@phdthesis{1258639,
  abstract     = {{The drylands of Chile are faced with continuous drought hazards that reduce their production potential and limit sustainable development. On the other hand, microcatchment harvesting has been promoted as a measure to reduce climate risk in drylands, but a quantitative assessment of their water use efficiency under a range of soil physical and climatic conditions has not yet been performed for those techniques actually used in Chile. 
The research in this dissertation was therefore divided in three main parts. In Part I the climatic uncertainty in the central-northern region of Chile was identified to evaluate the potential for the development of a drought early warning system. First, daily winter rainfall characteristics were compared for 42 stations in the region, and showed clear regional homogeneity as well as a strong interaction between seasonal rainfall amount and ENSO, suggesting potential predictability. This was further evaluated using three modelling approaches with increasing complexity. First, a simple univariate index of ENSO was found only well correlated for simultaneous (May or MJJA) values of the index, and thus not useful for prediction since lead times are insufficient for drought prediction. 
In a second approach, three General Circulation Models (GCMs) were used to forecast winter season rainfall amounts by a Canonical Correlation Analysis (CCA), finding the Climate Forecast System or CFS model most adequate for prediction purposes, due to its small bias and high skill. In a last approach, the rainfall field of the CFS was then used as a predictor for generating stochastic daily precipitation series in a non-homogeneous Hidden Markov Model (nHMM), from which seasonal rainfall characteristics were then calculated. 
The latter two methods produced similar and high correlations when hindcasting the seasonal precipitation in the period 1980-2005. They were also used to hindcast the Standardized Precipitation Index, and were able to reproduce the main seasonal characteristic (dry, normal or wet) in 88-92% of the time. This indicates that the proposed methodology is effective in predicting drought for the region adequately.
The last approach has, however, an added advantage for dryland management, because it can be used to derive daily information more relevant for dryland stakeholders, such as the accumulated daily precipitation deficit or the frequency of days exceeding a threshold precipitation (e.g. 15 mm dayP-1P).  Ultimately, these predictions can also be coupled with water balance or crop models to transform the current climatic drought early warning system towards hydrologic or agronomic applications. 
In the second part of this dissertation another source of variability in the central-northern Chilean drylands was investigated, focussing on the heterogeneity of soil physical properties in soils with high stoniness. This was especially important for the modelling approach developed in Part III, which requires initial estimates of hydrophysical model parameters. Therefore, a comparison was made between six different measurement methods to determine the saturated hydraulic conductivity (KRfsR), in order to evaluate how the selection of a particular method can influence the KRfsR estimate, as well as to compare the applicability of the different methods on soils with embedded stone fragments and suggest a reference method. For each measurement method, also a range of calculation procedures for KRfsR were compared that could add to this parameter uncertainty. 
A clear relation was found between the different measurement methods, consistent over three different locations. The single (SR) and double ring (DR) infiltrometers resulted in similar, not significantly different values of KRfsR that were significantly higher whenR Rcompared to the other four methods. This would suggest that the combination of ponding water and ring insertion, altering the soil structure in the immediate surroundings of the ring, was not sufficiently accounted for by these methods in the case of stony soils. The two borehole methods applied, the inverse auger hole (IA) and constant head well (CH) method, also resulted in similar KRfsR estimates, but proved difficult to apply in soils with high stoniness, such as at the Romeral location. The remaining two methods, the rainfall simulator and tension infiltrometer, were characterized by their non-invasive character, as such providing a flawless method to determine KRfsR in these soils. When taking the water requirement into account, it is clear that the tension infiltrometer can be proposed as a reference method for stony soils.
The effect of embedded stone fragments on hydrophysical characteristics was then investigated with more detail along a transect in a dryland watershed with highly variable gravimetric rock fragment content (RRmR) of 0.05 0.75 g gP-1P. A positive relation was found between RRmR and hydraulic conductivity, especially for high pressure potentials and under saturated conditions. The spatial variability of hydraulic conductivities measured along three transects in the watershed could be mainly related to differences in stone fragment contents, showing higher values on the eastern, drier slopes, in the riverbed, and on the steeper slopes, that are more prone to soil loss processes.
On the other hand, soil water retention was negatively influenced by stone fragment contents, since the additional macro pore space cannot contribute sufficiently to compensate for the loss of soil matrix when replaced by the rock fragments. In accordance to the observations for KRfsR, the stone fragment content influence on soil water retention was largest near saturation and for the lower pressure potentials up to -30 kPa, after which the influence of rock fragments became limited. 
Finally, in Part III different aspects of microcatchment water harvesting were investigated. As a transition from the previous part, we focused on the impact of infiltration trenches and reforestation on soil hydrophysical properties in two adjacent watersheds with different management. Thirteen years after implementation of infiltration trenches in a dryland watershed, clear differences in Normalized Difference Vegetation Index (NDVI) were observed when compared to an unmanaged watershed. When extending the analysis to soil properties, some differences were encountered. Opposite slopes in the watersheds had a different particle size distribution, with the slopes more exposed to wind-driven rains showing a depletion of erodible fine sand fraction, and this was more pronounced in the unmanaged watershed. As could be expected, in the managed watershed the difference between the slopes was less, suggesting that the soil conservation techniques and reforestation have decreased the soil erosion hazard.  
The influence of infiltration trenches on soil properties was less pronounced. The clay fraction inside the trenches tended to be higher than outside the trenches, but was not significant for all trenches. Organic carbon and nitrogen content were higher inside the trenches, but no significant differences in other properties, such as KRfsR and available soil water capacity, could be established.
A new framework was developed to evaluate the efficiency of microcatchment systems using a three-dimensional fully coupled surface-subsurface soil hydrological model, HydroGeoSphere (HGS). In a first step, the use of a rainfall simulator in combination with detailed soil water content measurements was evaluated to generate the data set required to auto-calibrate the HGS model. A sensitivity analysis indicated that six parameters contribute to the model solution, while three others were insensitive and were fixed to avoid model non-uniqueness. The remaining parameters proved not well correlated, as a requirement for parameter identifiability in the final model. As a last step in the evaluation of model well-posedness, data sets required to calibrate the model univocally were assessed for three different objective functions. Using both measured runoff data and soil moisture contents resulted in the lowest difference of modelled versus observed values. As a result, the calibrated model fitted well to the observed data series of soil moisture and runoff (correlation coefficient of 0.97), and a response surface analysis confirmed reaching a well defined global minimum during the inversion process. 
The final model also effectively simulated trench overfilling, while the moisture content profile of the slope showed a marked increase in the water content just below and near to the infiltration trench, confirming the water harvesting potential of these techniques for reforestation purposes. The spatial extent of this increase was limited to the immediate surroundings of the trench. Nevertheless, model simulations with different characteristic rainfall seasons typically observed in the region showed clearly that rainfall intensities were insufficient to generate enough runoff for efficient water harvesting, except during an extremely wet year. This indicates that the climatic and soil physical constraints operating at that location limit the applicability of the microcatchment technique.
This was further evaluated for a hillslope at a much more southern location, where the newly developed framework was used to determine water harvesting design schemes, considering interaction of different infiltration trenches. 
The hillslope model with six infiltration trenches was calibrated and validated using runoff observations from simulated and natural rainfall events during the rainfall seasons from 1996 to 1998, conserving general infiltration-runoff patterns. Under these climatic and soil physical conditions, important runoff reductions (44–90%) were simulated when compared to a bare plot without infiltration trenches, but significant runoff amounts could not be buffered under the (very) wet conditions occurring at least every eight years.
As such, two different strategies were developed to improve the design of the infiltration trenches for the study location. First, the empirical MAUCO approach was tested as a design method, resulting in a considerably drop in runoff amounts in both dry and wet years, confirming that this approach can be used to design soil conservation measures. Secondly, a new approach was then tested to select a correct infiltration trench design for an Eucalyptus globulus forest plantation with the HGS, based on a detailed water balance analysis. In order to increase the plant beneficial water fluxes (transpiration), the trench distance of 4m resulted in an acceptable trade-off between potential yield per tree and limited losses by runoff, deep drainage, evaporation, non-target transpiration and open water evaporation.
Finally, the evaluation of microcatchment systems was oriented towards aspects of tree survival and productivity. Here, three different microcatchment treatments (infiltration trenches, microterraces and rototilling) and two combined treatments were tested on tree plantation of Eucalyptus globulus and Pinus radiata in the central Chilean drylands. First, the behaviour of the soil water potential directly below the different treatments was compared, showing higher moisture contents during a longer period under the microterraces and on the plot where rototilling was observed when compared to a control plot without treatments, and these differences were consistent for the two different measurement locations. The moisture content was highest near to the trench (0.6 m), decreasing rapidly at larger distances, confirming the rather small spatial extent of the trench influence on water availability and tree growth.
Secondly, the different treatments showed a clear improvement of Eucalyptus tree survival rates after two dry seasons, while this was less clear for Pinus trees, due to the higher vulnerability of Eucalyptus to prolonged dry spells during establishment. Tree biomass proved (significantly) higher for the plot with microterraces and rototilling, but this was more pronounced for Eucalyptus than for Pinus trees. The trench had only a small effect on tree productivity, mostly on trees near to the trench. 
From this dissertation, it is clear that these techniques can result in additional sources of soil water for increased plant productivity, if they are correctly dimensioned and implemented.}},
  author       = {{Verbist, Koen}},
  isbn         = {{9789059894457}},
  keywords     = {{climatic variability,Rain water harvesting,Chile,drought,stoniness}},
  language     = {{eng}},
  pages        = {{III, XXIII, 224}},
  publisher    = {{Ghent University. Faculty of Bioscience Engineering}},
  school       = {{Ghent University}},
  title        = {{Climatic and soil physical constraints for efficient rain water harvesting in degraded lands of Chile}},
  year         = {{2011}},
}