Rößler, Ole: A Climate Change Impact Assessment Study on Mountain Soil Moisture with Emphasis on Epistemic Uncertainties. - Bonn, 2011. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5N-26000
@phdthesis{handle:20.500.11811/5013,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5N-26000,
author = {{Ole Rößler}},
title = {A Climate Change Impact Assessment Study on Mountain Soil Moisture with Emphasis on Epistemic Uncertainties},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2011,
month = aug,

note = {Mountains are expected to respond sensitive to climate change. Thus, sound climate change impact assessment studies focusing on mountain areas are strongly needed to estimate changes and to develop adaptation strategies. Nowadays, climate change impact assessment studies (CCIAS) are a common approach and many publications on hydrological responses to climate change have been published. Nonetheless, CCIAS focusing on soil moisture are widely missing especially at the catchment scale; even more, as to our knowledge there are only two studies on mountain soil moisture at a coarse scale. The wide neglect of soil moisture in climate change impact assessment studies contrasts the key role of soil moisture in ecosystems. This clearly shows the strong demand for CCIAS on mountain soil moisture. In this study, a commonly used CCIAS approach was used, comprising (1) of a physically based model that was calibrated and validated under recent climate conditions, (2) that was driven by downscaled regional climate models (RCMs) for a reference and a future scenario climate conditions. A major challenge in CCIAS is the propagation of uncertainties that questions the model results. In this study a special focus is set on the structural uncertainties originating from the use of downscaling approaches and climate models. Therefore, an analytic framework was developed based on the both concepts of uncertainty propagation and the uncertainty cascade. The concept comprehensively summarizes all uncertainties occurring in climate change impact assessment studies and illustrates how the uncertainties propagate. We conducted the CCIAS in a mountain catchment (160 km²) in the Swiss Alps at a high spatial resolution (50m). At first, the frequently used, physically based, distributed hydrological model was successfully applied to the catchment for recent years (2001-2007) to provide a sound calibration and validation. The potentials and the limitations of WaSiM-ETH to simulate soil moisture dynamics and patterns were shown by comparing model results with extensive soil moisture measurements at an hourly time step. While WaSiM-ETH was able to reproduce discharge with a high accuracy (R² = 0.95, ME = 0.8, IoA = 0.95), the simulation of soil moisture for different altitudes and land use types is partly limited, since the model was unable to model the total variability of the soil moisture dynamic, but tended to mean values. An adjusted RMSE of 8.0 Vol-% that takes the intra-plot variability into account was calculated for soil moisture. A necessary prerequisite is the validation of the ability of the downscaled RCM data to drive the hydrological model in such that the hydrological processes are reproduced. A comparative study was conducted based on two common downscaling approaches (statistical downscaling (SD) and direct use (DU)) and two RCMs (CHRM, REMO). Uncertainties were found to be unsteadily distributed, both in terms of variables and time. The “one” model approach that shows least uncertainty for all kinds of hydrological variables like discharge, actual evapotranspiration, and soil moisture was not found. This finding adds considerable value to the scientific discussion, since most previous studies focus on one variable or one downscaling approach alone. In addition, we evaluated the spatial uncertainties of soil moisture and evapotranspiration. We showed that the choice of downscaling approaches is of circumstantial relevance for discharge and water balance, while for all spatial variables, we found SD approaches to perform better than DU approaches. Next, we simulated the impact of climate change on mountain soil moisture by applying three different downscaling approaches and two RCMs. In addition to the SD and DU-models, the very popular delta change approach (Δ) was applied that scales the climate observation by adding the climate signal. Therefore, uncertainty assessment for the Δ-approach was not necessary. The use of multiple downscaling techniques in an ensemble forecast is new for soil moisture impact studies. The study proved the partly superior role of downscaling approaches when focusing on the impact per se under future climate and thereby contrasting findings of recent publications. Moreover, it questions results from studies that are based on one downscaling approach alone. The study provided detailed data on climate change impact on the hydrology of the catchment that are completely in line with previous findings. The high spatio-temporal resolution of the study add value to previous mountain soil moisture studies of Jasper et al. (2004, 2006) by providing site specific data on soil moisture decrease and drought stress potential at the catchment scale. The consensus of six models driven by two threefold downscaled RCM reveals the forested areas below 1800 m a.s.l. to be most affected by climate change in 2070-2100 (-10 vol-%). The variability of the results from the six ensembles were remarkably high, offering a bandwidth of possibilities from nearly unchanged soil moisture conditions to strong expansion of drought stress in the future. In addition we found uncertainties from the applied hydrological model and downscaling approaches in the magnitude of the predicted changes (+/- 10 vol-%). Therefore, the results have to be interpreted carefully. Probabilistic forecasting with several hundred model runs might confirm the found tendency of soil moisture decrease in future studies.},
url = {https://hdl.handle.net/20.500.11811/5013}
}

The following license files are associated with this item:

InCopyright