How To Use the Nexleaf BC System

Overview of Steps

1. Get an account.
2. Select a Calibration
3. Set up a Deployment for your Submissions.
4. Sample air through a filter during air quality measurements.
5. Place the loaded filter on the Nexleaf BC reference card
6. Capture a photo of the card.
7. Create a Submission with the photo
• Alternatively, you can Email the photo to the Nexleaf BC system.
• See our extended Email codes to use.
8. View your data online.
9. Export your data as a CSV file.

• See our Example Submissions and Download section for helpful images, below.
• Experimental Color information is also available for exploration of organic carbon.
• Or visit our Troubleshooting / FAQ section, below.

Detailed Steps

Don't have a login?

• Email us at: info@nexleaf.org and we will send you a login and temporary password (this may take a day or two, we apologize for the delay).

After you log in:

• You will see three new options on the Menu Bar: Deployments, Submission, and Export.
• If you just want to see how the system works, you can start after you log in by downloading an example set of reference card images and submit them for processing.

Get a Reference Card:

• Email us at: info@nexleaf.org and we will send you a Reference Card depending on the filters you will be using:
• 25 mm diameter filters - ask for a Small Radial Card.
• 47 mm diameter filters - ask for a Large Radial Card.
• Larger filter sizes - ask for a Linear Card.
• Each Reference Card will have an ID value that you will enter into the form for online submission.

Different instruments used for measuring BC in aerosols differ in their methods and thus are not easily inter-comparable (see EPA 2012 Black Carbon report to Congress, Appendix 1).

Three Calibrations are available for this image-based analysis, all related to the reference instrument used to originally measure BC.

• BC_TOT: Thermal/Optical Analysis

  Thermal-optical methods are the most commonly used. Since 1982, thermal-optical analysis methods have been applied to measurement of BC of ambient and source aerosols.

  Our calibration is based on the NIOSH TOT protocol.

• BC_Aeth: Aethalometer

  The Aethalometer is a device that measures of the attenuation of specific wavelengths of light through a quartz fiber filter as it loads over time.

  The device is manufactured by Magee Scientific

• BC_PASS: Photoacoustic Soot Spectrometer

  The PASS measures particulate light absorption (and scattering) at different wavelengths in the aerosol, not accumulated on a filter.

  The device used for our calibration is a Collocated PASS 1 (single wavelength) device manufactured by Droplet Measurement Technology.

Calibrations relate the darkness of the calibration square on the reference card to a loading of BC on a filter, as measured in the Nexleaf lab using both a flatbed scanner and a hand-held colorimeter (X-Rite i1Pro). The figure below represents the calibration values for the BC_TOT analysis and the table underneath are the values used.

Deployments are a useful way to organize your BC Submissions into groups.

1. Choose the Deployments menu option.
2. If this is your first Deployment, you will be presented with an empty list of Deployments you have control over.
3. Click on the "Create Deployment" button at the top right.    
4. On the new page, you will be presented with several fields to fill out. Again, the only field required is the Deployment Name.
   • Note: the Deployment Name must NOT contain spaces or strange characters. Keep it simple.
   • Also Note: you can go back and edit the Deployment information at any time.
   • Help about nearly each field is provided with a pop-up by cliking on the symbol. Try it!
5. After filling out the form, hit the "Create Deployment" button at the bottom of the page. You now have one new Deployment that you can start adding Submissions to!

Notes:

• You can set up as many Deployments as you want to contain groups of Submissions.
• The only required value is the Deployment Name. All other values are for your own record keeping.
• Each Deployment can have a set of default values that are applied to Submissions made for that Deployment. But you can easily override the defaults when making a Submission.

Aerosol sampling parameters depend on many factors. Below are some considerations but your sampling needs may be different (for example, if other analyses need to be made on the exposed filters or if certain particle sizes need to be excluded, then please consult other references):

Filter selection

• We have successfully used uncoated (no binder) Quartz Fiber Filters for BC loading measurements. These filters can also be used for Thermo-Optical (TO) analysis done by other labs.
• Although we have the option of submitting samples collected on Teflon filters, these have not been as rigorously tested.
• We do not officially endorse any products, however we have had success with these filters (image on the right):
  • "Pallflex" Air Monitoring Filters. Indeed, the Pall Corporation website has a lot of information about filter selection.
  • Prices currently run about $1.50 USD each filter (e.g., at Sigma-Aldrich)
• If want to do TO analysis (or some other), then you will have to store the filters (and they will definitely need refrigeration). For storing the individual filter samples, one option is the use of PetriSlides (model Z355445) from Millipore (SIGMA-ALDRICH is the vendor).

Filter loading

• Loading of black carbon on the filter should be above 5 ug cm^-2 and below about 25 ug cm^-2 for the highest accuracy. Values above or below this range can be analysed but are more prone to error.
• If you don't have a feeling for this, the filter should have enough BC on it to appear gray but not very black. Too light and our system will most likely over-estimate the BC and too dark and our system will most likely under-estimate the BC.

• The calculation to determine loading thus involves estimating or knowing:
  • The average* ambient BC concentration in the air (ug m^-3) during the sampling period.
  • The flow rate (liters/minutes) of the sampler.
  • The duration of sampling (minutes).
  • The filter loading area (cm²).
• For example:
  • Using a filter that is 3.14 cm² in area (a 24 mm filter with a loading radius of 1 cm):
  • If the 24-hour average* BC concentration in the air is 100 ug m^-3:
  • Sampling at 300 cc per minute for 24 hours will load the filter to 13.8 ug cm^-2
  • Sampling at 500 cc per minute for 24 hours will load the filter to 22.9 ug cm^-2
• Thus, you need to do some calculations to determine the flow rate, duration, and the filter size needed in order to get the loading to within our detectable range (between 5 - 25 ug cm^-2) assuming some sort of expected average* concentration of BC in the air (ug m^-3).

• Average* BC concentration may be difficult to estimate: for example, if the BC in the kitchen during cooking for 1 hour is expected to be 1000 ug m^-3, then for the rest of the 24 hour period the BC needs to be less than 58 ug m^-3 to not overload the filter assuming a 24-hour sample with the pump running at 500 cc min^-1 and a filter size of 3.14 cm².
• Note: the filter loading area (the dark circle on the filter that has collected BC) is NOT the total filter area.

Don't have a sampler?

Although we cannot officially endorse any commercial sampler, we do have some experience with:

• The SKC "Universal Pump" PCXR8, which runs about $1,300 USD. This is a very nice sampler.
• The "Air One" TI-004, sold by numerous outlets including Zefon International, which runs about $200 USD. It appears stable at flow rates above 300-500 cc per minute, however the manufacturer does not recommend a flow rate this low.
• The GEO3 Project has a BC sampling and assessment system that is for educational purposes. We have tested out their pump (sold separately from the Black Carbon Experiment Package, about $200?) and it seems quite stable at low flow rates.

We have no hands-on experience with these (and there are many more out there than just these):

• The Gilian Air Sampling Pumps by Sensidyne.
• The A. P. Buck set of indoor air samplers.

Samplers should have a constant flow rate and/or continuously log the flow rate. Additionally, you will need to know the exact duration of sampling to calculate BC concentration in the air.

Note: If you have a pump that does not sample stably at the required low flow rates, then a "Low flow adapter/constant pressure controller" is required. A description of a simple low flow adapter is here (about mid-page). Alternatively, the Gilian company offers a "Dual Port High/Low Flow Manifold", and at Buck they offer a simple flow splitter (that does not control for changes in flow rate between the two sides as the filter accumulates particles) that may be useful under certain conditions.

If you have experience with other pumps that you can recommend, please let us know!

• The filter should be placed on the white circle in the black field below the QR code for the linear reference card, or in the middle circle for the radial reference card.
• Our reference cards are created for filters that are 25 mm and 47 mm in diameter.
• Please see below in the Download section for examples of filter placement.

• For filters larger than 47 mm in diameter (that would use the large radial Reference Card), or for square filters, you must use the Linear Card with the filter covering the black area away from the QR code and then select "Used Fixed Detector" when making a Submission.

Photos should be taken indoors!

Because photos of the filters and cards needs to be under relatively uniform lighting, outdoor locations are not suitable for good photos (unless the day is very cloudy and you can guarantee uniform light). We recommend taking photos indoors to assure that lighting conditions are controlled.

Do's

• Do take the photo indoors away from windows and under uniform illumination.
• Do fill the frame of the photo with the reference card
• Do take the photo straight down rather than at an angle.

Do Nots

• Do not take pictures in direct sunlight or near direct sunlight.
• Do not use a flash.
• Do not take pictures near windows or open doors.

Please see below in the Download section for examples of photographs.

The core of any Submission is the photo of the filter on the reference card.

1. Choose the Submissions menu option.
2. If this is your first Submission, you will be presented with an empty list of Submissions for this Deployment.
3. Click on the "Create a New Submission" button at the bottom left.    
4. On the new page, you will be presented with several fields to fill out. Again, the only field required is the image of the reference card with the filter.
   • Note: you can go back and edit the Deployment information at any time.
   • Help about nearly each field is provided with a pop-up by cliking on the symbol. Try it!
5. After filling out the form, hit the "Submit for Processing" button at the bottom of the page. You now have one new Submission for your Deployment!

Notes:

• You can make an unlimited number of Submissions under any Deployment.
• Multiple photos can be submitted at one time, which will result in multiple Submissions being created, one for each photo.
• For your records, you can add more information to each Submission, although the only required bit of data is the photo.
• Default information is "inherited" from the Deployment that you are submitting under. But you can over-write that information at the time of submission.
• The filter loading radius (the dark circle on the filter that has collected BC) is the value to enter in the form under "Sample Filter Load Radius" to calculate BC concentration, NOT the actual filter radius.

New submissions uploaded through this website are processed every 10 minutes (each submission may take a few minutes, depending on the size of the image). Submissions that are emailed (see below) may take up to 20 minutes for processing.

Alternatively, the photo can be emailed to our system for processing using the special email you are provided when setting up a Deployment.

The email address

• Each Deployment has a unique email address, based on the Deployment name.
• You can find the Deployment email address by expanding the details of a Deployment on the Deployment list page by clicking on the symbol.
• You can also edit the email address to something you can more easily remember by editing the Deployment.
• Note: any new email address must not contain spaces or strange characters.
• Like submitting through the web, you can submit multiple photos in an email.

See below for special codes you can include in your email to add metadata to your submissions.

New submissions uploaded through this website are processed every 10 minutes (each submission may take a few minutes, depending on the size of the image). Submissions that are emailed may take up to 20 minutes for processing.

If you just created your Submission, chances are that the results are still "pending".

In about 10 minutes, the Submission will be processed and you will have your results.

• If you wish to edit any information contained in the submission, you can click on the "View/Edit" link to the right of the Submission under "Actions".
• Updated information will require you to re-send the data and you will have to wait 10 more minutes for the re-submissions to start processing and the new results visible (the BC loading results should not change but the BC concentration calculated may).

You can export Deployment and Submission data to a comma-separated ascii file (CSV).

1. To export your data, click on the Export menu option:
2. Choose either the "Export All Deployments" button at the top right of the page or choose the "Export" link under the "Actions" column next to each individual submission.
3. A CSV file will be automatically generated and downloading will begin.
4. Only processed Submissions will be exported.

Details on email Submissions

Emailed submissions can include codes to make submission from cell phones easier.

• None of the codes are required to make a submission by email, they just add information.
• Submission values (such as Flow Rate or duration) can be indicated in the body of the email by using the special codes below.
• Any information in the Subject line will be added to the "Notes" section.
• Any additional text in the body of the email, if not preceded by a special code, will be ignored.

Submission codes are a word or letter followed by a colon and then followed by the value desired.

For example, "duration: 24" sets the duration of sampling to 24 time units. Do not use quotes around these codes and do not include the braces ("[" or "]").

Example photographs and Results

• Below are links to various photos of reference cards with filters. You can download the originals and then submit them to our system to test out how things work.
• Right-click or Command-click on the text link to the left or on the image to save to your computer. Then submit as part of a BC Submission as described above.
• Example results of analysis are also provided by the link to the right of the photo.

Best possible photo with filter. BC loading result: 3.889 ug cm-2.		Link to Results
Good possible photo with filter. BC loading result: 9.311 ug cm-2.		Link to Results
Not great photo with filter, not great results. BC loading result: 0.532 ug cm-2. Anything less than 5 ug cm-2 is suspect. Note: Look at the Results page and the "debug image" to see that the calibrator was not sampled correctly.		Link to Results
OK to include sample information in the photo, as long as the reference card is large. BC loading result: 3.581 ug cm-2. Note: There is a light gradient across the card (lighter at one end than the other) which can adversely affect results.		Link to Results
OK to use over-sized filters, as long as the QR code and chart are visible AND "Use Fixed Detector" is chosen for the Submission. BC loading result: 9.233 ug cm-2.		Link to Results
Bad filter. Irregularities on the filter will prevent accurate readings. BC loading result: 5.165 ug cm-2.		Link to Results
Bad white balance. White balance of the camera is off, resulting in slightly strange results. BC loading result: 0.559 ug cm-2. Anything less than 5 ug cm-2 is suspect.		Link to Results
Bad glare on the card AND punches from the filter may interfere with the system. BC loading result: Failed to process.		Link to Results
Best possible photo with new radial card that may be less susceptible to light gradients (but cannot be used with larger filters). BC loading result: 4.049 ug cm-2.		Link to Results

Experimental Color Analysis

Card Color Correction

• Cameras sometimes introduce false colors in order to "color balance" the ambient light.
• Our Reference Cards have known color values that are used to "reset" the color of the entire image to something closer to the true values of the Reference Cards, what we would expect.
• We use a combination of "gray world" assumption adjustment, the "histogram stretch", and a graypoint adjustment, although the gray point is slightly colored.
• Color adjustment can improve the color cast of a photo but can also go awry and lead to undesirable colors, thus this adjustment is experimental and should be regarded with caution.
• RGB colors have been rounded to integers before export to CSV; sampled values are assumed to be sRGBB with an illuminant of "D65".

The color of BC loading results.

• Black carbon is assumed to be colorless, absorbing light across all visible wavelengths.
• Contaminants, such as dust or Organic Carbon, may be colored and this color may reveal some information about the source of aerosols collected.
• We offer pre- and post-color correction color values of the filter for correlating with sources.
• We display some of these color analysis results on a graph available on the Submissions page to suggest interpretation of colored values.

The Lab color space.

• The CIE Lab color space is designed to approximate human vision.
• We use the Lab color space to remove the lightness (L) component of the image of the filter and examine only the color components (the "a" and "b" aspects of the color space).
• The figure to the right shows the colors associated with the lightness (L) component equal to 75% and the "a" and "b" varying from -30 to +30 around a neutral gray at 0,0.
• The Lab results from Submissions are presented on the Detail page of each submission and are exported with the rest of the data.

The Lab color space, Organic Carbon, dust, and other "contaminants".

• Results from data sets shared with us indicated that different sources of aerosols have different colors associated with them.
• It may be possible to interpret color as indicating the presence of various other compounds, other than Black Carbon.
• Alternatively, coloration of the filter may be related to the age of the aerosol.
• Below is a chart of some of the data we have available to us, and can be used as inspiration for examining your colored filters more carefully:

Below is a similar graph but with a few single-source samples of aerosols generated in the laboratory (thanks to Dr. Jamie Schauer and his lab!):

Other color spaces we export.

We transform the filter color into other color spaces, which can be useful for "hybrid" or multi-dimensional color analysis, beyond the "a-b", 2-D space presented above. Although all spaces are derived from the RGB digital information, the transformations are non-linear and so may "tease" out more useful information.

• The HSV color space
• HSV is a cylindrical-coordinate representation of RGB. Like the Lab color space, the luminance (V for Value) is separated from the color (H for Hue) and Saturaton (S)
• One problem with the HSV color space is the discontinuous Hue, which ranges from 0 to 360 degrees; two perceptually similar colors can have Hues near zero and also near 360.
• HSV data for the color-corrected sample is exported along with the Lab color transform.
• The CIE Yxy color space
• CIE Yxy, like the Lab color space, was deliberately designed so that the brightness parameter (Y) was separate from the chromaticity of a color (x and y)
• Yxy data for the color-corrected sample is exported along with the Lab color transform.
• The Yuv color space
• Yuv is an analog encoding of color information. It is typically used as part of a color image pipeline and encodes a color image taking human perception into account (similar to Lab).
• One problem with the HSV color space is the discontinuous Hue, which ranges from 0 to 360 degrees; two perceptually similar colors can have Hues near zero and also near 360.
• Yuv data for the color-corrected sample is exported along with the Lab color transform.

Multidimensional color space estimate of Organic Carbon.

• Using a multiple linear regression with log of Organic Carbon as the response variable and all the color components (and log of BC) as independent variables, it may be possible to estimate OC within a deployment location (similar conditions).
• An easy to use Excel spreadsheet for multiple regression is offered here (fun to play with, trust at your own risk): http://udel.edu/~mcdonald/statmultreg.html
• For example, we ran independent multiple regressions for the sample data sets shown above, with the results below. Each resulting equation was different for estimating OC.
• The US EPA regression equation (r² = 0.656.) was:
• log_OC = -652.1069 + 0.31224 * log_EC - 0.509 * L - 2.360 * a - 0.446 * b + 0.6708 * V - 0.1351 * Y - 34.1832 * x + 19.4271 * y + 54.53003 * u

Multidimensional color space to separate sources.

• Using a Recursive Partitioning model (rpart for R), and more specifically a Multivariate Regression Tree (because we are separating multiple groups from each other), separation of sources may be possible.
• For example, we ran a multivariate regression using all color variables from the data set presented above, and found a "tree" for classifying the color results.
• As more data sets are available to us, correlating OC with source or age of aerosols may be possible using color information (see below).

Troubleshooting / FAQ

I don't see my Submission.

1. Refresh the Submissions web page.
2. Check other Deployments that you have access to. You may have accidentally sent the Submission to the wrong Deployment.
3. Check the "missing Submissions" page using the link at the bottom of the Deployments list page. Submissions are orphaned if the Deployment name was not readable for an email submission.
4. Send us an email and we'll see if it got accidentally deleted or some other mishap occurred.

The BC loading results from a Submisson are indicated as too low / high / error.

1. Very low BC loading (less than 5 ug cm^-2) or very high (greater than 25 ug cm^-2) may be less accurate than loadings in the middle of the range. Low or high values are flagged but not necessarily incorrect.
2. The Submission page has links to the debug image and debug chart at the bottom. Look at the debug image and verify that a blue circle containing a smaller a blue square and a blue '1' are centered over the loaded area of the filter in the image.
3. If the blue square '1' does not appear over the filter, consider using the "Fixed Detector" option on the Submission page.
4. Note: the filter radius for loading (the dark circle on the filter that has collected BC) is the value to enter in the form if you want to calculate BC concentration, NOT the actual filter radius. See Step 5 above.
5. Examine the squares where the calibration chart is samples to make sure that there is not a severe light gradient and that the squares are centered on the colored chart.
6. Try re-taking the photo under different lighting conditions and resubmitting.

My Submission has "failed".

1. The photo might not be the correct format.
2. The photo may be dimensionally (not file size) too large for our server (even if it got uploaded)-- try to keep the images around 1024 pixels on a side.
• Please use only: JPEG, GIF, PNG, or TIFF
3. The QR code might not have been read.
• Make sure that the entire QR code is visible in the photo.
4. Make sure you selected the correct Reference Card ID (selecting Radial for a Linear card and vice-versa will result in a failed Submission).
5. Try re-taking the photo under different lighting conditions and resubmitting.
6. Send us an email and we'll see if something weird is happening.

My Submission has "Suspicious Results" or has "Failed" because "Gradient not descending".

1. There is most likely a shadow that is falling across part of the calibration chart that is making some squares darker than expected.
2. Shadows over the calibration part of the card but not the filter will result in inaccurate or poor results. Example of a shadow on a linear card:
   
   
1. The white area of the card at the top right is a red rectangle and the labels "dark" and "light". This area should be uniformly light (it is white and not colored). This indicates a shadow occurring on the right-top side of the card.
2. A blue crescent has been placed over the area that there is a shadow falling on. This shadow covers some of the calibration squares but not all. This shadow also does not extend to the filter and so the filter is lighter than a set of the calibration squares used to measure the filter color.
3. Numbers on the calibration squares correspond to the red pixel value. These numbers should be decreasing from one to the next by about 13 pixel values (thus, for example, from 100 to 87 to 74 to 61, etc.). The numbers on the left of the card (220, 217, 200, 182, 155, 128) continue to the column to the right but should continue to drop (and they do drop, from 134 to 118, 102, 81, 59). However, the transition from 128 to 134 should have been from 128 to about 115 but instead it is 134, which is higher, not lower. This is because the shadow is covering the 128 square but not the 134 square, making 128 darker than it should be and 134 is lighter.

---

I accidentally deleted a Submission.

• Send us an email and we'll see if we can retrieve the Submission for you.

---

I want to remove a Deployment from my list.

• Send us an email and we can delete a Deployment and all the associated Submissions for you (or we can move the Submissions to a different Deployment).

---

Hint: To reduce work when making submissions within a Deployment that have different flow rates and durations...

• If the submissions share even the units for Sample Flow Rate (cc/minute or liters/minute) or units for Sample Duration (minutes or hours), then set those as Deployment defaults.