If you have point climate data sets, you can validate the predicted SWAT climate data ( like rainfall and potential evapotranspiration) with the observation series. What I understood from your question is you don't have a gauged catchment flow or constituent loading/ sediment data.
One way to cross validate is to use conceptual rainfall runoff models with the same grid climate and potential evapotranspiration data for the ungauged or poorly monitored sub-catchments of the bigger scale. You can also follow a marginalisation procedure. In any of the above methods, the result remains uncertain. Difficult to fix parameter value ranges ( optimised parameter value), perform local and global sensitivity analysis, and statisticize your result with the model performance rating criteria due to the absence of monitoring datasets. Check this out!