Nitrogen oxides play a key role in tropospheric chemistry; to study the distribution patterns of the corresponding anthropogenic emissions (fossil, industrial, waste), we use three independent data sources: GOME measurements of the tropospheric NO2 column density fields, the EDGAR 3 emission inventory as an estimation of the anthropogenic NOx emissions and nighttime images of worldwide human settlements seen by the DMSP OLS satellite instrument as a proxy for these emission patterns. The uncertainties are not known precisely for any of the fields. Using the MOZART-2 CTM, tropospheric column density fields are calculated from the emission estimates, and transformations are developed to turn the GOME columns into anthropogenic emission fields. Assuming the errors of the three data sources (GOME, EDGAR, lights) to be independent, we are able to determine ranges for the pattern errors of the column density fields and values for the pattern errors of the source fields by a correlation analysis that connects relative error (co)variances and correlation coefficients. That method was developed for this investigation but can generally be used to calculate relative error variances of data sets, if the errors of at least three of them can be assumed to be independent. We estimate the pattern error of the EDGAR 3 anthropogenic NOx emission field as (27 ¿ 5)%, which can be reduced by combining all fields to (15 ¿ 3)%. By determining outliers, we identify locations with high uncertainty that need further examination.