I am using gridded climate data to regress streamflows in a watershed using PCR. I found that more than 40% variance is explained? Is this a good approach?

Minimum of how much variance (PCA explained) do I need to accept the regression model?