To the root of vegetation-water interactions: Improving spatiotemporal variations in global models

Vegetation strongly influences evaporation from land by transporting water from the subsurface to the atmosphere through root water uptake. The amount and timing of this water flux depends on the aboveground (e.g., the amount of leaves) and belowground (e.g., the root extent) characteristics of the vegetation. Although these vegetation characteristics vary strongly both in space and time, there is lack of adequate representation of this vegetation variability in large scale hydrological and land surface models. This causes deficiencies in representing the associated variability in modeled water and energy states and fluxes, which introduces uncertainties in future predictions of the hydrological cycle, including hydrological extremes such as droughts and floods. To address this issue, this research aims to develop more realistic model representations of spatial and temporal vegetation variability, and explore their potential for improving modeled water fluxes in large scale hydrological and land surface models.

Chapter 2 focuses on model representations of spatial and temporal variability of aboveground vegetation characteristics based on satellite remote sensing data. Interannual variability of land cover and leaf area index (LAI) from latest global remote sensing datasets are integrated into the land surfacemodel Hydrology Tiled ECMWF Scheme for Surface Exchanges over Land (HTESSEL). Furthermore, datasets of LAI and the fraction of green vegetation cover are used to develop and integrate a spatially and temporally varying model parameterization of the effective vegetation cover. The effects of these three implementations on simulated hydrology are evaluated using offline (land-only) model simulations. The results show that the enhanced variability of aboveground vegetation characteristics considerably improves the simulated variability of evaporation and near-surface soil moisture. These improvements are connected to a framework that describes how the implemented vegetation variability influences internal model interactions between vegetation, soil moisture, and evaporation.

Chapter 3 evaluates how climate-controlled root zone parameters influence water flux simulations with the land surfacemodel HTESSEL. To this aim catchment scale root zone storage capacity Sr (mm), defined as the maximum volume of subsurfacemoisture that can be accessed by the vegetation roots, is estimated using the memory method. In this method Sr is derived from soil water deficits, reflecting the ability of vegetation to adapt to the local climate conditions by sizing their roots in such a way to guarantee continuous access to water, keeping memory of past water deficit conditions. Climatecontrolled Sr is estimated with the memory method for 15 catchments in Australia to adequately represent the spatial variability of the vegetation roots. These estimates are integrated into HTESSEL, replacing the static root representation based on soil types and uniform soil depth. The results of offline model simulations show that climatecontrolled Sr representation significantly improves the timing of modeled discharge in the study regions. This suggests that a climate-controlled representation of the model Sr has potential for improving water flux simulations by land surface models in a global context.

Chapter 4 presents the influence of irrigation on the estimation of Sr with the memory method. The memory method Sr is derived from the seasonal patterns of root zone water input and output. Besides precipitation as input, irrigation supplies additional water to the root zone in irrigated agricultural fields. However, the influence of irrigation on the memory method Sr estimates has not been assessed previously. In this study two methods based on different globally available irrigation datasets are developed to account for irrigation in the memory method for estimating Sr. The Sr estimates fromthese two methods are compared to a case without considering irrigation for a large sample of catchments globally. The results show, for the first time, that irrigation considerably reduces Sr in regions with extensive irrigation, highlighting the relevance of irrigation for adequately estimating ecosystem scale Sr.

Chapter 5 investigates the influence of climate, landscape, and vegetation variables on Sr globally. So far, there is limited insight on the controls of global-scale root development and their spatial variation. A random forest model is used to predict Sr as estimated with the memory method based on 21 variables for a large sample of catchments globally. The results indicate that hydro-climatic variables are the dominant, but spatially varying, driver of ecosystemscale Sr, while landscape and vegetation play aminor role. Based on the importance of the various drivers, a reduced parsimoniousmodel using four variables is used to predict Sr on a global scale. These predictions largely resemble other global estimates of root characteristics based on more complex methods and datasets. This indicates that the here developed parsimonious model to estimate global scale Sr based on four simple globally available variables adequately represents the spatial variability of Sr globally. Together with the results from Chapter 2, it can be concluded that integration of these estimates into large scale hydrological and land surface models has potential to improve model water fluxes.

The findings of this dissertation directly contribute to the large scale hydrological and climate model communities by providing methods to adequately represent spatial and temporal vegetation variability. The results demonstrate the potential of these methods to improve modeled water fluxes by large scale hydrological and land surface models, with major implications for the accuracy of hydrological and climate predictions. This dissertation lays the foundation for future research aimed at further improving the realism of model vegetation variability.