Purpose: A methodology was developed to evaluate and mitigate the impacts of particle size and post-depositional diagenesis when using mineral magnetic signatures to trace the sources of historically deposited sediment in farm dams in the South African Karoo. Materials and methods: Samples from a range of potential sediment sources were sieved to different particle size fractions, and the relationships between pairs of tracer signatures were established for each fraction. Non-conservatism of the magnetic signatures was determined by identifying whether the magnetic signatures of the farm dam sediments were within the range of those of the fractionated source samples. By fractionating the sediment source samples, the core samples were able to be traced using appropriately sized sources. Results and discussion: It was found that strong relationships existed between the pairs of tracer signatures at all particle size fractions. Relationships in the <32 μm fraction were significantly different to that of coarser fractions. It was also found that particle size had a large effect on all magnetic signatures and would prove to be a large source of uncertainty if not accounted for within any methodology developed for quantitative source discrimination and source apportionment. There was very little non-conservatism caused by diagenetic or biogenic processes in six of the seven dams sampled. In one dam, there was evidence to suggest that dissolution had probably caused the loss of almost all small superparamagnetic and stable single-domain grains. The other signatures associated with coarser magnetic grains in this dam were generally unaffected by the dissolution processes. Conclusions: The good preservation of magnetic signatures suggests that they can make reliable tracers over historical timescales (up to 164 years) in the Karoo and similar semi- arid catchments. However, the mitigation of particle size effects and screening for post-depositional alteration is an essential part of their use. The methodology presented in this paper is a potential way of recognising tracer non-conservatism and limiting its effects in future studies.