Soil moisture is one of the most important drivers of the hydrological cycle. Therefore, global soil moisture records are needed to study hydrology driven phenomena of the earth system such as climate change, vegetation growth, and many others. The most important sources for global soil moisture records are space borne microwave instruments. However, such satellite-derived soil moisture products are subject to errors and their correct interpretation and application requires an in-depth understanding of their accuracy. Triple collocation (TC) analysis is a method for estimating the individual signal- and random error variances of three collocated data sets with mutually uncorrelated errors without relying on a high-quality reference data set. It has therefore evolved as one of the most important methods for estimating error structures in remotely sensed soil moisture data sets. Nevertheless, the exploitation of the full potential of the TC method is still subject to ongoing research. On the other hand, TC analysis is based on a variety of assumption on the structure of the underlying data sets whose validity hasn't been fully investigated yet. This thesis further develops the TC method, aiming for an improved and more complete estimation of error structures in remotely sensed soil moisture data sets. Existing TC implementations are reviewed, assumptions underlying the method are evaluated, and novel generalizations and extensions to the method are proposed, which allow for a more objective interpretation of soil moisture data quality as well as for the additional estimation of spatial error auto-correlation and mutual error cross-correlation structures.