DHSVM Setup & Cal/Val


Setup and Calibration/Validation of DHSVM for Rios Jucu and Santa Maria da Vitoria

    DHSVM was initialized by running the entire simulation period ten times and saving the last date of the simulation. We then calibrated the model from 2009 – 2010 and validated the model for 2011 – 2012.  Calibration parameters were total soil depth, soil lateral and vertical hydraulic conductivities, and vegetation parameters. Values of vegetation parameters were initially set to correspond with Global Land Data Assimilation System (GLDAS) vegetation parameters mapped to the University of Maryland (UMD) classification scheme, and were then tuned to Southeastern Brazil based on descriptions by Allen et al. (2002), Chander and Markham (2003), and Galvincio, Correa and Araujo (2006).  Simulated hydrographs from January 2009 – December 2012 using the IEMA2007 landcover were compared to observed stream flows at Leopoldina for SMV and Fazenda Jucuruaba for the Jucu river. 

    The model captures the onset of the storm season, and estimated flows were relatively close to recorded values as denoted by the Nash-Sutcliffe efficiency measuring model performance (E=0.809 for Jucu and E=0.705 for SMV) Fig. 3.5 and Fig. 3.6 shows the pre-calibration and calibrated flows for both the Jucu and SMV. The average natural base flow was underestimated and this is can be explained by the fact that the model does not include deep groundwater components. The model is consistent in estimating dry season flows (Fig. 3.7) but has a wider-scatter of wet season flows. This indicates that the wet season flows are highly dependent on climate inputs (precipitation) during the storm period. We found that the model was most sensitive to precipitation inputs (see Tan et al., 2015), lateral soil hydraulic conductivity and soil depth. Final calibration parameters are listed in Table 1.


Table 1: Final values for DHSVM parameters after calibration and validation
Figure 1: SMV calibration and validation
Figure 2: Jucu calibration and validation
Figure 3: Comparison between calibrated and uncalibrated flows for Jucu (top) and SMV (bottom)
Figure 4: Average annual low-flow and high-flow comparison between simulated and observed discharge for Jucu and SMV