Refine
Document Type
Keywords
- Autocorrelation (1)
- Disaggregation (1)
- Intra-event intermittency (1)
- Precipitation event (1)
- Rainfall generator (1)
- Within-event variability (1)
Institute
In landscapes with heterogeneous vegetation structure, interception and throughfall patterns produce spatiotemporal
variability of soil moisture. This variability is important for eco-hydrological processes, in particular on
small spatial scales up to the catchment scale. Throughfall depends on vegetation structure, whereas vegetation
development is presumably co-determined by the spatio-temporal distribution of throughfall itself. In addition
to vegetation structure, meteorological factors like wind speed and rainfall intensity also have an impact on
throughfall.
The objective of this study is to quantify the influence of vegetation structure and meteorological variables
on spatial (and in the long run the temporal) variability of throughfall. For that purpose, we developed an
approach combining field methods, image analysis and multivariate statistics. The 6-ha constructed catchment
‚Hühnerwasser‘ (aka Chicken Creek, southern Brandenburg, Germany) offers ideal conditions for the investigation
of eco-hydrological feedback processes. After more than 10 years of development, vegetation structure on the
catchment is spatially heterogeneous and evolves through natural succession. Furthermore, complementary
meteorological data are available on-site.
Throughfall was measured using 50 tipping-bucket rain gauges, which are aligned along two transects in 0.5
and 1 m heights, covering the dominating vegetation types on the catchment (e.g., robinia, sallow thorn, reed,
reedgrass, herbs). The spatial distribution of vegetation structures around each measurement site was recorded
with hemispheric photographs, which were subsequently analyzed using image processing techniques. Two
weather stations provide reference values for precipitation and relevant meteorological variables for wind speed
and direction, air humidity, temperature and irradiation.
The amount and distribution of precipitation measured in scarcely vegetated areas of the catchment widely
correspond with values from the reference weather stations. Under dense vegetation, very heterogeneous values
were recorded, which can be explained by i) canopy interception, and ii) fetching effects. The results of this study
can serve as basis for interception models and may also contribute to complex eco-hydrological models.
Rainfall variability within a storm is of major importance for fast hydrological processes, e.g. surface runoff,
erosion and solute dissipation from surface soils. To investigate and simulate the impacts of within-storm variabilities on these processes, long time series of rainfall with high resolution are required. Yet, observed precipitation records of hourly or higher resolution are in most cases available only for a small number of stations and only for a few years. To obtain long time series of alternating rainfall events and interstorm periods while conserving the statistics of observed rainfall events, the Poisson model can be used. Multiplicative microcanonical random cascades have been widely applied to disaggregate rainfall time series from coarse to fine temporal resolution.
We present a new coupling approach of the Poisson rectangular pulse model and the multiplicative microcanonical random cascade model that preserves the characteristics of rainfall events as well as inter-storm periods. In the first step, a Poisson rectangular pulse model is applied to generate discrete rainfall events (duration and mean intensity) and inter-storm periods (duration). The rainfall events are subsequently disaggregated to high-resolution time series (user-specified, e.g. 10 min resolution) by a multiplicative microcanonical random cascade model. One of the challenges of coupling these models is to parameterize the cascade model for the event durations generated by the Poisson model. In fact, the cascade model is best suited to downscale rainfall data with constant time step such as daily precipitation data. Without starting from a fixed time step duration (e.g. daily), the disaggregation of events requires some modifications of the multiplicative microcanonical random cascade model proposed by Olsson (1998): Firstly, the parameterization of the cascade model for events of different durations requires continuous functions for the probabilities of the multiplicative weights, which we implemented through sigmoid functions. Secondly, the branching of the first and last box is constrained to preserve the rainfall event durations generated by the Poisson rectangular pulse model.
The event-based continuous time step rainfall generator has been developed and tested using 10 min and hourly rainfall data of four stations in North-Eastern Germany. The model performs well in comparison to observed rainfall in terms of event durations and mean event intensities as well as wet spell and dry spell durations. It is currently being tested using data from other stations across Germany and in different climate zones. Furthermore, the rainfall event generator is being applied in modelling approaches aimed at understanding the impact of rainfall variability on hydrological processes.
In evolving initial landscapes, vegetation development depends on a variety of feedback effects. One of the less understood feedback loops is the interaction between throughfall and plant canopy development. The amount of throughfall is governed by the characteristics of the vegetation canopy, whereas vegetation pattern evolution may in turn depend on the spatio-temporal distribution of throughfall. Meteorological factors that may influence throughfall, while at the same time interacting with the canopy, are e.g. wind speed, wind direction and rainfall intensity. Our objective is to investigate how throughfall, vegetation canopy and meteorological variables interact in an exemplary eco-hydrological system in its initial development phase, in which the canopy is very heterogeneous and rapidly changing. For that purpose, we developed a methodological approach combining field methods, raster image analysis and multivariate statistics. The research area for this study is the Hühnerwasser (‘Chicken Creek’) catchment in Lower Lusatia, Brandenburg, Germany, where after eight years of succession, the spatial distribution of plant species is highly heterogeneous, leading to increasingly differentiated throughfall patterns. The constructed 6-ha catchment offers ideal conditions for our study due to the rapidly changing vegetation structure and the availability of complementary monitoring data.Throughfall data were obtained by 50 tipping bucket rain gauges arranged in two transects and connected via a wireless sensor network that cover the predominant vegetation types on the catchment (locust copses, dense sallow thorn bushes and reeds, base herbaceous and medium-rise small-reed vegetation, and open areas covered by moss and lichens). The spatial configuration of the vegetation canopy for each measurement site was described via digital image analysis of hemispheric photographs of the canopy using the ArcGIS Spatial Analyst, GapLight and ImageJ software. Meteorological data from two on-site weather stations (wind direction, wind speed, air temperature, air humidity, insolation, soil temperature, precipitation) were provided by the ‘Research Platform Chicken Creek’. Data were combined and multivariate statistical analysis (PCA, cluster analysis, regression trees) were conducted using the R-software to i) obtain statistical indices describing the relevant characteristics of the data and ii) to identify the determining factors for throughfall intensity. The methodology is currently tested and results will be presented. Preliminary evaluation of the image analysis approach showed only marginal, systematic deviation of results for the different software tools applied, which makes the developed workflow a viable tool for canopy characterization. Results from this study will have a broad spectrum of possible applications, for instance the development / calibration of rainfall interception models, the incorporation into eco-hydrological models, or to test the fault tolerance of wireless rainfall sensor networks.
To simulate the impacts of within-storm rainfall variabilities on fast hydrological processes, long precipi-tation time series with high temporal resolution are required. Due to limited availability of observed data such time series are typically obtained from stochastic models. However, most existing rainfall models are limited in their ability to conserve rainfall event statistics which are relevant for hydrological pro-cesses. Poisson rectangular pulse models are widely applied to generate long time series of alternating precipitation events durations and mean intensities as well as interstorm period durations. Multiplicative microcanonical random cascade (MRC) models are used to disaggregate precipitation time series from coarse to fine temporal resolution.
To overcome the inconsistencies between the temporal structure of the Poisson rectangular pulse model and the MRC model, we developed a new coupling approach by introducing two modifications to the MRC model. These modifications comprise (a) a modified cascade model (“constrained cascade”) which preserves the event durations generated by the Poisson rectangular model by constraining the first and last interval of a precipitation event to contain precipitation and (b) continuous sigmoid func-tions of the multiplicative weights to consider the scale-dependency in the disaggregation of precipita-tion events of different durations. The constrained cascade model was evaluated in its ability to dis-aggregate observed precipitation events in comparison to existing MRC models. For that, we used a 20-year record of hourly precipitation at six stations across Germany. The constrained cascade model showed a pronounced better agreement with the observed data in terms of both the temporal pattern of the precipitation time series (e.g. the dry and wet spell durations and autocorrelations) and event char-acteristics (e.g. intra-event intermittency and intensity fluctuation within events). The constrained cas-cade model also slightly outperformed the other MRC models with respect to the intensity-frequency relationship.
To assess the performance of the coupled Poisson rectangular pulse and constrained cascade model, precipitation events were stochastically generated by the Poisson rectangular pulse model and then disaggregated by the constrained cascade model. We found that the coupled model performs satisfac-torily in terms of the temporal pattern of the precipitation time series, event characteristics and the in-tensity-frequency relationship.
Experimental catchments with well-known boundaries and characteristics may contribute valuable data to hydrological, critical zone and landscape evolution research. One of the most well-established and largest constructed catchments is the Chicken Creek catchment (6 ha area including a 0.4 ha pond, Brandenburg, Germany) representing an initial ecosystem undergoing a highly dynamic ecological development starting from clearly defined starting conditions. The water balance dynamics of the catchment was calculated using a simple mass balance approach to reveal the impact of ecological development during 12 years. Water storage in the catchment was calculated from a 3D-model of groundwater volumes, soil moisture measurements and water level recordings of the pond. The catchment water balance equation was resolved for evapotranspiration, the only part that was not measured directly. Due to the known boundary conditions and the inner structure of the catchment, we were able to quantify the different storage compartments and their role in hydrologic response. Our results indicate that for small catchments with a highly dynamic ecological development like the Chicken Creek, the knowledge about saturated and unsaturated storage volumes enables a good estimate and closure of the water balance using a rather simple approach, at least in annual resolution. We found a significant relationship between vegetation cover in the catchment and calculated ET. Time series of meteorological, hydrological, soil and vegetation data over 12 years enabled us to characterize the transient development of the catchment and to evaluate the effect of different feedback mechanisms on catchment hydrology. The dataset from the Chicken Creek catchment indicate at least three phases in ecosystem development, where initial abiotic feedbacks (e.g. erosion) were followed by more and
more biotic controls (e.g. biological soil crusts, vegetation succession and growth). Data from Chicken Creek in high spatial and temporal resolution provide a valuable database underlining the high importance of abiotic/biotic feedback effects that change the hydrologic functioning and response of the catchment more than the water balance itself revealed and thus have to be included in catchment models.
Difficulties in quantitatively closing the water balance of catchments arise when upscaling point measurements and from insufficient knowledge of the physical boundaries, inner structure, and storage volumes of natural catchments. In addition, there is a strong need for generalizing the relationship between catchment characteristics and hydrological response. Therefore, experimental catchments with well-known boundaries and conditions could contribute valuable data to hydrological and critical zone research. One of the most well-established and largest constructed catchments is the Chicken Creek catchment (6 ha including a pond, Brandenburg, Germany) representing an initial ecosystem undergoing highly dynamic ecological development starting from clearly defined starting conditions. Directly after completion of the construction, extensive monitoring equipment was installed to track the ecosystem development and to capture the
spatiotemporal variability of meteorological, hydrological, ecological, and soil conditions and vegetation succession. In this study, we focused on the water balance dynamics of the Chicken Creek catchment for the period 2005 to 2015 as influenced by ecological development. Water storage in the catchment was calculated from a three-dimensional model of groundwater volumes, soil moisture measurements, and water level recordings of the pond. The catchment water balance equation was resolved for evapotranspiration, the only part that was not measured directly. Time series of meteorological, hydrological, and ecological data for 10 yr enabled us to characterize the transient development of the catchment and to evaluate the effect of different feedback mechanisms on catchment hydrology.