Metadata, station magnitude contributions, gain(?) issues

[fmassin]

Hello,

While processing the Haiti earthquake with seiscomp, I noticed significant metadata issues on some (E,S)HZ channels. I attach issue seen with mB but similar things can be observed with other M types (in a smaller proportion of affected stations since mB seems to have a wider range of application).

Any thoughts or help?

Cheers

Fred

[ivor]

hi fred,

thanks for the information, we will have a look and see what we can find that might explain your results.

cheers,
richard

[fmassin]

Thanks for looking into it.

[ivor]

hi fred,

looking into this a little deeper and the source of the problem quickly became obvious. all of the stations in your plots have their gain set to 1 since these are units which have geophones (or some other type of seismometer) that is not known to us. since we don’t know the actual recording meter’s details, we are unable to provide the gain and overall sensitivity of the unit in the response file.

if you want to use a Shake station as part of magnitude calculations, or anything else where the gain necessarily needs to be accurate, you will need to exclude these types of stations.

you can do this (somehow) by looking at the sensor attribute contained in the stream element of the XML response file. it will have a value looking something like this:

sensor="Sensor-OTHER-VEL"

you can key off the word OTHER for all types of instruments and channels to identify which ones are unknown.

alternatively, you can also reject any channel where the value for the gain element is 1, which looks like this:

<gain>1</gain>

cheers,
richard

[fmassin]

Hi Igor,

I understand, thanks a lot. Isn’t it possible to not provide unknown responses? Just channel and up-stream informations?

Cheers

Fred

[ivor]

hi fred,

i will have a look, but at the moment, the method is to provide a value that is obviously wrong, where this can be used to key off of. providing all meta-data information, but not gain info, could very well break down-stream programs that are expecting it 100% of the time.

cheers,
richard

[fmassin]

Dear Richard,

Thanks a lot for looking into it. I understand and I don’t mean to be annoying. To my personal opinion, a gain of 1 is not obviously wrong until looking into individual magnitude residuals that not all people do that. For example, industry-derived MEM-based accelerometers typically have correct gains of 1… If a gain should be mandatorily be provided for downstream clients, it could be Nan ; and if it should "real"ly be, it could be 0 ?

Cheers

Fred