The data set below was part of a magnetic gradient survey performed at the 3030 Winchester site (results are discussed in more detail here). Data were collected at 0.5 meter transect intervals, with 8 samples collected per meter along each transect (sample density = 16 s/m). The sample density was experimentally "thinned" by removing alternate transects from the data file (processing and display parameters are otherwise identical). A transect interval to 1m (which has been very common practice on both historic and prehistoric sites) was found to dramatically reduce the detectability of hearth features.
Figure 1. Subset of magnetic survey data. Darker shades represent higher (positive) values. Five Prehistoric hearth features plainly appear as circular magnetic highs. When these features were excavated, a correlation was found between the amplitude and geometry of the geophysical anomalies and the volume and composition of the excavated hearths.
Figure 2. (below) The data set plotted above with alternate lines of data removed to simulate survey at 1m transect intervals (8 s/m). At this sampling interval, detection of hearth features is hit-and-miss. Only one of the hearths is plainly detected, and correlation between feature size/composition and anomaly strength/geometry is lost - in fact, the only hearth that is plainly detected (N15/E6) is the least substantial of the features within this area. While these hearth features create very distinct anomalies on an otherwise magnetically "quiet" site, their radius of detectability is quite small, and they are easily missed if the transect interval is too great. While 1m transect intervals can be appropriate where features of interest are large or substantial (e.g. many historic sites). Spatially small and very subtle anomalies require closer sampling intervals. Cost-effectiveness is best served when the sampling strategy is optimized to address specific research interests, feature composition, and site conditions with budgetary constraints.
See also: 3030 Winchester Page