The analysis and mapping of agronomic and environmental spatial data require observations to be comparable. Heterogeneous spatial datasets are those for which the observations of different datasets cannot be directly compared because they have not been collected under the same set of acquisition conditions, for instance within the same time period (if the variable of interest varies across time), with consistent sensors or under similar management practices (if the management practices impact the measured value) among others. When heterogeneous acquisition conditions take place, there is a need for harmonization procedures to make possible the comparison of such observations. This analysis details and compares four automated methodologies that could be used to harmonize heterogeneous spatial agricultural datasets so that the data can be analysed and mapped conjointly. The theory and derivation of each approach, including a novel, local spatial approach is given. These methods aim to minimize the occurrence of discrepancies (discontinuities) in the data. The four approaches were evaluated and compared with a sensitivity analysis on simulated datasets with known characteristics. Results showed that none of the four methods consistently delivered a better harmonization accuracy. The accuracy and preferred choice for the harmonization procedures was shown to be influenced by (i) within-field spatial structures of the datasets, (ii) differences in acquisition conditions between the heterogeneous spatial datasets, and (iii) the spatial resolution of the simulated data. The four approaches were used to harmonize real within-field grain yield datasets and a discussion to help users select an appropriate harmonization methodology proposed. Despite significant improvements in dataset harmonization, discontinuities were not entirely removed and some uncertainty remained.