@article{fdi:010091991,
  title = {{D}enoising of geodetic time series using spatiotemporal graph neural networks : application to slow slip event extraction},
  author = {{C}ostantino, {G}. and {G}iffard-{R}oisin, {S}ophie and {D}alla {M}ura, {M}. and {S}ocquet, {A}.},
  editor = {},
  language = {{ENG}},
  abstract = {{G}eospatial data have been transformative for the monitoring of the {E}arth, yet, as in the case of (geo) physical monitoring, the measurements can have variable spatial and temporal sampling and may be associated with a significant level of perturbations degrading the signal quality. {D}enoising geospatial data is, therefore, essential, yet often challenging because the observations may comprise noise coming from different sources, including both environmental signals and instrumental artifacts, which can be spatially and temporally correlated, thus hard to disentangle. {T}his study addresses the denoising of multivariate time series acquired by irregularly distributed networks of sensors, requiring specific methods to handle the spatiotemporal correlation of the noise and the signal of interest. {S}pecifically, our method focuses on the denoising of geodetic position time series, used to monitor ground displacement worldwide with centimeter-to-millimeter precision. {A}mong the signals affecting global navigation satellite system ({GNSS}) data, slow slip events ({SSE}s) are of interest to seismologists. {T}hese are transients of deformation that are weakly emerging compared to other signals. {H}ere, we design {SSE}denoiser, a multistation spatiotemporal graph-based attentive denoiser that learns latent characteristics of {GNSS} noise to reveal {SSE}-related displacement with submillimeter precision. {I}t is based on the key combination of graph recurrent networks and spatiotemporal {T}ransformers. {T}he proposed method is applied to the {C}ascadia subduction zone, where {SSE}s occur along with bursts of tectonic tremors, a seismic rumbling identified from independent seismic recordings. {T}he extracted events match the spatiotemporal evolution of tremors. {T}his good space-time correlation of the denoised {GNSS} signals with the tremors validates the proposed denoising procedure.},
  keywords = {denoising ; geodesy ; {D}eep learning ; geospatial data ; global navigation satellite system ({GNSS}) ; global positioning system ({GPS}) ; graph neural networks ; multistation ; seismology ; slow slip events ({SSE}s) ; spatiotemporal ; spatiotemporal attention ; time-series analysis},
  booktitle = {},
  journal = {{IEEE} {J}ournal of {S}elected {T}opics in {A}pplied {E}arth {O}bservations and {R}emote {S}ensing},
  volume = {17},
  numero = {},
  pages = {17567--17579},
  ISSN = {1939-1404},
  year = {2024},
  DOI = {10.1109/jstars.2024.3465270},
  URL = {https://www.documentation.ird.fr/hor/fdi:010091991},
}