vertical cross sections via GMG Vote Maps

[amrelsharkawy]

I’ve created vote maps following the instructions from GMG' GitHub, and I’ve attached a Paraview screenshot for your reference. However, when I try to create a slice along the profile through the 3D vote map, I encounter an error (pasted below). Do you have any suggestions for resolving this?

+++ The error message I got +++

---

********* Profile 1 *******

---

lon start 9.0
lat start 43.0
lon end 15.0
lat end 49.0
ERROR: LoadError: InexactError: Int64(NaN)
Stacktrace:
  [1] Int64
    @ ./float.jl:994 [inlined]
  [2] convert
    @ ./number.jl:7 [inlined]
  [3] _round_convert
    @ ./rounding.jl:480 [inlined]
  [4] round
    @ ./rounding.jl:479 [inlined]
  [5] round
    @ ./rounding.jl:477 [inlined]
  [6] _broadcast_getindex_evalf
    @ ./broadcast.jl:678 [inlined]
  [7] _broadcast_getindex
    @ ./broadcast.jl:661 [inlined]
  [8] getindex
    @ ./broadcast.jl:610 [inlined]
  [9] macro expansion
    @ ./broadcast.jl:973 [inlined]
 [10] macro expansion
    @ ./simdloop.jl:77 [inlined]
 [11] copyto!
    @ ./broadcast.jl:972 [inlined]
 [12] copyto!
    @ ./broadcast.jl:925 [inlined]
 [13] copy
    @ ./broadcast.jl:897 [inlined]
 [14] materialize(bc::Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{3}, Nothing, typeof(round), Tuple{Base.RefValue{Type{Int64}}, Array{Float64, 3}}})
    @ Base.Broadcast ./broadcast.jl:872
 [15] interpolate_datafields(V::GeoData, Lon::Array{Float64, 3}, Lat::Array{Float64, 3}, Depth::Array{Float64, 3})
    @ GeophysicalModelGenerator ~/.julia/packages/GeophysicalModelGenerator/Q1va2/src/utils.jl:1067
 [16] combine_vol_data(VolData::@NamedTuple{Zhu2015::GeoData, Giacomuzzi2023::GeoData, UU07_Amaru2007::GeoData, MIT08_Li2008::GeoData, Piromallo2003::GeoData, Koulakov2009::GeoData, Paffrath2021::GeoData, Zhao2016::GeoData, Timko2023_Mantle::GeoData, Lippitsch2003::GeoData, Mitterbauer2011::GeoData, Plomerova2022::GeoData, VoteMap_3D::GeoData}; lat::Nothing, lon::Nothing, depth::Nothing, dims::Tuple{Int64, Int64, Int64}, dataset_preferred::Int64)
    @ GeophysicalModelGenerator ~/.julia/packages/GeophysicalModelGenerator/Q1va2/src/ProfileProcessing.jl:245
 [17] combine_vol_data(VolData::@NamedTuple{Zhu2015::GeoData, Giacomuzzi2023::GeoData, UU07_Amaru2007::GeoData, MIT08_Li2008::GeoData, Piromallo2003::GeoData, Koulakov2009::GeoData, Paffrath2021::GeoData, Zhao2016::GeoData, Timko2023_Mantle::GeoData, Lippitsch2003::GeoData, Mitterbauer2011::GeoData, Plomerova2022::GeoData, VoteMap_3D::GeoData})
    @ GeophysicalModelGenerator ~/.julia/packages/GeophysicalModelGenerator/Q1va2/src/ProfileProcessing.jl:218
 [18] extract_ProfileData(ProfileCoordFile::String, ProfileNumber::Int64, DataSetFile::String; DimsVolCross::Tuple{Int64, Int64}, DepthVol::Nothing, DimsSurfCross::Tuple{Int64}, WidthPointProfile::Unitful.Quantity{Int64, 𝐋, Unitful.FreeUnits{(km,), 𝐋, nothing}})
    @ GeophysicalModelGenerator ~/.julia/packages/GeophysicalModelGenerator/Q1va2/src/ProfileProcessing.jl:422
 [19] extract_ProfileData
    @ ~/.julia/packages/GeophysicalModelGenerator/Q1va2/src/ProfileProcessing.jl:404 [inlined]
 [20] top-level scope
    @ /meissnerdata/seismologie/models/GMG_Datasets/PickingData/MediT/Process_easternAlps.jl:67
in expression starting at /meissnerdata/seismologie/models/GMG_Datasets/PickingData/MediT/Process_easternAlps.jl:59

[albert-de-montserrat]

Could you share the exact lines of code to reproduce this error? Some NaNs where produced at some point.

[amrelsharkawy]

I used the following command, where volumetric data are list in file_datasets and the coordinates of the profile are given in a file_profiles,
Profile = extract_ProfileData(file_profiles,iprof,file_datasets,DimsVolCross=(500,500),DimsSurfCross=(200,),WidthPointProfile=50km)

For me, The error occurs within the interpolate_datafields function, where attempts to round a 3D array of Float64 values to Int64, but encounters NaN (Not a Number) values that cannot be converted exactly to integers. The NANs are likely coming from the creation of the VoteMap itself, I was wondering if there is away to overcome this while creating the 3D vote map.

[albert-de-montserrat]

The vote map does not seem to contain NaNs

using JLD2, GeophysicalModelGenerator

Pwave_Zhao      = load("Zhao_etal_2016_data/Zhao_etal_2016_data/Zhao_Pwave.jld2","Data_set_Zhao2016_Vp");
PSwave_Koulakov = load("Koulakov_EuropeAsia/Koulakov_EuropeAsia/Koulakov_Europe.jld2","DataKoulakov_Alps");
Pwave_Paffrath  = load("Paffrath_Pwave_Alps_SE_2021/Paffrath_Pwave_Alps_SE_2021/Paffrath_Pwave.jld2","Data_set_Paffrath2021_Vp");

vmap = votemap( 
    [Pwave_Paffrath,       PSwave_Koulakov,    Pwave_Zhao],
    ["dVp_Percentage>3.0","dVp_percentage>2.0","dVp_Percentage>2.0"], 
    dims=(100,100,100)
);

julia> any(isnan, vmap.fields[1])
false

I am missing some information to reproduce your error, what are file_profiles, iprof, and file_datasets exactly?

[amrelsharkawy]

Case 1): Creating vertical cross section through a 3D Vote map only → succeeds (see the plot - you can also see that I can include seismicity and Moho data with it successfully)
Case 2): Creating vertical cross section through a list of several tomography models only (40+ models without vote map) → succeeds (as shown previously)

Case 3)❌ Creating vertical cross section through list contains the Vote map + tomography models together (full Datasets.txt: 40 models + votemap) → fails

Why ==>Incompatibility issue: combine_vol_data() fails only when votemap::GeoData is combined with tomography::GeoData objects in the same @NamedTuple. Vote map creation alone and pure tomography combinations both work.

ERROR: LoadError: InexactError: Int64(NaN)
@ ./float.jl:923
...
@ utils.jl:1067 # interpolate_datafields round()
@ ProfileProcessing.jl:245 # combine_vol_data

The point is Vote map integer values (0,1,2,3...) conflict with tomography Float64 during common-grid interpolation → NaN propagation → round(Int64, NaN) crash.

Here are a couple of point:

• Does combine_vol_data() expect consistent eltype between vote maps (Int) and tomography (Float64)?
• Should interpolate_datafields() handle mixed Int64/Float64 + NaN gracefully?