Prediction of the Earth Orientation Parameters
Introduction to the methods
The specifications of the methods that are currently
operational are mentioned in the following table.
EAM is equivalent to the component-wise summation of AAM, HAM, OAM, and SLAM (i.e., EAM = AAM + HAM + OAM + SLAM).
Acknowledgements
The data sources used in the current framework are taken from different datacenters. These are as follows.
IERS EOP series can be accessed here.
SYRTE EOP and EAM series can be accessed here.
JPL EOP series can be accessed here.
GFZ EAM series can be accessed here.
Methods description
| Method | Input EOP | Input EAM | Algorithm | Forecasting horizon (days) | Comments |
|---|---|---|---|---|---|
| lod-edlstm-10d-iers-gfz | IERS length of day | GFZ AAM χ3 | Encoder-decoder long short-term memory | 10 | |
| lod-edlstm-10d-syrte-gfz | SYRTE length of day | GFZ AAM χ3 | Encoder-decoder long short-term memory | 10 | |
| lod-edlstm full eam-10d-iers-gfz | IERS length of day | GFZ EAM χ3 | Encoder-decoder long short-term memory | 10 | |
| lod-edlstm full eam-10d-syrte-gfz | SYRTE length of day | GFZ EAM χ3 | Encoder-decoder long short-term memory | 10 | |
| lod-larf-10d-iers-gfz | IERS length of day | GFZ AAM χ3 | Stacking of long short-term memory layers augmented with residual learning and fast attention | 10 | For further information please read our publication "Small Geodetic Datasets and Deep Networks: Attention-Based Residual LSTM Autoencoder Stacking for Geodetic Time Series", found here |
| lod-larf-10d-syrte-gfz | SYRTE length of day | GFZ AAM χ3 | Stacking of long short-term memory layers augmented with residual learning and fast attention | 10 | |
| lod-lstm-10d-iers-gfz | IERS length of day | GFZ AAM χ3 | Long short-term memory | 10 | |
| lod-lstm-10d-syrte-gfz | SYRTE length of day | GFZ AAM χ3 | Long short-term memory | 10 | |
| lod-quantum lstm-iers-gfz | IERS length of day | GFZ AAM χ3 | Quantum long short-term memory | 10 | For further information please refer to our publication |
| lod-quantum lstm-10d-syrte-gfz | SYRTE length of day | GFZ AAM χ3 | Quantum long short-term memory | 10 | |
| lod-recursive ode-10d-iers-gfz | IERS length of day | GFZ AAM χ3 | First order neural ODEs | 10 | |
| lod-recursive ode-10d-syrte-gfz | SYRTE length of day | GFZ AAM χ3 | First order neural ODEs | 10 | |
| lod-recursive ode full eam-10d-iers-gfz | IERS length of day | GFZ EAM χ3 | First order neural ODEs | 10 | |
| lod-recursive ode full eam-10d-syrte-gfz | SYRTE length of day | GFZ EAM χ3 | First order neural ODEs | 10 | |
| lod-recursive ode full eam-30d-iers-gfz | IERS length of day | GFZ EAM χ3 | First order neural ODEs | 30 | |
| lod-recursive ode full eam-30d-syrte-gfz | SYRTE length of day | GFZ EAM χ3 | First order neural ODEs | 30 | |
| lod-rem-10d-iers-gfz | IERS length of day | GFZ AAM χ3 | Revised Multilayer Perceptrons augmented with residual learning | 10 | |
| lod-rem-10d-syrte-gfz | SYRTE length of day | GFZ AAM χ3 | Revised Multilayer Perceptrons augmented with residual learning | 10 | |
| lod-rem-30d-iers-gfz | IERS length of day | GFZ AAM χ3 | Revised Multilayer Perceptrons augmented with residual learning | 30 | |
| lod-rem-30d-syrte-gfz | SYRTE length of day | GFZ AAM χ3 | Revised Multilayer Perceptrons augmented with residual learning | 30 | |
| lod-rem full eam-30d-iers-gfz | IERS length of day | GFZ EAM χ3 | Revised Multilayer Perceptrons augmented with residual learning | 30 | |
| lod-rem full eam-30d-syrte-gfz | SYRTE length of day | GFZ EAM χ3 | Revised Multilayer Perceptrons augmented with residual learning | 30 | |
| lod-trend only-10d-iers-gfz | IERS length of day | - | Extrapolation of length of day's secular trend | 10 | Residual part of length of day is not predicted |
| lod-trend only-10d-syrte-gfz | SYRTE length of day | - | Extrapolation of length of day's secular trend | 10 | Residual part of length of day is not predicted |
| dut1-model residuals full eam-30d-iers-gfz | IERS dUT1 | GFZ EAM χ3 | First order neural ODEs + Longth short term memory for modelling residuals | 30 | |
| dut1-model residuals full eam-30d-syrte-gfz | SYRTE dUT1 | GFZ EAM χ3 | First order neural ODEs + Longth short term memory for modelling residuals | 30 | |
| dut1-recursive ode-10d-iers-gfz | IERS dUT1 | GFZ AAM χ3 | First order neural ODEs | 10 | |
| dut1-recursive ode-10d-iers-none | IERS dUT1 | - | First order neural ODEs | 10 | |
| dut1-recursive ode-10d-syrte-gfz | SYRTE dUT1 | GFZ AAM χ3 | First order neural ODEs | 10 | |
| dut1-recursive ode-10d-syrte-none | SYRTE dUT1 | - | First order neural ODEs | 10 | |
| dut1-recursive ode-30d-iers-gfz | IERS dUT1 | GFZ AAM χ3 | First order neural ODEs | 30 | |
| dut1-recursive ode-30d-iers-none | IERS dUT1 | - | First order neural ODEs | 30 | |
| dut1-recursive ode-30d-syrte-gfz | SYRTE dUT1 | GFZ AAM χ3 | First order neural ODEs | 30 | |
| dut1-recursive ode-30d-syrte-none | SYRTE dUT1 | - | First order neural ODEs | 30 | |
| dut1-recursive ode assimilated full eam-10d-iers-gfz | IERS dUT1 | GFZ EAM χ3 | First order neural ODEs + assimilation of IERS predictions to the model | 10 | |
| dut1-recursive ode assimilated full eam-10d-syrte-gfz | SYRTE dUT1 | GFZ EAM χ3 | First order neural ODEs + assimilation of SYRTE prediction to the model | 10 | |
| dut1-recursive ode assimilated full eam-30d-iers-gfz | IERS dUT1 | GFZ EAM χ3 | First order neural ODEs + assimilation of IERS predictions to the model | 30 | |
| dut1-recursive ode assimilated full eam-30d-syrte-gfz | SYRTE dUT1 | GFZ EAM χ3 | First order neural ODEs + assimilation of SYRTE prediction to the model | 30 | |
| dut1-recursive ode full eam-10d-iers-gfz | IERS dUT1 | GFZ EAM χ3 | First order neural ODEs | 10 | |
| dut1-recursive ode full eam-10d-syrte-gfz | SYRTE dUT1 | GFZ EAM χ3 | First order neural ODEs | 10 | |
| dut1-recursive ode full eam-30d-iers-gfz | IERS dUT1 | GFZ EAM χ3 | First order neural ODEs | 30 | |
| dut1-recursive ode full eam-30d-syrte-gfz | SYRTE dUT1 | GFZ EAM χ3 | First order neural ODEs | 30 | |
| pm-model residuals full eam-30d-iers-gfz | IERS polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + Longth short term memory for modelling residuals | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-model residuals full eam-30d-jpl-gfz | JPL polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + Longth short term memory for modelling residuals | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-model residuals full eam-30d-syrte-gfz | SYRTE polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + Longth short term memory for modelling residuals | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-pcnn assimilated full eam-30d-iers-gfz | IERS polar motion | GFZ EAM χ1 and χ2 | Physically constrained neural networks + assimilation of IERS predictions to the model | 30 | Both polar motion components are inserted to one model and are predicted simultaneously |
| pm-pcnn assimilated full eam-30d-jpl-gfz | JPL polar motion | GFZ EAM χ1 and χ2 | Physically constrained neural networks + assimilation of JPL predictions to the model | 30 | Both polar motion components are inserted to one model and are predicted simultaneously |
| pm-pcnn assimilated full eam-30d-syrte-gfz | SYRTE polar motion | GFZ EAM χ1 and χ2 | Physically constrained neural networks + assimilation of SYRTE predictions to the model | 30 | Both polar motion components are inserted to one model and are predicted simultaneously |
| pm-recursive ode-10d-iers-gfz | IERS polar motion | GFZ AAM χ1 and χ2 | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-10d-iers-none | IERS polar motion | - | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-10d-jpl-gfz | JPL polar motion | GFZ AAM χ1 and χ2 | First order neural ODEs | 10 | Both polar motion components are inserted to one model |
| pm-recursive ode-10d-jpl-none | JPL polar motion | - | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-10d-syrte-gfz | SYRTE polar motion | GFZ AAM χ1 and χ2 | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-10d-syrte-none | SYRTE polar motion | - | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-30d-iers-gfz | IERS polar motion | GFZ AAM χ1 and χ2 | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-30d-iers-none | IERS polar motion | - | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-30d-jpl-gfz | JPL polar motion | GFZ AAM χ1 and χ2 | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-30d-jpl-none | JPL polar motion | - | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-30d-syrte-gfz | SYRTE polar motion | GFZ AAM χ1 and χ2 | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode-30d-syrte-none | SYRTE polar motion | - | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode assimilated full eam-10d-iers-gfz | IERS polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + assimilation of IERS predictions to the model | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode assimilated full eam-10d-jpl-gfz | JPL polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + assimilation of JPL predictions to the model | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode assimilated full eam-10d-syrte-gfz | SYRTE polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + assimilation of SYRTE predictions to the model | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode assimilated full eam-30d-iers-gfz | IERS polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + assimilation of IERS predictions to the model | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode assimilated full eam-30d-jpl-gfz | JPL polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + assimilation of JPL predictions to the model | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode assimilated full eam-30d-syrte-gfz | SYRTE polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs + assimilation of SYRTE predictions to the model | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode full eam-10d-iers-gfz | IERS polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode full eam-10d-jpl-gfz | JPL polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode full eam-10d-syrte-gfz | SYRTE polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs | 10 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode full eam-30d-iers-gfz | IERS polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode full eam-30d-jpl-gfz | JPL polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| pm-recursive ode full eam-30d-syrte-gfz | SYRTE polar motion | GFZ EAM χ1 and χ2 | First order neural ODEs | 30 | Both polar motion components are inserted to one model, but are predicted separately |
| cpo-recursive ode-10d-iers-gfz | IERS celestial pole offsets | GFZ AAM χ1, χ2, and χ3 | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-10d-iers-none | IERS celestial pole offsets | - | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-jpl-gfz | JPL celestial pole offsets | GFZ AAM χ1, χ2, and χ3 | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-10d-jpl-none | JPL celestial pole offsets | - | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-10d-syrte-gfz | SYRTE celestial pole offsets | GFZ AAM χ1, χ2, and χ3 | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-10d-syrte-none | SYRTE celestial pole offsets | - | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-30d-iers-gfz | IERS celestial pole offsets | GFZ AAM χ1, χ2, and χ3 | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-30d-iers-none | IERS celestial pole offsets | - | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-30d-jpl-gfz | JPL celestial pole offsets | GFZ AAM χ1, χ2, and χ3 | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-30d-jpl-none | JPL celestial pole offsets | - | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-30d-syrte-gfz | SYRTE celestial pole offsets | GFZ AAM χ1, χ2, and χ3 | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode-30d-syrte-none | SYRTE celestial pole offsets | - | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode full eam-10d-iers-gfz | IERS celestial pole offsets | GFZ EAM χ1, χ2, and χ3 | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode full eam-10d-jpl-gfz | JPL celestial pole offsets | GFZ EAM χ1, χ2, and χ3 | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode full eam-10d-syrte-gfz | SYRTE celestial pole offsets | GFZ EAM χ1, χ2, and χ3 | First order neural ODEs | 10 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode full eam-30d-iers-gfz | IERS celestial pole offsets | GFZ EAM χ1, χ2, and χ3 | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode full eam-30d-jpl-gfz | JPL celestial pole offsets | GFZ EAM χ1, χ2, and χ3 | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| cpo-recursive ode full eam-30d-syrte-gfz | SYRTE celestial pole offsets | GFZ EAM χ1, χ2, and χ3 | First order neural ODEs | 30 | Both celestial pole offsets components are inserted to one model, but are predicted separately |
| eop-efonode-367d-iers-gfz | IERS all EOPs | GFZ EAM χ1, χ2, and χ3 | Expressive first order neural ODEs | 367 | Simultaneous learning of deterministic and stochastic signals |
| eop-efonode-367d-syrte-gfz | SYRTE all EOPs | GFZ EAM χ1, χ2, and χ3 | Expressive first order neural ODEs | 367 | Simultaneous learning of deterministic and stochastic signals |
| reference-iers-none | IERS all EOPs except for length of day | - | 10 | Used as reference for comparison purposes | |
| reference-syrte-none | SYRTE full set of EOPs | - | 10 | Used as reference for comparison purposes |