A precise understanding of crustal structure is essential to the fields of geodynamics, seismology and certain branches of geophysics. A boundary between the crust and the mantle is known as the Mohorovičić discontinuity, simply referred to as the 'Moho'. Moho geometry and depth have been extensively studied in Europe, but there are still regions with little information about it. One such area is the northern Baltics, Latvia in particular. So far, only one seismic refraction profile, spanning from Sovetsk (Kaliningrad) to KohtlaJärve (Estonia), has been used to study the deep structure of the Earth in Latvia. We propose gravity inversion (Parker-Oldenburg algorithm) to gain more insight into the Moho depth of Latvia. Multiple gravity sources are combined into a single dataset with the regressionkriging method. Gravity data are then iteratively filtered with various wavelength lowpass filters. We use different combinations of these filtered datasets and varying input parameters -mean depth to the Moho and density contrast between the crust and the mantle -to carry out multiple iterations of the inversion, validating the results by seismic refraction profiles available for Latvia. The calculated Moho depth varies from 41.5 km in the southern and northeastern parts of Latvia to 46.5 km in the northern part of Latvia and the Gulf of Riga. We conclude that gravity inversion with the Parker-Oldenburg algorithm can be used as an alternative to the seismic exploration of the Moho, especially in places where there is a shortage of earlier seismic data. The obtained results also show that it is necessary to create multiple models with various combinations of input values when using the Parker-Oldenburg inversion algorithm.