Some details on the necessity of the changes:
-	the error was a bug in my code that used log10 instead of the natural log to calculate the RMSE, resulting 	in an underestimate of the RMSE by a factor of 1/log10(e) = 1/0.434 = 2.303
-	I’ve also added some additional checks for outliers, resulting in a more reliable calculation of RMSE
-	I previously provided a single percentage RMSE=E%, where E = 100*RMSE(ln(x)); I’ve now changed this to a 	more readily-understood RMSE=+F/-G% range (where F = 100*(exp(RMSE(ln(x)))-1) & G = 100*(np.exp(-	RMSE(ln(x)))-1). which better represents the range as percentages
-	As an example of the impacts of these changes:
	o	An error reported previously as RMSE=14% would be corrected by a factor of ~2.3 to RMSE=32%
	o	Reported as a range, this would be given as RMSE = +37/-27% 
	o	These ranges of percentage retrieval uncertainties are now consistent with those presented elsewhere 		in the literature (e.g. for IWC in Hogan, Mittermaier and Illingworth, 2006).
