Measuring and modelling spatiotemporal changes in hydrological response after partial deforestation

Abstract

Vegetation plays an important role in the hydrological cycle, as it governs the partitioning of water fluxes and therewith affects the functioning of the system. Deforestation can cause a highly non-linear response of the natural system and may change the interaction between the land surface and the atmosphere, flow conditions, groundwater recharge and soil moisture storage, which in turn affects the quality and amount of available water resources. To be able to predict changes of deforestation and other land use management activities, there is a need for comprehensive understanding of the hydrological effects of such activities. Although the effects of land use change on hydrology have been studied intensively, predicting the effects of land use change on hydrological states and fluxes remains challenging. Existing paired catchment studies mostly focus on yearly discharge, often do not consider changes in subsurface storage and evapotranspiration, and lack information at the intra-annual time scale. Additionally, soil hydrological processes are often not considered. Thus, only few datasets are available to accurately describe, model, and predict detailed changes in spatiotemporal patterns of hydrological fluxes and states due to land use change. The aim of this thesis is to improve understanding of the rapid system changes related to deforestation by analysing an innovative dataset and evaluating the predictive ability of a distributed hydrological model. In order to achieve this aim, four steps that represent the individual sub-aims of this project were followed. In the first step, hydrological changes in spatiotemporal fluxes related to partial deforestation measures were defined with a focus on discharge, actual evapotranspiration, and soil moisture. In a second step, the spatial and temporal characteristics of water movement in the vadose zone (piston flow, preferential flow) and the factors that control these processes were assessed. In a third step, changes in spatial and temporal characteristics of water movement in the vadose zone related to partial deforestation were defined. In a final step, the effects of partial deforestation were simulated with a distributed hydrological model (ParFlow-CLM) and were compared with the observed changes to test the predictive ability of the model. For this thesis, data from the Wüstebach catchment established within the TERENO (TERrestrial Environmental Observatories) network in Germany have been used. This catchment provides an unique monitoring setup to investigate the major components of the water balance (evapotranspiration, discharge, precipitation) and the spatiotemporal distribution of soil moisture before and after a partial deforestation. Given the large amount of previous work, the thesis starts with an overview of the state-of-the-art for investigating the hydrological impact of deforestation and other land use changes with a focus on the impacts on discharge, actual evapotranspiration and soil moisture storage. After evaluating data-driven studies, existing modelling studies are evaluated by comparing the study area characteristics, the applied hydrological models, and the calibration and validation procedures. Next, the study area is introduced and the measurement setup in the Wüstebach catchment is explained in detail. To put this thesis in the context of previous work, the main insights about this catchment obtained in previous studies is also briefly reviewed. To analyse the hydrological impact of deforestation, five years of measured hydrological data from the Wüstebach catchment were analysed, including all major water budget components three years before and two years after a partial deforestation. A data-driven approach was used to understand changes and related feedback mechanisms in spatiotemporal hydrological response patterns. As expected from earlier studies, it was found that partial deforestation caused a decrease in evapotranspiration and an increase in discharge. A closer look at the high-resolution datasets revealed new insights into the intra-annual variability and relationships between the water balance components. The overall decrease in evapotranspiration caused a large increase in soil water storage in the deforested region, especially during the summer period, which in turn caused an increase in the frequency of high discharge in the same period. Although the evapotranspiration in the forested region was larger on average, the deforested region showed a higher evapotranspiration during part of the summer period on several occasions. This was related to the wetter conditions in the deforested area accompanied with the emergence of grass vegetation. At the same time, wetter soil moisture conditions in the deforested area increased the spatial variance of soil moisture in the summer and therewith altered the relationship between spatial mean and variance. Altogether, this data-based analysis illustrates that detailed spatiotemporal monitoring can provide new insights into the hydrological effects of partial deforestation. Next, soil moisture sensor response time analysis was used on the 5-year soil moisture monitoring dataset to identify factors that control preferential and sequential flow before and after partial deforestation. For this, the sensor response times at 5, 20 and 50 cm depth were classified into one of four classes: (1) non-sequential preferential flow, (2) velocity-based preferential flow, (3) sequential flow, and (4) no response. For the three years before deforestation, it was found that the spatial occurrence of preferential flow was governed by small-scale soil and biological features and local processes, and showed no obvious relationship with any of the selected catchment-wide spatial attributes. Event-based occurrence of preferential flow was highly affected by precipitation amount, with a nearly catchment-wide preferential response during large storm events. During intermediate events, preferential flow was controlled by small-scale heterogeneity, instead of showing catchment-wide patterns. The effect of antecedent catchment wetness on the occurrence of preferential flow was generally less profound, although a clear negative relationship between the mean soil moisture content and the percentage of preferential flow was found for precipitation events larger than 25 mm. Overall, the results of this analysis before deforestation demonstrate that sensor response time analysis can offer insights into the spatio-temporal interrelationships of preferential flow occurrence. In a next step, sensor response time analysis was also applied to the 2- year soil moisture monitoring dataset obtained after partial deforestation to analyse the effects of the partial deforestation on flow conditions in the vadose zone of the Wüstebach catchment. Results of this analysis showed that partial deforestation increased sequential flow occurrence and decreased the occurrence of no flow in the deforested area. Similar precipitation conditions after deforestation caused more sequential flow in the deforested area, which was related to higher antecedent moisture and missing interception. Results of this analysis demonstrated that the combination of a sensor response time analysis and a soil moisture dataset that includes pre- and post-deforestation conditions can offer new insights in preferential and sequential flow conditions after land use change. In a final step, the five-year long hydrological dataset was used to evaluate the ability of the ParFlow-CLM model to predict hydrological effects of partial deforestation. ParFlow-CLM simulations in the Wüstebach catchment were performed for a three year spin-up period, a three year control period where the entire catchment was forested and a two year post-treatment period, where the hydrological effects of partial deforestation were simulated. The results showed that ParFlow-CLM did not only capture low and intermediate discharge conditions, but was also able to correctly predict evapotranspiration fluxes in the catchment before and after partial deforestation. Also, observed spatiotemporal soil moisture patterns and post-deforestation related changes were fairly well represented. At the same time, this model evaluation informed about current model limitations that could be improved to obtain even better predictions. Modelling results have shown that the global plant parameterization strategy within CLM 3.5 may not always be directly transferrable to small catchments. Peak flow conditions and observed soil wetness increases after deforestation were underestimated by the model. This could be addressed by improving the soil parameterization and the soil process description (preferential and lateral flow). Overall, the results of this model application in non-stationary conditions clearly illustrate the potential of distributed hydrological models to forecast non-linear system changes. The thesis concludes with a synthesis of the main outcomes and a discussion of possible future research activities. Overall, the combination of the Wüstebach dataset and the ParFlow-CLM model simulations have provided new hydrological insights in spatiotemporal system changes related to deforestation. Extrapolation of this study to other research areas and other modelling platforms could provide new understanding on the hydrological effects of deforestation and other land use change related activities.