Landscape Controls on Hydrological Processes under Rainfall- and Radiation-driven Conditions

Abstract:

1. Introduction

Hydrological modelling frameworks are often based on model units which supposedly show similar hydrological process response to main meteorological drivers. Processes considered in the models could be soil moisture response, change in groundwater table or evapotranspiration dynamics and model units might be separated according to site characteristics like land use, geology or slope position which are likely to control hydrological processes. However, the similarity in process response as a result of these controlling factors has rarely been demonstrated.

Hydrological processes exhibit temporal variability which depends to a large extent on the dynamics in their meteorological drivers. Two primary meteorological drivers are solar radiation and precipitation, affecting evapotranspiration and soil water dynamics. An assessment of the main landscape or site controls on hydrological processes could therefore be conducted separately for contrasting “rainfall-driven” and “radiation-driven” conditions. Monitoring variables which directly respond to these conditions can then serve as indicator variables for hydrological processes and the assessment of their controls. Examples for suitable monitoring variables are soil moisture content and sap flow: Soil moisture response to rainfall events is a good indicator for controls of water storage and transport processes under rainfall-driven conditions whereas sap flow, being mainly dependent on radiation input, can help identify controls of evapotranspiration under radiation-driven conditions.

The research question we want to examine in this study is to what extent site characteristics such as geology, land use, slope position and aspect control hydrological processes at the catchment scale, using the examples of soil moisture and sap flow dynamics, and how the importance of these controls varies between rainfall-driven and radiation-driven conditions.

2. Methods

Our study area is the Attert catchment, situated in the west of Luxemburg and partly in Belgium. It comprises an area of about 290 km² and three geological units, predominantly Devonian schists of the Ardennes massif in the northwest , Triassic sandy marls and a small area underlain by Luxemburg sandstone (Jurassic) on the southern catchment border. Mean annual precipitation for the study area is approximately 850 mm, and land use varies from mainly pasture and agriculture in the marls area, mainly forests in the sandstone to a mixture of agriculture and pasture on the plateaus and forests on the steep slopes of the schist area.

Within the CAOS research unit a monitoring network of 46 sensor clusters was installed, covering the different geological units, two land use types, different slope positions and aspects. These characteristics are thought to be the main landscape controls on hydrological processes at the sites. At each cluster site a comprehensive set of measurements is recorded to assess these processes. In this study we focus on two of the monitoring parameters: 1) Soil moisture is measured at each site in three profiles at the depths of 10, 30 and 50 cm. 2) Sap flow sensors of the heat pulse velocity type were installed in four trees at all forest sites. The trees were selected as representative examples of the local forest structure, based on stand characterisation surveys.

Data analysis for soil moisture, serving as target parameter for rainfall-driven conditions, focused on the dataset from the schist area because for these sites longer time series were available. Dominant controls on soil moisture dynamics were analysed in two different ways: 1) Temporal persistence of soil moisture patterns was assessed by arranging each sensor’s deviation from the spatial mean at all time steps according to their mean value. In these plots the influences of land use, slope position and aspect were evaluated by counting the number of sensors which showed larger or smaller mean differences than the averaged spatial mean of the catchment. 2) Soil moisture response to three individual rainfall events in May, July and October, respectively, was examined by counting responding sensors in comparison to the total amount of sensors, by looking at lag times and peak amplitudes and by determining thresholds in plots of soil moisture response related to cumulative rainfall. The events differed in intensities and overall moisture conditions: The May event had two peaks and amounted to 12 mm in total, but was preceded by a relatively wet month. The July event followed a dry month and had one large peak of 11 mm in total. The October event was again preceded by a wet month and was less spiky than the other two events, amounting to a total of 14 mm. For each of these events and for all three analysis approaches (relative sensor response, lag times/amplitudes, thresholds) we assessed the importance of land use, slope position and aspect.

Sap flow data was used to assess dominant controls on evapotranspiration, assuming sap flow represented a mainly radiation-driven process. We used linear models on daily aggregated sap velocities and a set of tree-specific (species, diameter at breast height, tree height), stand-specific (basal area) and site-specific (geology, slope position, aspect) predictors, determining their relative importance for the model for each day.

3. Results

Controls on soil moisture dynamics

The analyses of temporal persistence of soil moisture patterns revealed an influence of land use on soil moisture dynamics. Especially at the 30 cm soil depth, large contrasts between wetter pastures and drier forests were visible. While the effect of slope position was somewhat ambiguous, aspect had a marked effect on the temporal persistence of soil moisture patterns, north-facing slopes being consistently wetter than south-facing slopes.

Sensor response to rainfall events differed between forest and pasture, between upslope and downslope position, between north- and south-facing slopes and between the three events (Fig. 1A). Pastures showed a higher relative sensor response than forests for the shallower soil depths. Downslope sensors showed higher relative response than upslope sensors, predominantly in the upper soil depths. However, the relative sensor response patterns differed greatly between the three events, for example the more continuous and larger October event led to more sensors responding in larger soil depths, especially in forests and upslope positions.

Lag times between the between the onset of the rainfall event and the maximum rise of soil moisture response varied mainly due to event characteristics, for example the high-intensity event in the dry month of July resulted in very short lag times in all depths irrespective of land use, slope or aspect, while the event in May triggered a more differentiated response, with shorter lag times at the 10 cm depth and a delayed response in 30 cm as well as larger 10 cm response variability in pastures and downslope sensors. The amplitude of the response exhibited larger values for 10 cm pasture sensors for the May and July event compared to forests, and marginally larger 10 cm downslope amplitudes in July. However, for both lag times and amplitudes there was large variability including many outliers in the datasets.

The analyses of thresholds in soil moisture response relative to cumulative precipitation input revealed consistently lower average thresholds for pastures compared to forest sensors, which was especially pronounced for the 10 cm depth of the July event. Downslope sensors were also responding already at slightly lower rainfall input compared to upslope sensors, while aspect did not have a clear effect on thresholds. A summary of the dominant controls on soil moisture is shown in Table 1.

Controls on daily sap flow velocities

Sap velocities were strongly dependent on the individual trees’ characteristics, especially species and DBH, however, the relative importance of these predictors changed over time (Fig. 1B). In June-August, for example, basal area had a large effect on sap velocities while the tree-specific parameters lost importance. Site-specific parameters were generally less dominant, but aspect sometimes explained more than 10% of the variance, for example in September. The contributions of the different controls varied over time, as did the overall explained variance which was highest during the late summer when precipitation was absent or low, so in truly radiation-driven conditions. A summary of the dominant controls on sap flow is shown in Table 1.

Table 1: Dominant controls on soil moisture and sap flow as identified in the different analyses, XX: strong influence, X: influence, (X): little influence, -: no influence

Dataset/Analysis	Sensor depth	Land use	Slope pos.	Aspect	Geology	Tree par.	Stand par.
Soil moisture dynamics
Temporal persistence	X	X	(X)	X
Event sensor response	X	X	X	X
Event lag/amplitude	X	X	(X)	(X)
Event thresholds	(X)	X	X	-
Daily sap flow velocities
Relative importance of predictors			(X)	X	(X)	XX	X

4. Discussion

Controls on soil moisture dynamics

Land-use influence on soil moisture response was visible in all of our soil moisture analyses. Reasons for this strong signal are probably differences in rooting depths and management or treading effects on pastures, leading to differing pore structures and pore distributions for the forest and pasture sites, which in turn influence infiltration and percolation behaviour. Additionally, rainfall input at the forest sites is redistributed by the canopy and the resulting throughfall is more variable than rainfall on the pastures and is likely to exhibit local hotspots of infiltration at dripping points or due to the channelling effect of stem flow.

Slope position mainly influenced sensor response and response thresholds. This could indicate differences in the proximity to groundwater close to the streams and also to differences in soil textures such as more loamy layers in the subsurface of downslope sites. Additionally the connection of downslope plots to their respective upslope areas might lead to differences in soil moisture dynamics.

Aspect affected mostly the temporal persistence of soil moisture patterns and sensor response. The differences in soil moisture dynamics on north- and south-facing slopes are probably mainly attributable to difference in evapotranspiration, however, for forest also differences in stand structure and management could play a role.

However, comparison between the different events showed that event characteristics and initial soil moisture state greatly influenced the magnitude and patterns of response. Therefore, results for single rainfall events should always be viewed within the context of antecedent precipitation and catchment state

Controls on daily sap flow velocities

Sap flow and thus evapotranspiration was mainly controlled by tree and stand parameters, however, also the site characteristics explained considerable parts of the overall variance. Geology determines the drainage patterns in the catchment and thus the trees’ access to deeper water resources. Slope position can also influence the access to water, however there are only very few downslope forest sites so this influence might be underrepresented in the dataset. Lastly, the differences in radiation input for different aspects determine potential evapotranspiration. Especially in the truly radiation-driven period July-August with almost no rainfall, the models were more reliable for explaining the variation in the data.

5. Conclusions and outlook

Site characteristics such as land use, slope position, aspect and geology exerted considerable control over the examined soil moisture and sap flow dynamics, however, for sap flow the tree- and stand-specific controls were clearly more pronounced. Nevertheless, the approach of using site characteristics to differentiate areas of similar hydrological functioning seems viable, bearing in mind that response to individual rainfall events can differ greatly and should be viewed in the context of antecedent precipitation and catchment state. If soil moisture dynamics is viewed as an indicator variable for more rainfall-driven conditions and sap flow for more radiation-driven conditions, differences can be seen in the importance of slope position for the former and the importance of aspect for the latter. Further analyses will build on these first results, especially extending the analysis of radiation-driven conditions to soil moisture by focusing on the summer recession period or short drainage periods without rain. Additionally, the influence of rainfall input on sap flow will be examined, for example by estimating the decline in sap velocities during rainfall events and assessing the controls on the resulting pattern.