great ques.

Hi Benjamin,

Hitachi F7000 could be good if the signal of your sample is enough (the signal/noise ratio is not high (SNR=800 http://www.hitachi-hightech.com/global/product_detail/?pn=ana-f7000) for hitachi while the signal/noise ratio is much higher (SNR>20000 http://www.horiba.com/scientific/products/fluorescence-spectroscopy/steady-state/aqualog/aqualog-r-our-compact-benchtop-fluorometer-for-cdom-13031/) and could be raised through increasing the integration time for Horiba Aqualog). So I think it could be better to check if Hitachi is suitable for your groundwater samples. If yes, Hitachi could be a better choice due to cheapness and easiness and also fastness. Aqualog integrates a lot functions in the software Origin (normalization, IFE correction, etc.) and it measures the absorbance at the same time. Aqualog could be very fast (only about 30 seconds) if the signal is high , and the time depends on the signal since the integration time need to be increased if the signal is low. For some low concentration oceanic samples, the integration time could be 10s or higher (the default is 0.1s), the time for one sample could be 100 times of the default (50 minutes).

Best regards

Penghui

Ben,

I've had great experiences with the Aqualog on a variety of DOM (rainwater, snow, riverine, lacustrine, and peat-derived). For boreal DOM, you'll be looking at a probably 2 minutes per acquisition for a large EEM. Plan on having to dilute 1:10 if you're above 30 or so mg L-1 [DOC]. Make sure a daily blank comes along for the ride. 

The IFE correction, etc in Horiba's software is suboptimal for a research environment; I'd recommend exporting to Matlab and performing these corrections offline. There are some excellent papers on this subject, and many labs have code already written. 

The software package that comes with the Horiba is a bit demanding for the laptop that they spec. You're much better of getting a PC with at least 12 Gb RAM. The more, the merrier. 

Finally, for PARAFAC work, PLS Toolbox from Eigenvector is an indispensable tool if you hate working from the command line. It's a solid GUI, and not terribly expensive. The support is actually quite excellent too. 

If you'd like to chat further, I'd be happy to answer any questions you have. 

Best,

Tim

Hi Ben,

We have been using the first-generation Aqualog since 2012 to characterize DOM from a variety of sources (i.e. urban wastewater, groundwater, agricultural runoff, produced waters from oil and gas), and are hoping to procure a new Aqualog with a wider scan range (200-800nm) to develop methods to detect algal pigments.

We use the instrument in the mode to simultaneously collect absorbance spectra and fluorescence EEMs. Scan time is approximately 2 minutes per sample, and we often analyze more than 50 samples per day. All scans are baseline-corrected using organic-free (blank) water. We allow the baseline correction to be automatically applied by the Horiba software. However, all other data corrections (e.g. normalize to daily water Raman, correct for IFE), processing (e.g. multiply up diluted samples, calculate designated slopes and indices), and manipulations (e.g. Rayleigh trims, vectorize fluorescence data, compile summary data) take place outside the Horiba software using Matlab code.

I agree with the above comment by Timothy, purchase a PC rather than the laptop recommended by Horiba. I've also found Horiba technical support to be outstanding.

If you end up going with the Aqualog and have operating questions, let me know. I’d be happy to discuss. We’re also really close to publishing a formal SOP…https://www.researchgate.net/project/Procedures-for-the-measurement-of-absorbance-and-fluorescence-from-dissolved-organic-matter-using-the-Horiba-Scientific-AquaLogR-Fluorometer.

Cheers!

Angela

Hi Angela, Tim, and Penghui,

Thank you all for the excellent information. This really simplifies our decision process and will help us get up and running soon.

One follow-up question, that you might have already answered. Horiba proposes a bundled version of EV-Solo Eigenvector software. Is that needed to export the files or can all the post-processing be done in Matlab/R with the raw files from the instrument controller/software?

Best,

Ben

I'm not familiar with EV-Solo Eigenvector Software. It looks like it's a stats package, which you might want for data interpretation. We use JMP 12.2 (SAS Inst.) for PCA, DA, PLS, etc...

We're able to export the raw files just fine without the Eigenvector Software add-on.

Angela

Don't bother with EV-Solo. Rather, get the full PLS toolbox from Eigenvector. This is a matter of personal preference, but I like the way in which the GUI works. The full package also allows you to work from the command line and to import toolboxes from the Matlab community. 

For making beautiful EEM graphics and doing some really basic regional integration, OriginLab (Origin Pro) is handy...but wildly expensive if your university doesn't carry a floating license. 

Cheers,

Tim

Whatever spectro you use, I am interested to integrate it to my R package (https://cran.r-project.org/web/packages/eemR/index.html). Please contact me if you want!

Thanks everyone. We are setting up our fluorometer next month and I will be in contact. 

Tim, we have an Aqualog fluorometer, and I an trying to export the EEM files to the PLS Toolbox, but I don't know how. The Aqualog software produces a table with excitation wavelengths as rows, and emission wavelengths as columns (or viceversa, I don't remember now), and I can export it either as a DAT or a CSV file. Can I use either of thiese files in PLS?

Thanks!

Gerardo

Gerardo,

You'll wan't a .csv for PLS Toolbox. It's best if you import a "stack" of .csv's, as Matlab is then able to collate them into a proper dataset. To do this, you'll first export from Originlab (or Aqualog's native software) into your .csv format: one at a time, or write a script to do it for you. We prefer to store these in files based on the date of acquisition to post-process (Raman normalize, IFE correction, etc) using our own scripts.

Next, in Matlab, you'll open PLS Toolbox. From there, you'll select

File-->Import Data-->Text File-->

A browser window will spawn, navigate to your files, highlight them, hit enter. Now you'll be prompted to tell PLS Toolbox "how" you want the files loaded. Since these are pretty simple files, you may choose to select "graphical" wherein you specify that the first column is an axis variable (absorbance) and the first row is an axis variable (emission); I prefer to select "strict, fast" parsing. It seems to be robust to any weirdness that happens behind the scenes during the initial conversion to .csv.

The final, and most important, box to check is "auto build three-way arrays." This ensures that you end up with a dataset in which each EEM is a page in a book, rather than appending each EEM onto a single plane. If the import process hangs up or returns an error, it may be that your .csv's are a different size (can't composite EEMs with different XY dimensions), or you perhaps exported the wrong tab from the Aqualog software. Finding which one is the culprit will be an iterative process.

Finally, before you try to do any multivariate work with your new stack of EEMs, it's best to shuffle the data. This way, the model will have much better odds of finding the true best-fit, rather than a local best-fit. This will also ensure that your split-half validation is balanced--this is critical to proper validation.

Last I checked, this line of code worked in Matlab to accomplish this feat:

A = EEMs_2013_2014_Combined1;

n = 480;

ix = randperm(n);

Ashuffled = A(ix,:,:);

Where the compiled dataset I wanted to shuffle was called "EEMS_2013_2014_Combined1"

and it was composed of 480 separate EEMS.

The resulting "Ashuffled" dataset file is the one used for modelling.

Hope this helps!

Tim

Thanks a lot! I will try it.

Gerardo

Thanks a lot for all these intersting information.