This directory contains:





a) Figures 3, 4, and 5 from the article for all S4* models





b) Figure 7 shows per spatial level for each S4* model (with respect to the monthly mean streamflow forecasts at 0-month lead time):


	
	1) the sample autocorrelation function

	
	2) a quantile plot against the Gaussian distribution


	
	both for paired differences of absolute residuals with respect to the preMet model (see Eq. 3 in the article)


	
	3) a scatterplot of predictions versus observations





c) Figure 8 shows for the refRun model and the y_{i,30} predictand at spatial level one
:



	1) the regression coefficients (for predictors and predictands standardised to mean 0 and standard deviation 1)

	
	2) the predictors aggregation periods a_{i,j} (see Eq. 1 in the article)


	
	n = 100 due to the bootstrap resampling