Multi-scale assessment of interactions between surface water and groundwater fluxes in hard rock, water limited environment
Tanvir Hassan is a PhD student in the department of Water Resources. Promotor is dr. M.W. Lubczynski from the Faculty Geo-information and Earth Observation (ITC).
The multi-scale assessment of surface-groundwater interactions in hard rock (granite), water limited environments was the main objective of this PhD study carried out first in the ~80 km2 Sardon catchment (Western Spain) and finally in a small, 7.6 ha Trabadillo study area (TSA) extracted from that Sardon catchment. To address that objective, three different thematic studies, each with different specific objective, had been carried out, applying various techniques which included: field data acquisition, automated monitoring, remote sensing and numerical, integrated hydrologic modelling using GSFLOW code.
The first specific objective (Chapter 2) addresses the assessment of dynamics of surface-groundwater interactions at the Sardon catchment characterized by: (i) low catchment storage, shallow drainage base, and shallow water table; (ii) spatially dense, intermittent drainage system matching faults and fractures; (iii) high temporal variability of rainfall with common droughts and floods; and (iv) typical for the western Iberian Peninsula silvopastoral agroforestry land cover, in Spain known as dehesa, but in Portugal as montado. The GSFLOW model of the Sardon catchment (Chapter 2) was calibrated using 16 years data (14 head observations and streamflow measurement at the catchment outlet) and validated using 2 years data.
The main findings of that model simulation are: (1) intense groundwater exfiltration (especially near stream channels where groundwater is the shallowest), associated with abrupt water table rises, even up to the ground surface, triggered by erratic, intense rainfall events; the exfiltration substantially affected net recharge and water balance; (2) groundwater flow characterized by short flow-path and short residence time, both implying mosaic type of spatially intercalated recharge and discharge zones; (3) declining trend of the catchment groundwater outflow; and (4) large spatial and temporal variability of infiltration largely influenced by LULC type. That impact of land cover upon dynamics of surface-groundwater interactions, triggered expansion of that research, further continued in Chapter 3 and 4, but also underpinned some hardcoded deficiencies of GSFLOW, related to formulation of driving forces, mainly of rainfall interception loss.
That gave a rise to the second specific objective presented in Chapter 3, addressing spatiotemporal tree rainfall interception loss in the Sardon catchment. For that objective, experimental measurements of throughfall and stemflow were carried out during two hydrological years, on selected evergreen Quercus ilex (Q.i.) and deciduous Quercus pyrenaica (Q.p.) oaks, the only two tree species present in the Sardon catchment. These measurements, together with rainfall measurements, allowed to estimate tree rainfall interception loss, which was further temporally extrapolated using Gash model and spatiotemporally upscaled into the Sardon catchment, with help of very high-resolution images of QuickBird and WorldView-2. The main findings of this study are: (1) the yearly rainfall interception loss per tree species was larger in wet than in dry hydrological years (HY), but if expressed as % of rainfall, then it was the opposite; for example in dry HY2012, with rainfall = 335 mm, tree rainfall interception loss was 50.7% for Q.i. and 15.8% for Q.p., while in medium-wet HY2013 with = 672 mm, 45.7% for Q.i. and 9.8% for Q.p.; the rainfall interception loss of evergreen Q.i. was 3 - 4 times larger than of deciduous Q.p.; (2) species-dependent, biophysical Gash model parameters (further used in Chapter 4), i.e. canopy storage capacity (1.75 and 0.66 mm for Q.i. and Q.p., respectively) and fractional canopy cover (0.69 and 0.26 for Q.i. and Q.p., respectively) were defined; and (3) spatial (plot or catchment) dependence of tree rainfall interception loss on: i) species-specific, experimentally defined reference rainfall interception loss; and ii) canopy coverage per species definable on multispectral, very high resolution images (fundaments of the proposed method).
The Chapter 4 corresponding with the third specific objective, focusses on simulating surface-groundwater interactions in the TSA, at very fine spatial (5x5 m grid) and temporal (daily) resolution, using long time-series observation, to assess the net recharge dependence upon LULC type expressed through hydrological terrain units (hydrotopes). The input data for the TSA model have undergone lots of improvements as compared to the Sardon model input and longer, 20-year model calibration was constrained by more data types (hydraulic heads, soil moisture profiles, and MODIS product). The very high spatial resolution of the TSA model allowed simulation of surface-groundwater interactions as dependent on hydrotopes (e.g. area occupied by tree canopy or rock outcrop). The findings of this study confirmed all the main findings of the Chapter 2, but on top of them, added the following new findings: (1) large spatial, hydrotope-dependent variability of water fluxes despite of uniform rainfall input, primarily influenced by rainfall interception loss; (2) the applied hydrotope modeling approach with very high resolution grid allowed to quantify all water balance components including the net recharge per hydrotope; (3) the spatial definition of hydrotopes was largely dependent on species-dependent lateral root extent (in this study uncertain because assigned based on literature studies); and (4) the model simulation of the entire TSA, matched well the MODIS estimate of , except of incident simulated maxima larger than the MODIS maxima.
The presented PhD study identified various research gaps that need further actions. The largely unknown hydrological process of groundwater exfiltration, characteristic for shallow water table environments, when simulated at two different models applying two different spatial grid scales (Sardon model Chapter 2 and TSA model Chapter 4), in both cases, consistently provided substantial contribution of groundwater exfiltration to water balances; however, the obtained values were not field validated; therefore, experimental measurements of groundwater exfiltration are recommended. Also, more experimental measurements of reference tree rainfall interception loss at various tree individuals of different tree species are recommended for further validation of the proposed, novel method of remote sensing upscaling of tree rainfall interception loss. The well-known problem of uncertainty in simulation of groundwater evapotranspiration, deserves better software implementation and experimental validation. Also, improvements in root parameterization and in description of root architecture, especially of lateral root extent, are important and are needed for adequate definition of hydrotopes. Considering grassland hydrotope processes of rainfall interception loss and evapotranspiration, they were derived exclusively by remote sensing, so dedicated hydrological grassland studies are recommended, also because in water limited environments, grasslands represent dominant areas (in Sardon and in TSA more than 80% of the area), being responsible for relatively large water loss.
The hydrotope modeling study provided interesting and useful experience, showing dynamics of surface-groundwater interactions and hydrotope-dependent variability of net recharge; it also showed that trees reduce net recharge, having negative impact on water resources; as such, the integrated hydrological modelling at the hydrotope scale can be used not only for water management but also for landscape management, for example predicting impact of afforestation or deforestation on water table depth and on water resources. The follow up studies should target at even denser grids within all modeling domains. Currently such studies are realistic only at the plot-scale; but improvements in spatial resolution of remote sensing products and in computer efficiency, allow to expect that in near future; such studies will be realistic also at larger scales such as the catchment scale.
