calculate absorption, not absorbance.

'Would a calibrated filter work to narrow down the problem of the source calibration ?' Hm, no i think the wavelength calibration with fluorescent light gives us really narrow peaks already. We would rather need a source with a stable intensity profile, or at least stable ratios. It's tricky, calibration lamps are far from being affordable. I guess the task is to find a cheap compromise.

'Are you looking at a specific wavelength for the absorbance or a range ? I'm guessing you are looking at absorbance at some other colour than green.' Yes, ratio algorithms are the usual approach. This is a good introduction. However, there are also some algorithms that use the full spectral information, e.g. some neural network approaches. Blue/green ratios are working well in open ocean waters, where all other optically active components are driven by the phytoplankton biomass. In coastal areas and inland waters, the situation is more complex and more sophisticated spectral-unmixing algorithms have to be applied.

'Forgive the questions, but water quality monitoring around these parts is rather important, the school I teach in runs a small aquaculture set up and getting a handle on the build up of algae would be excellent. I would also be interested in seeing if it could be tied in to pre-emptive chlorination of swimming pools, picking the (exponential ?) rise in algae as an indicator it is time to chlorinate, rather than rely on the measured value of the chlorine which is wasteful.' Cool idea! The issue with swimming pools is that you usually get a significant amount of bottom reflectance into your measurement. This makes the whole radiative transfer much more complex, as many of the assumptions for analytical solutions are not valid anymore and you need to know your pool's bottom albedo. It'd rather measure samples in that case.

@webbhm: lab analysis of water samples is still considered the most accurate measurement (even though, there are many uncertainties involved as well). usually chlorophyll-a concentration is determined with absorption measurements of filter-pads. fluorometry I mostly heard of in combination with 'living' samples/cells. this project uses in-vivo fluorescence, measured with open source components.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Instead of a calibrated light source, what about scanning a sample and non-sample simultaneously, so that you can actually measure the input light and sample with the same conditions? The slit is wide enough that you should be able to get 2 readings at once, just cover half of the slit with the sample... just an idea.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


@warren: as sensitivity is my main concern, i'd really avoid masking even more intensity. i'm currently experimenting a bit with possible diffusors, e.g. baking paper or a table tennis ball in front of the slit, which all dampen the signal even more.

Reply to this comment...


I ended up trying some semi-opaque plastic but indeed the light has to be really strong as a result: http://publiclab.org/notes/warren/07-31-2013/modifications-and-tweaks-for-flare-spectrometry

Potentially you could use some kind of concentrator like a reflector from a lamp to get more light in there... just thinking out loud: http://www.amazon.com/Bayco-SL-300N4-Clamp-Aluminum-Reflector/dp/B007RKKEHA/

reflector

Reply to this comment...


@wattminder - have you seen some of the recent work with the Riffle, for conductivity monitoring of water? http://publiclab.org/wiki/riffle

Also folks are trying to make a "most wanted list" of water contaminants: http://publiclab.org/wiki/common-water-contaminants

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


How did you resolve these issues?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Login to comment.