obspy remove_sensitivity problem with the tsunami monitoring gauges after the FDSN station upgrade to StationXML1.2

[elidana]

Reporting here an issue found by one of our users.

This has been noted after the recent upgrade of the FDSN station webservice to deliver StationXML 1.2 (see https://github.com/GeoNet/help/blob/main/data-access-changes/fdsn-station.md for details)

I am trying to plot buoy data from GEONET. The following code snippet worked a few months ago:

from obspy.clients.fdsn import Client
from obspy import UTCDateTime

client = Client("GEONET")
starttime = UTCDateTime("2025-07-29T20:00:00.000")

endtime = starttime + 3600*24

network='NZ'
station='CHIT'
location ='41'

st = client.get_waveforms(network=network, location=location, station=station, channel="LTZ",starttime=starttime, endtime=endtime,attach_response=True)

st.remove_sensitivity()
st.plot()

The error I am getting is:

AttributeError: 'NoneType' object has no attribute 'value'

and the data are not converted from V to m.
This can be seen if you print
st[0].stats.response

Any idea what I am doing wrong?

[elidana]

@Thomas-Benson confirmed that the update in stationxml version appears to be what broke your code.

Between stationxml 1.0 and 1.2 there are several changes to Polynomial response types, since stationxml1.0 only implemented relatively stripped down version of polynomial responses.

It looks like the problem you are encountering is a problem with how obspy reads the new polynomial responses, since it appeared to have the same problem when I checked with the example polynomial response provided by FDSN

https://docs.fdsn.org/projects/stationxml/en/latest/response.html#setra-270

Channel Response
    From UNKNOWN () to UNKNOWN ()
    Overall Sensitivity: UNKNOWN defined at UNKNOWN Hz
    3 stages:
        Stage 1: PolynomialResponseStage from mbar to V, gain: UNKNOWN
        Stage 2: ResponseStage from V to V, gain: 1
        Stage 3: CoefficientsTypeResponseStage from V to count, gain: 51

based on the obspy release log, recent version should account for StationXML 1.2 (see https://github.com/obspy/obspy/releases, starting from v. 1.3.1)

what Obspy version have we been using to verify this issue?

[Thomas-Benson]

I am using obspy 1.4.1. Also it looks like the change in polynomial responses was introduced in stationxml 1.1 which should be earlier still.
It also looks like the remove_response function (which is the more complete version of remove_sensitivity) doesn't handle polynomial responses either, it just raises NotImplementedError: PolynomialResponseStage not yet implemented. Please contact the developers. 🫠

[Thomas-Benson]

I found another workaround, it's not particularly elegant but it works on attached responses without needing to download a seperate file. I read through the old stationxml 1.0 encoder and copied the simplified gains into python. This should add enough info to a streams attached response to use remove_sensitivities, although I'll leave removing the full polynomial response up to the Obspy devs to implement
https://github.com/GeoNet/delta/blob/c4ff23d3f90caf85412231477cbb87314dc5481a/internal/stationxml/encoder10.go#L235

from obspy.core.inventory.response import PolynomialResponseStage, InstrumentSensitivity

def calculate_polynomial_sensitivity(st):
    for tr in st: 
        if not hasattr(tr.stats,"response"): #skip channels with no attached response
            continue
        if tr.stats.response.instrument_sensitivity != None: #skip channels that already have sensitivities            
            continue
        gain = 1
        stages = tr.stats.response.response_stages
        for s in stages:
            if type(s) == PolynomialResponseStage:
                gain /= s._coefficients[1] #asssume gain is related to the second coefficient
            else:
                gain *= s.stage_gain
        freq = stages[-1].stage_gain_frequency
        input_units = stages[0].input_units
        output_units = stages[-1].output_units        
        tr.stats.response.instrument_sensitivity = InstrumentSensitivity(gain,freq,input_units,output_units)

#function operates in place on the stream
calculate_polynomial_sensitivity(st)

[elidana]

this is absolutely brilliant, thanks @Thomas-Benson !!! <3 <3
and it seems that someone can probably raise an issue with the obspy developers

[timronan]

This is really useful information. Thanks for reporting and documenting this issue.

[elidana]

@timronan the user who bumped into this issue has now opened an issue with Obspy, if you're interested: obspy/obspy#3632

[timronan]

Thanks for the link, I will watch this issue.