Outliers and the standard deviation of the observations can both have a significant impact on the adjustment results. The relative-gravity meter—both metal and quartz-spring versions—is suspect to periodically recording a "bad" measurement. At times these may be tares (offsets that persist across subsequent measurements) or extreme drift, but at other times, only a single reading seems to be bad (based on repeat observations at that station). The ability to quickly and iteratively remove potential bad measurements is an important function of GSadjust.
Similarly, the standard deviation of the observations may need updating when adjusting a network. The weight of each observation is proportional to 1 over the variance (standard-deviation squared). For example, an observation (delta-g or datum) with standard deviation of 5 µGal would have 1/25 the weight of an observation with a standard deviation of 1 µGal. This range of standard deviations (or greater) is not uncommon for field surveys.
Outliers
Observations can be removed from the adjustment by unchecking the appropriate checkbox. If the checkbox next to a station in the Data tree view is unchecked, the respective delta-g table on the Drift tab will be updated. In this case, the delta-g table on the Network adjustment tab will need to be updated using one of the "Populate delta table..." commands (the delta-g tables also need to be updated after changing the drift method).
In addition to unchecking individual station occupations, delta-g's can be unchecked on the Network adjustment tab. The delta-g table does not need to be updated, but the "Adjust network" command must be run to update the adjustment results. After the initial adjustment, by sorting the delta-g table by the adjustment residual (by clicking the column header), the worst observations can be identified and potentially removed. By cross-referencing the residuals with stations and occupation times, it may be possible to identify problem stations, or periods of bad data.
Datum observations can also be removed from the adjustment by unchecking the respective checkbox. The "Adjust network" command must be run to update the adjustment results.
Standard deviation
The standard deviation of the observations may need updating based on the adjustment results. As noted, it may be undesirable to have a wide range of standard deviations, thus giving some undue weight. A second reason to update the standard deviation of the observations is to meet the the Chi-square test criteria. if the a priori standard deviations are too low (that is, the observations are worse than estimated), the Chi-square test will be rejected.
Standard deviations of delta-g's and datums can be edited directly by double-clicking the respective cells on the Network adjustment tab. More efficiently, a minimum standard deviation, additive term, or multiplier can be applied to all of the delta-g's in a survey or campaign in the Network adjustment options dialog.
