You are on page 1of 1

ID: 74205

Combining point and distributed snowpack data with landscape based discretization for hydrologic modeling of the snow-dominated Maipo River basin,
in the semi-arid Andes of Central Chile.
Videla, Yohann* & McPhee, James**

*Department of Geology, University of Chile, (yv.giering@gmail.com)


**Department of Civil Engineering, University of Chile, Chile (jmcphee@ing.uchile.cl)

Abstract

The 5000-km2 upper Maipo River Basin, in central Chiles Andes, has an adequate streamgage network but almost no
meteorological or snow accumulation data. Therefore, hydrologic model parameterization is strongly subject to model errors
stemming from input and model-state uncertainty. In this research, we apply the Cold Regions Hydrologic Model (CRHM) to
the basin, force it with reanalysis data downscaled to an appropriate resolution, and inform a parsimonious basin
discretization, based on the hydrologic response unit concept, with distributed data on snowpack properties obtained
through snow surveys for two seasons. With minimal calibration the model is able to reproduce the seasonal accumulation
and melt cycle as recorded in the one snow pillow available for the basin, and although a bias in maximum accumulation
persists, snowpack persistence in time is appropriately simulated based on snow water equivalent and snow covered area
observations. Blowing snow events were simulated by the model whenever daily wind speed surpassed 8 m/s, although the
use of daily instead of hourly data to force the model suggests that this phenomenon could be underestimated. We
investigate the representation of snow redistribution by the model, and compare it with small-scale observations of
wintertime snow accumulation on glaciers, in a first step towards characterizing ice distribution within a HRU spatial
discretization.
Although built at a different spatial scale, we present a comparison of simulated results with distributed snow depth data
obtained within a 40 km2 sub-basin of the main Maipo watershed in two snow surveys carried out at the end of winter
seasons 2011 and 2015, and compare basin-wide SWE estimates with a regression tree extrapolation of the observed data.

3.- Methodology - Distribution of mean annual rainfall

The forcing variables (precipitation and temperature) were estimated according to a topographic gradient. Then, using the
inverse distance interpolation method "IDW"(Teegavarapu et al., 2006), precipitation in each HRU can be calculated. This
way, the P_URH multiplier can be estimated based on a precipitation adjust observed in the neighboring stations according
to the following equation (Figure 3).
The use of correction factors or "multipliers precipitation" (P_mult table 1) in each HRU is a way to improve the estimation of
rain on them. Typically, such factors are determined as a function of the average elevation of the modeling unit or HRU.

5.- Results Snow modules

The results of the module "PBSM" for the simulated period are shown at Figure 5.
There is an overestimation of 12% above than observed in the SWE and Snow depth, however the accumulation period and
thawing speak were well modeled. Any relation to gravitational effects and wind exposure was only found at sites where the
snow accumulation is not explained by orographic moisture convergence. .

Table 1.- P mult. factor

1.- Study Area

In its upper course, the river runs as an entrenched torrent through the Andes mountains. Here, it receives three major
tributaries: El Volcn River, Yeso River and Colorado River. After leaving the Andes, the Maipo flows through the valley that
bears its name.

Figure 5.- Snow model evaluation

The mean anual discharge present a snowfall distribution character pattern, while the area is characterized by a mountain
climate (Figure 1)

6.- Results Percent of snow accumulation and SCA

The snow at each HRU was calculated and compared to the percentage of cells corresponding to MODIS-fractional images
at the same geographical coordinates.
The results show that there is a good correlation between the different products (figure 6).

Figure 3.- Simplification of the spatial distribution of mean annual rainfall

4.- Methodology - Flowchart simplified model

Figure 1.- Study area and hydrometeorological data

The meteorological variables were entered in the relevant modules, and calculated for each defined HRU (Figure 4),
allowing a unique connection between geomorphic and topographic parameters of the basin and hydrometeorological
observations.

2.- Methodology - Parameterizing modules and HRU's

Figure 6.- HRU's evaluation v/s Fractional Snow covered Area

The catchment area is modeled by Hydrologic Response Units (HRU's). The HRU's are characterized using topographic
variables, such as elevation, slope and aspect, and geographic variables, such as soil type, vegetation type and
precipitation distribution, through overlaying using a GIS-based approach..

7.- Compared with observed and Conclusions

This work's main objective was to model and identify the main driving processes of snow accumulation in the semiarid
Andes of central Chile, at different spatial configuration HUR's. This was compared and analysed with a distributed
snowcover dataset (figure 7), containing snow depth and density measurements along the region, in a relatively wide range
of altitude (2000 to 3900 m) and latitude (32.4S to 34.0S).

45 HRU's were created by combining physiographic criteria and are shown in Figure 2.

Figure 4.- Flowchart simplified model

Figure 2. HRU's configutation for CRHM Model

The model also supports the concept of distinct landscape elements (Hydrological Response Units or HRUs), which may be
linked episodically in process-specific sequences such as blowing snow, overland flow, organic layer subsurface flow,
mineral interflow, groundwater flow, and streamflow.
Figure 3.- Simplification of the spatial distribution of mean annual rainfall

Figure 7. HRUs comparation with Snow sampling points


These results follow an on-going effort of the University of Chile (RHMA) to build a novel regional dataset for testing and
implementing new modelling techniques for a better understanding of the spatial and temporal dynamics of snow in the
semiarid Andes Cordillera of central Chile.