Is removing bad/noisy data acceptable?

Short Answer: Yes, it is acceptable if done procedurally—up to about 20% of the total with most data sets

Long Answer: Each measurement you take is an integrated response from the entire half or full-space. As such, the measurement carries information about the entire half or full-space. Data redundancy due to over-sampling ensures removal of some data is not fatal on the model's resolution.

Hand-picking data points to remove is not recommended as it is typically not a defensible practice. Also, massive data removal of a large number of data points will cause poor resolution in any area without data coverage.

When removing data, you will want to only remove ≤5% of data per iteration until you get to ≤10% RMS.

Related Material:

Learn how to remove misfit data using the Histogram Filtering Tool in EarthImager™ 2D.