|Despite recent improved methods, obtaining paleointensities from lavas is still a complicated process. With the IZZI-Thellier method, samples are heated which makes them prone to (chemical) alteration, and success rates are often low. The pseudo-Thellier method avoids heating the samples but uses alternating ?fields to replace the natural remanent magnetization (NRM) of a sample with an anhysteretic remanent magnetization (ARM). However, the resulting calibration relationship which converts the pseudo-Thellier slope to a paleointensity, misses the origin. Also, various samples can not be used because their B1/2ARM, a grain size selection criteria, is outside the accepted boundaries of 23 to 63mT. These and other previous paleointensity techniques assume all grains in a sample behave in a similar manner as they did when they acquired their NRM during cooling. The newly proposed End-Member Modelling Analysis (EMMA) avoids making grain size selections and assumes the magnetic remanence in a sample is carried by an assemblage of iron-oxides. Each one has their own mapping factor between the NRM acquired during cooling in a natural fi?eld and the ARM acquired in a laboratory fi?eld. The EMMA-model was previously capable of calculating accurate results for a real dataset with data from Hawaii and Reunion of which the paleointensity was known. They cover a small range of paleointensities from 35 to 40uT. Here I expand this calibration dataset with new data from recent lava flows of Iceland and New-Zealand, which cooled in ?fields of >50uT. The samples from these locations are subjected to all three paleointensity methods, and to a rock-magnetic study to identify the magnetic remanence carriers of the samples. The IZZI-Thellier method was very successful and
con?firms the assumed high paleointensities of the samples. The pseudo-Thellier method gave almost no results. The B1/2ARM of Iceland was too low, and of New-Zealand too high for most samples to be interpreted and for most other results the calibration relation calculated lower paleointensities than expected. The most accurate EMMA-model was based on a 3 end-member model with outliers, overprints and unsuccessful locations removed and ultimately contained 189 samples from Hawaii, Reunion, Tenerife, Iceland and New-Zealand. It calculated an average difference in paleointensity of -0.056uT and an absolute difference in paleointensity of 6.497uT. The successful samples had a B1/2ARM ranging from 16.3 to 124.7 mT which is a much wider range than would be selected with the pseudo-Thellier method. The three different end-members which the model determines for each sample seem to correlate to different Curie temperatures, which indicates the end-members are based on specifi?c iron-oxide compositions. Future studies should expand the dataset further with more samples from high and low magnetic fi?elds, with different types of compositions and therefore different possible end-members. A larger dataset could also shed light on why some locations are unsuccessful and how to remove bad measurements from the dataset. Then the EMMA-model can be applied to sites for which the reference paleointensity is not known to improve future geomagnetic fi?eld models.