Questions About Using Geopsy

[schmanse]

Hi Marc,

I am currently using Geopsy for HVSR analysis and have several questions regarding its use.

1. I wanted to assess whether the analyzed data meets the SESAME criteria (e.g., whether the frequency with an A/2 value falls within a specific range) and tried to download the graph. To do this, I followed the process: H/V Results -> Properties -> Layer -> result::Average -> Actions -> Save, and saved it as a .txt file for use. Is this the correct method to obtain the information I need?
2. In general, I understand that the frequency (f) with the highest A value is considered f₀. Why do the A_peak values obtained through the mentioned process sometimes differ from the f₀ and A₀ values shown in the Results?
3. Based on the above, what is the difference between A₀ and A_peak?
4. The SESAME criteria include "Reliable" and "Clear peak" criteria. Are these empirically defined? If they are empirical, could there be limitations when applying them in other countries?
5. When the "Overlap" option for window length is enabled, the standard deviation either increases or decreases. Under what circumstances does each occur?
6. To reduce the standard deviation, how should STA and LTA be adjusted in general practice?

Best regards,
Shin

[admin]

Hi Shin,

I wanted to assess whether the analyzed data meets the SESAME criteria (e.g., whether the frequency with an A/2 value falls within a specific range) and tried to download the graph. To do this, I followed the process: H/V Results -> Properties -> Layer -> result::Average -> Actions -> Save, and saved it as a .txt file for use. Is this the correct method to obtain the information I need?

It is better to select the HV plots you want to export and go to the menu "Tools/Save results". The exported text file contains more information.

2. In general, I understand that the frequency (f) with the highest A value is considered f₀. Why do the A_peak values obtained through the mentioned process sometimes differ from the f₀ and A₀ values shown in the Results?

What is A_peak? I can't find this quantity in the SESAME criteria.

In the header of the results you will find f_{0 from average} and f_{0 from windows} is computed by considering an admissible range around f_{0 from average}. Each time window is considered individually and a peak is identified within the range around f_{0 from average}. The list of peak frequencies for each time window are then used to compute statistics about f_{0 from windows}. f_{0 from windows} is not considered at all by the SESAME criteria. The condition iv) where f_peak appears is just that a peak must appear on the one standard deviation curves within 5% of the peak observed for the average curve.

3. Based on the above, what is the difference between A₀ and A_peak?

I can't find this quantity in the SESAME criteria.

4. The SESAME criteria include "Reliable" and "Clear peak" criteria. Are these empirically defined? If they are empirical, could there be limitations when applying them in other countries?

The definition of a clear peak is a bit arbitrary. Consider these criteria as indicative. The database used to infer these conditions was the databases available twenty years ago during SESAME project. HVSR is a simple method with usually complex results. If the investigated site is 1D and with a clear contrast, expect a nice and stable (in time and in space) peak. For other situations, expect lateral variations of the results and instabilities over time (depending of source directions). Site characterization must not be limited to HVSR for complex situations: ambient vibration arrays, MASW, borehole and other geophysical experiments should be combined to provide a reliable image of the ground structure.

5. When the "Overlap" option for window length is enabled, the standard deviation either increases or decreases. Under what circumstances does each occur?

Overlap should be used only if the amount of available data is small, to produce more time windows. A high overlap forces the re-use of the same data, it should artificially decrease the standard deviation. Can you quantify the effect? Is it a significant variation? This is part of the epistemic uncertainty of the results.

6. To reduce the standard deviation, how should STA and LTA be adjusted in general practice?

Choices for LTA and STA are subjective. If you suspect parts of the signal to be affected by close and transient sources (e.g. traffic), STA and LTA can help to discard some parts of the signal. A good test is to process the rejected time windows only (invert window selection). Does it produce significantly different curves?

[schmanse]

Hi, Marc,

Thank you for your response.
It seems that many of my questions have been resolved.
However, I think some of my previous questions might not have been clear enough, so I’d like to ask for clarification once again.

[Regarding Question 1]

When I saved the data via the mentioned path, I got a log file and an hv file.
Can I open the log file as a .txt file to check the data?

When I opened the log file, it contained:
- Each software version
- Signal used
- Parameters used
- Information for each window.
However, there were no graph coordinates included, so it seems difficult to check the Criteria for a clear H/V peak (Conditions 1 and 2) from this file.
How would you recommend verifying Criteria 1 and 2 in this case?

[Regarding Question 2]

I understand that f₀ {from average} is calculated as a statistical value derived from f₀ {from windows}, which are values selected based on specific windows.

So, does this mean that f_peak is the frequency corresponding to the peak value observed in the average curve, and this value should fall within the "gray area" of the HVSR curve?

Additionally, when I visually inspect the HVSR curve, the frequency at which the average curve has its maximum value is not always the same as f₀.
In such cases, how is f₀ determined? Could you explain the detailed process?

[Regarding Question 5]

If I suspect the signal is being affected by transient sources, is it better to increase the STA duration?
Or would adjusting the Min STA/LTA or Max STA/LTA values be a better approach?
Would a parametric study be helpful in this case?

[Additional Question]

Since the software was developed as part of the SESAME project, is there a built-in feature to automatically verify whether the results meet the criteria mentioned in the SESAME guidelines?

The attached file shows cases where the "intuitive f_peak of the average curve" and the "final selected f₀" differ, as I wanted to understand the criteria for selecting f₀ in Geopsy.

Best regards,
Shin

[admin]

1. The curves are stored in the .hv file. First two criteria can be easily checked from them.
2. f₀ {from average} is the maximum observed on the average curve. f₀ {from windows} (the gray bands indicate the average and the standard deviation) is computed for each individual time windows and summarized by an average and its standard deviation.
3. A parametric study is necessary. No definitive recommendation have been issued until now. It is too case dependent.
4. During SESAME, my main interest was the inversion of dispersion curves from array measurements. I added H/V to geopsy at the end of SESAME. The official piece of software produced by SESAME was JSESAME which probably includes these criteria. Deciding if a peak is good or bad remains still a bit arbitrary. I'm not sure if these criteria were built after a quantitative analysis. They mostly reflect what's an expert would define as a good peak. A peak that does not follow these rules but that can be observed in many places and stable over time is also a good peak. Whatever the shape of the H/V curve, it is a signature of the underground structure and of the wavefield.

[schmanse]

Hi Marc,

Thank you for your excellent response. However, I still have some unresolved questions, so I would like to ask further:

1. You mentioned that the 1st and 2nd criteria of a clear peak can be verified, but there is no indication for f0/4, 4f0 or A0/2. How can these values be identified?

2. Is f_peak the point that is intuitively identified on the average HVSR curve?
There are many cases where f0 and f_peak are different. In such cases, what is the selection criterion for f0?
As far as I know, f0 is the frequency at which A0 is at its maximum, and I understand that f0 should not differ from f_peak

3. Therefore, I would like to understand the machanism by which f0 is calculated in Geopsy.

4. After saving as an hv file, should I load it in Geopsy by going to the results window and selecting Tools -> Load results?
Could you confirm if the results obtained in this way are displayed as shown in the attached file?

Thank you again for your kind response.