Magnetometry on the Lagash Archaeological Project: Salvaging a Flawed Dataset


The spring 2022 season (3LAP) commenced with placing three new 10×10 meter trenches in Area H. Guided by GeoTIFFs of the magnetometry survey conducted during the 2019 season (1LAP), these trenches were located so as to find walls and streets revealed by magnetometry. Though the general character of the subsurface architecture reflected what was shown in the magnetogram, direct correlation between features in the magnetometry data and the excavations was not to be had. Trench 3, in particular, was extended 5 meters north in order to better intercept structures that the magnetometry showed there, with mixed results. Our assumption at the time was that the numerous Late ED kilns in the area complicated the magnetometry, and that the difficulty in finding direct matches between the magnetometry and the excavated architecture may have been further hampered inconsistencies in the GNSS readings due to different equipment employed between 1LAP and 3LAP. Attempts to re-plot the original 1LAP magnetometry data were unsuccessful and eventually abandoned.




Above: Dr. Paul Zimmerman with the Lagash Archaeological Project’s Sensys MXPDA magnetometry cart

A new Sensys MXPDA magnetometry system consisting of three FGM650/3 sensors mounted on a push cart was purchased for the project’s fall 2022 season (4LAP), and Dr. Paul Zimmerman (with training from Geoffrey Jones of Archaeo-Physics, LLC) devised a survey strategy to complement and extend the 1LAP survey. Though the new data clearly showed signatures of subsurface features such as kilns, walls, and roads, there was no correlation between the 4LAP and 1LAP magnetometry when those two surveys abutted or overlapped. Though the hardware used in the 1LAP survey was different than that used in the 4LAP survey, both systems had similar sensors with the same inter-sensor spacing, and their data should have been comparable. So, during the 4LAP field season a second attempt was made to re-plot the 1LAP data from the source files, in hopes of merging the two datasets. Armed with our new understanding of magnetometry data, the error in the 1LAP data was immediately made clear: the first two day’s data from 2019 were recorded in the wrong direction, with repercussions that affected all subsequent survey units.


Namely, those first two days’ data (collected 3 April, 2019 and 4 April, 2019) were reversed east-to-west. Instead of starting in the southwest corner of each 20×20 meter survey unit and walking north-south transects in boustrophedon fashion proceeding west to east, the surveyors began in the southeast corner and proceeded westward, making those two days’ data a mirror image of what the processing software expected. On the third day of the survey (6 April 2019), this mistake was noticed, and all subsequent survey units were collected as intended from west to east. An attempt was then made to reconcile the first two days’ data with subsequent days’ surveys by flipping the reversed units horizontally. But instead of flipping them east-to-west in place, as should have been done, the survey units were rolled on their east edge and then 20 meters was added to the X coordinates of every other grid square in order to align them with the surveys of the first two days. This effectively displaced the 1LAP magnetometry results 20 meters to the east, explaining the lack of correlation between that dataset and the 4LAP dataset or features found in trenches dug in the 1LAP, 3LAP, and 4LAP seasons. (2LAP was devoted to mapping and coring, and no trenches were excavated that season.)

Overview of LAP trenches in Area H, showing the 20 meter eastward displacement of the 1LAP magnetogram (grey) versus two survey units of 4LAP magnetometry data (greens).

Plots of the first two days’ data from the 1LAP magnetometry survey (red) displayed atop the finished 1LAP magnetogram, with its 20 meter eastward shift (grey). The red dashed line indicates the position of the easternmost of these strips when flipped on its eastern edge, resulting in the accidental displacement of the overall magnetogram.


Completed magnetogram of the 4LAP dataset merged with the 1LAP dataset once its spatial registration had been corrected.

Fortunately, once the error was discovered and corrected, we were able to fully integrate the 1LAP dataset with our newly-collected 4LAP magnetometry. The icing on the cake, however, is that not only did the two surveys’ data now match, but they also both matched exactly what has been revealed in excavation. On this solid foundation, we are confident that future seasons will bring greater understanding of ancient Lagash through additional geophysical survey and excavation, as well as through novel computational analyses of our magnetometry data.