The GEOMAGIA database

Available Models

This section gives an overview of the available global geomagnetic field models that can be calculated using the GEOMAGIA50 web site. Monika Korte (GFZ Potsdam) and Cathy Constable (Scripps Institution of Oceanography) provided the codes and coefficients for the calculations. Additional information can be found here.

Model predictions from regions sparse in data may have uncertainties larger than the estimates provided by the bootstrap calculations of CALS3k.4, CALS10k.1b, ARCH3k.1 and SED3k.1. Uncertainties are shown in the last three columns of the output file.

A description of the column headers in the model output files can be found at the bottom of the page.

Current Models:

(a) CALS3k.4

Constructed in 2011 using a combination of the available archeomagnetic and sediment data covering the past 3 ka (Korte and Constable, 2011). Compared with CALS3k.3, 163 additional archeomagnetic data and 13 newly published sediment records were used in the model construction. This resulted in notable differences between the two models over South-East Asia, Alaska and Siberia.

CALS3k.4 is constrained closely to gufm1 (Jackson et al., 2000) for the past four centuries. It can be used for calculations of the field over the entire globe. It will produce bootstrap estimates of uncertainty.

A zip folder containing model overlays of declination, inclination and intensity for Google Earth can be download here (courtesy of Alexandra Lodge).

(b) CALS10k.1b

Covers the past 10 ka and incorporates the largest number of data to date. CALS10k.1b supersedes CALS7K.2. It is based on sediment, lava and archeological data available up to 2011. The data compilation is dominated by sediment data. The final model is an average obtained from bootstrap sampling (the 'b' in the model name denoting bootstrap sampling) to account for uncertainties in palaeomagnetic and chronological data. Consequently, it is smoothed strongly compared with CALS3k.3 and CALS3k.4 to allow reconstruction of earlier epochs.

CALS10k.1 is constrained closely to gufm1 (Jackson et al., 2000) for the past four centuries. It can be used for calculations of the field over the entire globe. It will produce bootstrap estimates of uncertainty. The model is described in Korte et al. (2011).

A zip folder containing model overlays of declination, inclination and intensity for Google Earth can be download here (courtesy of Alexandra Lodge).

Constructed using available archeomagnetic data up to 2009. It covers the past 3 ka. Data are strongly biased towards the Northern Hemisphere and Europe in particular. The model gives reasonable field values for the Northern Hemisphere, but should not be used for global studies or Southern Hemisphere field predictions.

ARCH3k.1 is constrained to agree loosely with the historical gufm1 model of Jackson et al., 2000 for the last 400 years. It will produce bootstrap estimates of uncertainty. Further details can be found in Korte et al. (2009).

A zip folder containing model overlays of declination, inclination and intensity for Google Earth can be download here (courtesy of Alexandra Lodge).

(d) SED3k.1

Constructed using available sediment data up to 2009. It covers the past 3 ka. Data have a better global distribution (less biased towards the Northern Hemisphere) than ARCH3k.1 and can be used for prediction in the Southern Hemisphere. The model output is smoothed in time as a result of the sedimentary recording process and the methods of sub-sampling employed.

SED3k.1 is constrained to agree loosely with the historical gufm1 model of Jackson et al., 2000 for the last 400 years. It will produce bootstrap estimates of uncertainty. Further details can be found in Korte et al. (2009).

A zip folder containing model overlays of declination, inclination and intensity for Google Earth can be download here (courtesy of Alexandra Lodge).

(e) pfm9k.1a

The pfm9k.1a model (Nilsson et al., 2014) spans the past 9000 years. It uses the same dataset used for CALS10k.1b (Korte et al. (2011)), but introduces new data treatments, particularly for the sedimentary data. These include redistributing the weight given to different data types and data sources, iteratively recalibrating relative declination data and adjusting the timescales of the sediment records using a preliminary model. These data treatments, particularly the timescale adjustments, reduce inconsistencies in the database and enable pfm9k.1a to capture larger amplitude PSV variations. The model is valid up to 1900 AD and unlike CALS10k.1b is not constrained by gufm1 (Jackson et al., 2000) for the recent times.

A zip folder containing model overlays of declination, inclination and intensity for Google Earth can be download here (courtesy of Alexandra Lodge).

(f) ARCH-UK.1

The ARCH-UK.1 model is only applicable for the UK and only valid across the latitude range 49°N-61°N and the longitude range 11°W-2°E. It is global model of the field (using the same dataset as ARCH10k.1 (Constable et al., 2016), GEOMAGIA50.v3 30th April 2015), but the UK data are weighted by four times more than the global data. The model spans 10,000 BCE to 1990 CE. The final model is an ensemble of 2000 individual models. Uncertainties on the model are the standard deviation of all ensemble predictions. See Batt et al., Journal of Archaelogical Science, 2017, for further details on the modelling procedure.

(g) CALS10k.2

Covers the past 10 ka and incorporates nearly the same sediment data as CALS10k.1b and an updated set of lava and archeological data, the same as ARCH10k.1. CALS10k.2 is of higher temporal and spatial resolution than CALS10k.1b due to improved data uncertainty estimates for the sediment records. It is not an average of bootstrap sampling. It is constrained closely to gufm1 (Jackson et al., 2000) for the past four centuries. The model is described in Constable et al., 2016.

(h) ARCH10k.1

Constructed using all available archeomagnetic and lava flow data from GEOMAGIA50.v3 up to April 30, 2015. Due to very uneven data coverage it should not be used to infer global characteristics and even regional predictions should be regarded with caution both for ages > 3ka, when numbers of data are sparse everywhere, and in general in any regions devoid of data. The model is described in Constable et al., 2016.

(i) HFM.OL1.AL1

Covers the same time span and incorporates the same data as CALS10k.2, but differs somewhat in modelling strategies like calibrations for relative paleointensities and iterative outlier rejection. The fit to all data is comparable for the two models, but HFM.OL1.A1 has somewhat higher temporal but somewhat lower spatial resolution than CALS10k.2. It is not constrained by any other model at the recent end. The model is described in Constable et al., 2016 and Panovska et al, 2015.

Superseded Models:

(a) CALS3k.3

Superseded by CALS3k.4. The CALS3k.3 model was constructed using archeomagnetic and sediment data available in 2009. It comprises the two datasets used to construct ARCH3k.1 and SED3k.1. CALS3k.4 is now recommended. Further details can be found in Korte et al. (2009) and Korte and Constable (2011).

(b) CALS7K.2

An early model described in Korte and Constable, 2005. It was constructed using full-vector archeomagnetic data and directional data from a limited number of sediment cores. It covers the past 7 ka. The dipole moment is underestimated as (i) no sedimentary intensity data and (2) only limited archeomagnetic intensity data before ~1000 BC, were included in the model construction. CALS7K.2 is superseded by CALS10k.1b.

Model Output Column Descriptors:

Column	Description
AGE_BP	Time in years BP (1950 AD is 0 years BP)
AGE_AD	Time in years AD/BC
SITE_LAT	Latitude (°) of the Country/State/Region/Sea or Site or Core selected to query
SITE_LON	Longitude between 0° and 360° of the Country/State/Region/Sea or Site or Core selected to query
INCL	Inclination (°)
DECL	Declination (°)
INT	Intensity (μT)
VADM	Virtual Axial Dipole Moment (x10²² Am²). See here.
EI	Bootstrap uncertainty estimates of Inclination (°)
ED	Bootstrap uncertainty estimates of declination (°)
EF	Bootstrap uncertainty estimates of intensity (μT)
EM	Bootstrap uncertainty estimates of Virtual Axial Dipole Moment (x10²² Am²)