Självständigt arbete Nr 91
Where does the stream begin?
Stream initiation under variable wetness conditions in a boreal landscape Where does the stream begin?
Stream initiation under variable
wetness conditions in a boreal landscape
Evelina Gallon and Sanna Lindberg
Evelina Gallon and Sanna Lindberg
Uppsala universitet, Institutionen för geovetenskaper Kandidatexamen i Geovetenskap, 180 hp
Självständigt arbete i geovetenskap, 15 hp Tryckt hos Institutionen för geovetenskaper Geotryckeriet, Uppsala universitet, Uppsala, 2014.
The understanding of where the streams begin is an important factor in both hydrology and geomorphology, as well as for land use activities. Despite this, only a few research projects have been done in a snowmelt-dominated boreal landscape. The main objective in this study is to see if one could predict where a stream initiates by knowing possible controlling factors. Data points from stream initiation points in the boreal landscape of Krycklan, situated 50 km Northwest of Umeå, were analyzed. Krycklan is a well-known research area and a lot of research projects have been done here. The landscape is diverse and most of the streams have been modified by human impact. The data points were collected with help of a Global Positioning System (GPS) during three different sampling campaigns; May 2012, May 2013 and August 2013. Maps were made from a Digital Elevation Model (DEM) showing slope, elevation and contributing areas for the stream heads. No clear relationship between the contributing area and slope could be found in this area, although it has been demonstrated in more semi-arid climates. The results show that other factors than the contributing area, elevation and slope seem to have a greater impact for the initiation of streams in a boreal landscape. The results were expected because of the modifications done for the streams.
Självständigt arbete Nr 91
Where does the stream begin?
Stream initiation under variable wetness conditions in a boreal landscape
Evelina Gallon and Sanna Lindberg
Sammanfattning
Att förstå vart källor börjar är viktigt både för hydrologiska och geomorfologiska processer, såväl som för markanvändning. Trots detta är väldigt lite forskning gjord inom detta område i ett snösmältningsdominerat nordligt klimat. Syftet med denna studie har vart att försöka förstå vilka faktorer som kan påverka källornas startpunkt.
Data har analyserats från källornas startpunkter från det nordliga avrinningsområdet, Krycklan, belägen 50 km nordväst om Umeå. Krycklan är ett känt forskningsområde där många olika forskningsprojekt har gjorts. Landskapet är varierande och de flesta bäckarna har blivit modifierade av människor. Punkterna togs med hjälp av en Global Positioning System (GPS) under tre perioder; maj 2012, maj 2013 och augusti 2013.
Utifrån en Digital Höjdmodell (DHM) gjordes rasterkartor som visade lutning, höjd och avrinningsområden för enskilda källor. Ingen tydlig relation mellan
avrinningsområdens area och lutning kunde hittas i detta område, men har påvisats i tidigare studier i torrare klimat. En viss korrelation kunde ses mellan storleken av avrinningsområdet och markens fuktighet. Resultaten påvisar att andra faktorer än avrinningsområdets area, höjd och lutning verkar ha en större påverkan för källornas början i detta nordliga landskap. Eftersom bäckarna är så modifierade i detta område så var resultatet förväntat.
Abstract
The understanding of where the streams begin is an important factor in both
hydrology and geomorphology, as well as for land use activities. Despite this, only a few research projects have been done in a snowmelt-dominated boreal landscape.
The main objective in this study is to see if one could predict where a stream initiates by knowing possible controlling factors. Data points from stream initiation points in the boreal landscape of Krycklan, situated 50 km Northwest of Umeå, were analyzed.
Krycklan is a well-known research area and a lot of research projects have been done here. The landscape is diverse and most of the streams have been modified by human impact. The data points were collected with help of a Global Positioning System (GPS) during three different sampling campaigns; May 2012, May 2013 and August 2013. Maps were made from a Digital Elevation Model (DEM) showing slope, elevation and contributing areas for the stream heads. No clear relationship between the contributing area and slope could be found in this area, although it has been demonstrated in more semi-arid climates. The results show that other factors than the contributing area, elevation and slope seem to have a greater impact for the initiation of streams in a boreal landscape. The results were expected because of the modifications done for the streams.
Table of Content
1. Introduction ... 1
1.1 Definitions and Basic Concepts ... 1
1.2 Literature Review ... 2
1.3 Objectives ... 3
2. Methods ... 3
2.1 Study area ... 3
2.2 Data ... 6
2.2.1 Sampling procedures ... 6
2.2.2 Processing and Statistical Analysis... 7
3. Results ... 9
3.1 Processing and Statistical Analysis ... 9
3.1.1 Diagrams for contributing area and elevation ... 9
3.1.2 Boxplots ... 11
4. Discussion ... 13
5. Conclusion ... 15
6. Acknowledgements ... 15
References ... 16
Appendix ... 18
1
1. Introduction
1.1 Definitions and Basic Concepts
The appearance of the stream network in a boreal landscape differs from other regions. Boreal streams drain the landscape, which often consist of wetlands, forest and lakes (Laudon et. al., 2007). Because of the diversity of a boreal landscape the hydrology and stream network is complex (McEachern et. al., 2006). The size of the streams can also vary during the seasons because of the change in flow during spring and autumn (Buffam et. al., 2007). The stream network will increase during higher flows and shrink when the flow becomes less, and therefore the location of the stream initiation, also called the stream head, can differ. Exactly how the stream network change during the course of a year is still not completely understood (Nihm, 2012).
Streams consist of a moving body of water, which flows in a channel (also called the stream bed). The channel can either have continuous flow all year round, or have periods when there is no water (Langbein & Iseri, 1960). Therefore streams can have intermittent, ephemeral or a perennial flow. A perennial stream means that the stream has a yearly flow, whereas an intermittent stream can dry out during drier parts of the year. An ephemeral flow occurs only during higher
precipitation events (Langbein & Iseri, 1960).
Channel head has been defined as the location furthest upslope of a water flow with definable banks (Montgomery & Dietrich, 1988, 1989; Henkle et. al., 2011). The understanding of where and why the channel initiates at a certain point is an important part in knowing the hydrological, geomorphological and geochemical processes (Henkle et. al., 2011; Jason et. al., 2012). The channel head represents the start of a stream, it also transport sediment and supplies organisms with
nutrition’s (Mazurek, 2011; Henkle et. al., 2011). It also defines the transition zone were the slope will turn into a channel. The stream head does not always have the same position as the channel head. The stream head has a perennial flow and its position can vary during the course of the year, whereas the channel head has intermittent and ephemeral flow which is then present above the stream head (Henkle et. al., 2011).
The point on where the stream will initiate depends on many factors, such as spreading due to hillslope processes and high discharge occurrences.
Hillslope processes can fill channel heads and cause the channel head to move downwards, but during greater discharge events the channel head can propagate upwards again (Montgomery & Dietrich, 1989). In a geochemical perspective, the channel head can serve as a transport access for pollutants to enter the stream network (Jason et. al., 2012). The channel head is also important for understanding how landscape activities influence the natural flows and headwater channels. Land use activities such as road constructions and logging can give higher peak flows than natural and can also increase the erosion rate (Jaeger et. al., 2007; Jones et al., 2000; Wemple et. al., 1996). The initiation points of a stream network can often be wrongly represented in maps due to the fact that the climate, landscape and land use influence the position of the stream heads (Jaeger et. al., 2007). Land use affects the hydrology and can disturb subsurface as well as overland flow which in turn can cause changes in the stream network (La Marche & Lettenmaier, 2001).
The processes contributing to the initiation of channels differs in different types of environments. In steeper terrains landsliding and seepage erosion are the
2 dominant factors that control channel initiation (Montgomery & Dietrich, 1989).
Seepage erosion takes place when there is a mass removal caused by seepage (Pornprommin & Izumi, 2010). In a more gentle sloped terrain the main process is overland flow (Montgomery & Dietrich, 1989). There are three different types of overland flow; (1) infiltration excess overland flow, (2) saturated overland flow and (3) interflow. In boreal landscape the saturated overland flow is the most common.
Saturated overland flow occurs when the soil has become saturated and the
rainwater cannot infiltrate, which results in water flowing on the surface and creates ephemeral channels (Beven, 2004; Mosley & McKerchar, 1993). This is why overland flow is not a dominant factor in steeper terrains, where the ground seldom gets
oversaturated.
According to how the channel heads have formed, they can be either gradual or abrupt. A gradual channel head has not a well-defined initiation point, whereas an abrupt channel head has. Abrupt channel heads can be formed in distinctive headcuts (Montgomery & Dietrich, 1989).
Another controlling factor for a stream to initiate is that the ground water table is close to the surface, and that the ground is not too dry. A common method to predict stream initiation points is to look at different wetness conditions and their spatial distribution by using the Topographic Wetness Index (TWI), TWI builds on the assumption that the water is flowing in the direction of the local slope, where the assumption is made that the groundwater table mainly follows the topography. TWI works better in steeper terrains than in flat terrains, because the slope is too low in flat areas and the flow directions of the water gets hard to predict (Grabs et al., 2009).
A higher TWI value predicts a wetter area, and a lower value a drier area. Assuming that all other conditions are the same (geology etc.), two areas with the same TWI values are thought to have a comparable hydrological response to precipitation (Cheng-Zhi Qin et. al., 2011).
1.2 Literature Review
Although the literature does not provide many articles on the initiation of channels, a few studies have tried to find a relationship that would predict the start of a channel (Henkle et. al., 2011; Jaeger et. al., 2007; Montgomery & Dietrich, 1988, 1989).
In a study made by Montgomery & Dietrich (1988), data was collected in the semi-arid to humid landscape of Oregon and California, and the results supported an inverse source area–slope relationship. The result showed that channel initiations in a landscape with a steep local valley gradient are caused by landsliding. The source area-slope relationship did not support channel initiation caused by overland flow.
Continuing studies by Montgomery & Dietrich (1989) confirmed previous work from 1988 that mathematically proved the source area-slope relationship by using statistical correlation, for a wide range of slopes.
Henkle et. al. (2011) did further work on finding parameters that could control channel initiation. The results showed that the relationship for drainage area and slope, as found by Montgomery & Dietrich (1989), were poor for a semi-arid climate with a higher elevation. It also showed that subsurface and surface processes may interact and affect the initiation point of the channel.
Jaeger et. al. (2007) investigated a forested area in the Northern part of USA to see if one could predict channel initiation in a specific lithology (sandstone
3 and basalt). The source area was defined with both a Digital Elevation Model (DEM) and with a GPS in the field. The results displayed no secure relationship when plotting local slope against source area. The reason for this was because the DEM derived source-area did not match the real source-area derived from the GPS. The source-area also varied with lithology.
Studies were also made on how well a DEM predicts and depicts
landscape features and stream beginnings (McMaster, 2002; Zhang & Montgomery, 1994). McMaster (2002) did a study on how the resolution on a DEM affects the accuracy of the results. The results showed that if the resolution were smaller than the length of a hillslope, the values gotten through the DEM would be incorrect. The results also demonstrated that if the resolution is greater than 180 m, the prediction of streams initiation points decline.
1.3 Objectives
Only very few research projects were done in a snowmelt-dominated area, which motivated us to focus on stream initiation in a boreal landscape in the northern part of Sweden. The main objective is to see if one could predict where a stream initiates by knowing possible controlling factors such as slope, elevation, TWI and contributing area, and compare them with each other to see if there is any correlation between them. The second objective is to see how stream initiation varies during dry and wet conditions over the year. One hypothesis is that the manmade ditches can affect the results and give less of a correlation as found before by others that only have had natural channel heads.
2. Methods 2.1 Study area
Our study was performed on the Krycklan catchment, which is situated in the
northern part of Sweden, 50 km northwest of Umeå (Figure 1). Krycklan is a research area and has been since 1980. As from 2002 the research area, called the Krycklan Catchment Study (KCS), expanded to today’s size with an area of 6790 ha (67.9 km2). Today it can be subdivided into 18 smaller subcatchments, and the data points collected for this study were mainly collected in subcatchments 1,7 and 9 (Table 1, Figure 2). The research made in this area includes research on forests, mires, streams, groundwater, lakes and soils. The aim of KCS is to give a prime field research infrastructure in a boreal landscape (Laudon et. al., 2013). It is one of few research areas in the world that is dominated by snowmelt.
4 Figure 1. Location of the Krycklan Catchment Area (left) and a digital Elevation Model map
over the Krycklan catchment (right).
The topography in Krycklan is higher in the north and levels off towards the south, with elevation ranging from 114 to 405 meters above sea level (Laudon et.
al., 2013). The mean annual precipitation in the area from 1969-1990 is 600 mm/year (SMHI, 2013). Much of the precipitation comes as snow during the winter and
contributes to the stream discharge when the snow starts to melt in the spring. Mean temperature over 30 years is 1.8°C (Laudon et. al., 2013).
The main geology is gneissic rocks and quartz rich sedimentary rocks from the “svekokareliska” orogenesis that took place 2850-1870 Ma ago, although there are almost no visible outcrops (SGU, 2014). More than half of the Quaternary deposits are till (51%), sorted sediments (30%) and the rest is mostly peat (10%) and some thin soils. The sorted sediments can mostly be found in the lower parts of the catchment, whereas till, peat and visible outcrops can be found at higher elevations (Figure 2).
5 Figure 2. Geology map of Krycklan (top) and subcatchments 1, 7 and 9 (bottom left to
bottom right) with the measured data points.
The landscape is dominated by forest (87%) and streams. There are also some mires in the area (9%) as well as some smaller lakes. Visible outcrops only represent 1% (Laudon et. al., 2013). Today there are no forestry operations in the central parts of the catchment, but it has been practiced before. This is the reason to why there are so many ditches in the area. To increase the productivity of the forestry, natural streams got dug deeper and new ditches were dug as well. The ditches were dug mostly in till soil, because of the natural drainage capacity of the sorted sediments. The stream network connects through the whole of Krycklan but unmodified natural streams are rare.
Table 1. Catchment characteristics of subcatchments 1, 7 and 9. (Modified from Laudon et.
al. (2013))
Catchment Area (ha) Lakes (%) Forest (%) Mire (%) Till (%) Thin Soils (%)
Rock outcrops (%)
C1 48 0,0 98,0 2,0 92,1 7,9 0,0
C7 47 0,0 82,0 18,0 65,2 15,4 0,0
C9 288 1,5 84,4 14,1 69,1 6,8 1,7
6
2.2 Data
2.2.1 Sampling procedures
A dominant factor when choosing channel heads in previous made studies has been that they have been far away from roads and in areas that have been undisturbed for a long time, to have such a natural and undamaged channel head as possible
(Henkle et al., 2011; Jaeger et al., 2007; Montgomery & Dietrich, 1989). The data sets collected for this research have been fairly close to roads, but always upstream of the roads which mean that they should not have been disturbed by the road systems downstream. Many of the selected channel heads are manmade ditches, which gives an opportunity to study the influence of landscape alterations.
The channel head locations in the study area were collected with help of Global Positioning System (GPS). Data were collected during three different
sampling campaigns, two during the spring flood (May 2012 and May 2013) and one during the drier part of the year in August 2013 (Table 2). Sixteen data points were collected in catchment 1. Catchment 9 is a bigger catchment and includes catchment 7, so of the total forty-two data points collected in catchment 9, twenty-eight of them belong to catchment 7.
Table 2. The Different types of stream heads in the mapped area.
The most of the stream heads mapped were manmade ditches, some were natural streams and a few were road ditches (Table 2). Because of the
accessibility, the most of the measured stream heads were quite close to roads in the area, but the stream heads were always picked so they would be upstream of the road to get as little disturbance as possible (Figure 3). The mapping positions were also chosen according to where other research has been done (e.g. discharge stations) and to get such a good variation as possible to reflect the true situation in Krycklan. The most of the channel heads geometry (channel width, water table depth, water width, channel depth) was measured. In May 2012, some of the ditches were still covered with snow, others had a continuous flow, discontinuous flow or no visible flow. For the points collected in May 2013 some of the channel heads were also covered with snow, and some were damaged by tree cutting processes. Overall the flow in the streams was less for May 2013 than in May 2012. In August 2013 the most of the channel heads were wet, but only a few were connected to the main stream and had a visible flow.
The points taken in May and August 2013 were supposed to match the streams measured in May 2012, but due to some errors with the coordinates this was not always possible, and some new places were added as well. The channel heads were mainly sampled in till, but some were also mapped in sediment soil.
The discharge was always higher in May than in August. When looking at the discharge series for the outlet of catchment 7 (Figure 4), one could see that the discharge peak was much higher for May 2012 than for May 2013 which gives a perception on how wet it is in the catchments during the sampling periods.
Manmade Ditch Road Ditch Natural Stream Total Data Points
May 2012 37 2 10 49
May 2013 16 0 6 22
August 2013 20 0 3 23
7 Figure 3. Roads and the collected data points in the Krycklan catchment.
Figure 4. The discharge series from catchment 7.
2.2.2 Processing and Statistical Analysis
All mapped channel heads were processed in Esri´s ArcGIS with the help of a DEM map showing the elevation in the area. When taking measuring points with a GPS the accuracy is not exact which needs to be taken into account. Also the DEM has a resolution of 5x5 meters, so the value of the data point in the model may not be entirely correct. Therefore a 10-meter buffer zone was created in ArcGIS for each data point to help get more reliable values.
8 The slope was not measured in the field instead it was determined in ArcGIS by using a raster of the slope of Krycklan’s catchment area. Raster for
contributing area and elevation derived from the DEM were also used. The data sets were analyzed in Microsoft Excel and scatter plots were made and linear regressions were performed.
The two parameters, contributing area and maximum elevation were examined. The size of the contributing area needed to initiate flow at wetter
conditions can be assumed to be smaller than the size needed when the conditions are drier. But for a specific point/channel head the contributing area is supposed to be constant. The contributing area is the area upslope for a specific stream head.
The size of the contributing area itself does not change, but the location of the stream head can vary during drier and wetter conditions, when the stream head propagates upwards during wetter conditions its contributing area will be smaller, and the
opposite for drier conditions (Figure 5). The slope may influence the size needed for a contributing area to initiate flow. A high elevation with a steep terrain should be drier than a low elevation with a flat slope, because the steep terrain helps to drain the ground.
Diagrams were plotted showing maximum elevation against mean slope as well as the logarithm of the maximum contributing area against mean slope. A linear regression line was drawn for every dataset to see if there were any visible relationships (Table 3). The fit of the regression lines was examined with help of the R-square measure. The p-value was used to analyze the significance of the trends of the regression lines. Boxplots displaying the logarithm of the contributing area, slope and TWI were created for every data set in MATLAB. Because the most of the data points were collected in May 2012, individual boxplots were made for only this
sampling period as well. The TWI can be calculated by knowing the contributing area for a certain point per unit contour length, divided by the local slope (Equation1). The contour length in this case is the length of the grid-size in DEM, which is 5 meters.
Figure 5. Simplified sketch of the contributing area for a channel head in a bigger stream network.
slope local
length contour unit
area ng contributi ln
TWI (1)
9
3. Results
3.1 Processing and Statistical Analysis
3.1.1 Diagrams for contributing area and elevation
For the diagrams, points are divided into natural streams and manmade ditches to see if there are any changes in the correlation pattern, and to see if the correlation differs for the natural streams. The regression lines fitted to the scatter points of the logarithm of the contributing area (m2) versus slope (m/m) (Figure 6) shows a wide range in R-square values ranging between 0.003 and 0.71 (Table 3). The highest R- square value (0.71) is found for the natural streams from May 2012 data set. The manmade ditches from the same data set have a much lower R-square value of 0.016. When looking at the regression lines, one can see that the most of them have a negative trend (Figure 6a), which indicates that less slope is needed for bigger contributing areas for a stream to initiate. However, for the natural streams in May 2013 (Figure 6c) the regression line shows a weak positive trend instead, which would actually suggest the opposite.
Figure 6. Logarithm of the contributing area plotted against the slope for all data points (a), manmade ditches (b) and natural streams (c).
a b
c
10 Table 3. Statistics for Log (contributing area) plotted against slope (m/m). Numbers in bold represent a statistically significant slope of the trendline (i.e. p-value below significance level of 0.05)
May 2012 May 2013 August 2013
All Ditch Natural Stream All Ditch Natural Stream All Ditch Natural Stream
R square 0.09 0.02 0.71 0.05 0.21 0.00 0.01 0.00 0.06
P-value 0.039 0.440 0.002 0.319 0.086 0.913 0.618 0.920 0.841
Lower 95% -9.901 -7.823 -20.652 -17.780 -39.003 -27.582 -14.138 -12.370 -492.519
Upper 95% -0.256 3.474 -6.551 6.081 2.913 30.008 8.596 11.228 473.133
The regression lines for the diagrams displaying the elevation versus slope (Figure 7) have R-square values ranging from 0.007 to 0.24 (Table 4). This indicates that the relationship between these two parameters is poor if any. The first impression is that these diagrams are more scattered than the diagrams for the logarithm of the contributing area (Figure 6). After dividing the data sets in natural streams and manmade ditches the relationship remains poor. The highest Rsquare value is found for the data points for the natural streams in August 2013 with a value of 0.96. This value is not trustworthy as only three natural streams were mapped, which are too few to get a reliable result. There is a negative trend for contributing area against slope for the dataset from August 2013. For all three data sets
displaying maximum elevation against slope at least one of the linear regressions is showing a positive relationship.
The diagram with all measured values, the points for May 2013 shows a weak positive relationship (Figure 6a), the same is true for the diagram showing ditches (Figure 6b), which indicates that more slope is needed at higher elevations for a stream to initiate. For the natural streams, however, the dataset from August 2013 (Figure 6c) shows the positive trend seen for May 2013 instead. Although, two of three datasets shows a negative trend for every diagram displaying the elevation against the slope, indicating that less slope is needed at higher elevations for a stream to initiate instead.
Additional statistics, such as p-value and confidence interval of 95 %, are also calculated. The p-values (describing the probability for the hypothesis to be correct) are used as an indicator to see whether the trend of the regression lines is statistically significant. The p-values are relatively high for all the data sets (Tables 3 and 4), which suggests that most trends are not significant and that there is no clear relationship. The only three data sets with significant trends are, (1) the natural streams in May 2012 displaying logarithm of the contributing area versus slope (Figure 6c), (2) the ditches in August 2013 displaying the elevation versus slope (Figure 7b) and (3) the natural streams in August 2013 displaying the elevation versus slope (Figure 7c).
11 Figure 7. Maximum elevation plotted against the slope for all data points (a), manmade ditches (b) and natural streams (c)
Table 4. Statistics for elevation plotted against slope (m/m). Numbers in bold represent a statistically significant slope of the trendline (i.e. p-value below significance level of 0.05)
May 2012 May 2013 August 2013
All Ditch Natural
Stream All Ditch Natural Stream All Ditch Natural
Stream
R square 0.01 0.01 0.01 0.00 0.01 0.21 0.24 0.43 0.96
P-value 0.627 0.612 0.773 0.872 0.801 0.363 0.014 0.002 0.023
Lower 95% -301.93 -311.30 -574.50 -219.89 -667.97 -799.99 -678.80 -903.46 108.68
Upper 95% 183.93 185.90 443.10 258.26 848.21 368.28 -85.09 -245.24 257.03
3.1.2 Boxplots
The data set from May 2012 contained the most sampling points and is therefore further analyzed. The boxplot displaying the logarithm of the contributing area for May 2012 (Figure 8a) shows that the ditches have a slightly narrower range of values than for the natural streams. The mean and the median of the logarithm of the
contributing area are higher for natural streams than for manmade ditches.
Considering the slope (Figure 8b) the pattern seems to be similar, with wider spread values as well as slightly higher mean and median for the natural streams. When looking at the TWI, the boxplots for manmade ditches and natural streams seems to have almost the same distribution, as well as mean and median (Figure 8c).
Boxplots for all three data sets are also examined (Appendix1), to compare the wetter seasons values against drier and to see how the median and mean changes during the different sampling periods. For all the boxplots the datasets
a b
c
12 are divided into manmade ditches and natural streams, to see if there are any
differences in correlation. For the slope, the boxplots displays a similar trend for all three data sets (Appendix1.1). The values for the natural stream vary more than for the manmade ditches, although the ditches have more outliers. Boxplots for the TWI are quite similar for both ditches and natural streams. Median and mean are almost the same as well (Appendix 1.2).
In the boxplots showing the logarithm of the contributing area, the median and mean differs for all three data sets, the values also differs between natural streams and manmade ditches (Appendix 1.3). The drier season in August 2013 shows a greater contributing area than for the wetter season in May 2012, 2013.
The median is greater for natural streams than ditches in May 2012, 2013 but for August it is the opposite.
That the contributing area expands in August can also be supported with the help of plotting the mean contributing area for each data set against the mean discharge during the sampling period (Figure 9). The diagram shows that for the wetter seasons in May 2012 and 2013, the mean contributing area is smaller than for the drier season in August 2013.
Figure 8. The boxplot displays the logarithm of the catchment area (a), the slope (b) and TWI (c) for manmade ditches and natural streams in May 2012.
a b
c
13 Figure 9. The mean contributing area plotted against the mean discharge for natural streams
and manmade ditches, for the three sampling campaigns.
4. Discussion
Although previous studies have shown a negative relationship for area plotted against slope (Henkle et. al. 2011; Montgomery & Dietrich, 1988, 1989), no clear correlation could be found for this area. One possible reason for this is that many of the measured points are manmade ditches and not natural streams. Another reason is that only 21-23 data points were sampled in May 2013 and August 2013, this is quite few to get a reliable result. As mentioned earlier, few studies have been done in a snowmelt dominated area. This can have an impact on how streams begin, and therefore the parameters looked at may not have the same strong influence as in previous researched areas.
For all the diagrams displaying contributing area, the R-square value of the regression line is very small, which means that the regression lines are not good representations of the data points in these cases. This, in turn, makes is difficult to draw conclusions about the relationship between the parameters. The high R-square value seen for the natural streams for the data points collected in August 2013
cannot be taken into account, because only three values were measured which is too few to give a reliable result.
The low correlation seen for the most of the data sets was expected for the manmade ditches, but not for the natural streams. The reason to why the
correlation is so poor for natural streams as well can be that there were too few measuring points. In the area of Krycklan, natural streams are rare, and some of them can have been modified in some ways as well. Only the bigger streams in Krycklan are largely unmodified. When digging ditches in a watershed one disturbs the natural drainage patterns, and this could be a reason to why the relationship is almost nonexistent.
When looking at the diagrams for the elevation plotted against the slope, the regression lines for the three data sets vary a lot. The trends vary from negative to positive to almost no trend. Thus, there seems to be no clear relationship between these parameters. Although the negative trend seems to be the dominant, but more points would have to be taken to get a more reliable value of the R-square value.
The p-value, explaining how good the hypothesis is, had very high values for almost all the data series. Only five (three of them found in August 2013 for the elevation versus slope) p-values were lower than 5%, which means that in the
14 confidence interval of 95%, the slope of the line will never be zero. The other p-
values were much higher than 5%, which means that the confidence interval will range from negative to positive and also that the hypothesis must be rejected. There is therefore a great chance that the slope of the line at some time will take the value of zero, which in turn means that the results from the diagrams are not reliable and that there is no proof that there actually could be a relationship between the
examined parameters. The p-value also varies in size for manmade ditches and natural streams and no clear relation could be found between them either.
The p-value for August 2013, displaying the elevation against slope, is lower than 5% for all diagrams. This could indicate that there is some relationship between the elevation and slope. Which would mean that in higher elevations, less of a slope is needed for a stream to initiate, and the opposite for lower elevations. But it is not the elevation alone that is the reason for this. In higher elevations the
conditions changes, the soil becomes thinner and can have a different composition than in the valleys. Krycklan has been affected by the latest glaciation and this has influenced the sediments that have been deposited here. Because of this, thinner beds of till are found in higher elevations, and in the valleys one can find thicker deposits of more sorted sediments instead. The till is shallower than the sorted sediments, because of this till has more shallow flow paths than the deeper sorted sediments. This in turn results in more frequent convergence of flow in the till, that can appear as a stream at the surface.
The boxplots for May 2012 displaying slope indicate that the slope values are more similar for the ditches than for the natural streams. This is probably because when people drained the area from water, they did it on lower, flat
elevations. The initiation of natural streams, on the other hand, depends on several parameters, so only looking at the slope does not tell us that much. The TWI does not change that much between ditches and natural streams. The TWI is a measure on how wet the ground is, and because the ditches and the natural streams are measured during the same period they should have similar values. For the logarithm of the contributing area the natural streams seems to have a little bit more variation in size than the ditches. This can be for the same reason as the slope. If the ditches are dug in areas with similar slopes and soil, the catchment area for this point will also be similar.
The boxplots for all the three data sets displaying slope plotted against data sets for natural streams shows a wider range of values than for the ditches. This is probably because of the same reason as mentioned for the data set for May 2012.
One did not dug ditches in steeper terrains, because they drain themselves and therefore the slope values for the manmade ditches are more alike than for the natural streams.
The boxplots showing the contributing area displays a more similar pattern between natural streams and manmade ditches. The contributing area for August is somewhat bigger than for May 2012 and 2013, which should be the case for the drier periods, because the contributing area for each stream head expands.
Also the diagram for the contributing areas against the mean discharge shows the same. The mean discharge in May 2013 was much higher than the mean discharge in May 2012, which explains why the mean contributing area is bigger for May 2012.
In August 2013, when the mean discharge is at its lowest, the contributing area is the biggest. When the discharge is at its lowest, the ground is at its driest. This means that the groundwater table is lower, and therefore subsurface flow will propagate further downslope. During higher discharge events, the groundwater table is higher
15 and the subsurface flow can propagate upslope instead. When the stream head propagates upslope, its contributing area will be smaller and the opposite way around during drier conditions, which is what our results displays.
The TWI values for the three data sets are similar for both ditches and natural streams. The mean is somewhat bigger for May 2013 than May 2012. This should probably be the other way around, because it is wetter in May 2012 than in May 2013, so we cannot explain the higher values for May 2013.
For future work one should try to collect more data points, and try to find an area that has not been so modified by human activity. All the ditches in this area can effects the results, so collecting more natural stream heads would be of interest.
However, great parts of the Swedish landscape have been modified by human activity, so understanding how this affects the landscape is of interest.
Since this is one of few studies made on stream initiation in a boreal landscape and only four parameters were looked at in this study, one should try looking at other indices such as wetness index, flow path and convergence to see if there is some relationship. One could also measure the distance between the points taken in May and the points taken in August, to see how the stream head varies during wet conditions compared to drier conditions. In this study only the depth of the manmade ditches was measured, not the depths for the natural streams. This is a thing that could be good to do in future work. Knowing the depth and which ditches that have water and which do not can be important in understanding where the flow begins, i.e. where the groundwater intersects the surface. The water table in a ditch can be lower than the groundwater table, but a deep groundwater table would possibly not initiate surface flow it the ditch was not dug.
It could also be good to analyze the depth of the sediment and till and how the depth varies with elevation. This can give a hint on the capacity of the soil to store water, or how fast the water is released.
5. Conclusion
Elevation, catchment area and slope seem to be three factors that may impact the stream initiation, but they do seem to be of less importance in a boreal landscape than in semi-arid to arid environments. Due to many uncertainties in the field measurements, it was difficult to see a clear relationship in the study, especially when comparing drier and wetter season.
We could, however, show that the contributing area is larger for drier conditions compared to wetter conditions. Furthermore, our study points towards the fact that the topography seems to be more important for the initiation point of natural streams than for manmade ditches.
6. Acknowledgements
We would like to give a special thanks to our helpful supervisors, Claudia
Teutschbein and Reinert Huseby Karlsen. They have helped us understand all the new terms and guide us in the right direction. We would also like to thank Thomas Grabs that has taken the time to read through our work and has also given help and information.
16
References
Beven, K. (2004). Robert E. Horton’s perceptual model of infiltration processes.
Hydrological Processes. p. 3447-3460. John Wiley and Sons, Ltd.
Buffam, I., Laudon, H., Temnerud, J., Mörth, C.-M., Bishop, K., 2007. Landscape- scale variability of acidity and dissolved organic carbon during spring flood in a boreal stream network. Journal of Geophysical Research:
Biogeosciences 112, G01022. doi:10.1029/2006JG000218
Grabs, T., Seibert, J., Bishop, K., Laudon. H. (2009). Modeling spatial patterns of saturated areas: A comparison of the topographic wetness index and a dynamic distributed model. Journal of Hydrology 373, p. 15-23.
doi:10.1016/j.jhydrol.2009.03.031
Henkle, J. E., Wohl, E., Beckman, N. (2011). Locations of channel heads in the semiarid Colorado Front Range, USA. Geomorphology 129, p. 309- 319. USA, Colorado State University.
Jaeger, K. L., Montgomery D. R., Bolton, S. M. (2007). Channel and Perennial Flow Initation in Headwater Streams: Managment Implications of Variability in Source-Area Size. Environ Manage, vol. 40. p. 775-786.
Jones, J.A., Swanson, F. J., Wemple, B. C., Snyder, K. U. (2000). Effects of Roads on Hydrology, Geomorphology, and Disturbance Patches in Stream Networks. Conservation Biology, vol. 14. No. 1. p. 76-85. ISI Journal Citation Reports, USA.
Julian, J. P., Elmore, A. J., Guinn, S. M. (2012). Channel head locations in forested watersheds across the mid-Atlantic United States: A physiographic analysis. Geomorphology 177-178, p. 194-203. Elsevier B.V. USA.
La Marche, J. L., Lettenmaier D. P. (2001). Effects of forest roads on flood flows in the Deschutes River, Washington. Earth Processes and Landforms. p.
115-134. John Wiley and Sons, Ltd.
Langbein, W.B., Iseri, K. T. (1960). General Introduction and Hydrological DefinitionsI.
Manual of Hydrology: Part 1. General Surface-Water Techniques. United states Government printing office, Washington: 1960.
Laudon, H., Sjöblom, V., Buffam, I., Seibert, J., Mörth, M., 2007. The role of
catchment scale and landscape characteristics for runoff generation of boreal streams. Journal of Hydrology 344, 198–209.
doi:10.1016/j.jhydrol.2007.07.010
Laudon, H., Taberman, I., Ågren, A., Futter, M., Ottoson-Löfvenius, M., Bishop, K.
(2013). The Krycklan Catchment Study - A flagship infrastructure for hydrology, biogeochemistry, and climate research in the boreal landscape. Water Resources Research, vol. 49. p. 7154-7158.
Mazurek, M. (2011). Geomorphological Processes In Channel Heads Initiated by Groundwater Outflows (The Parsęta Catchment, North-western Poland), Quaestiones Geographicae 30(3). p.33-45
McEachern, P., Prepas, E.E., Chanasyk, D.S., 2006. Landscape control of water chemistry in northern boreal streams of Alberta. Journal of Hydrology 323, p. 303–324. doi:10.1016/j.jhydrol.2005.09.016
Montgomery, D. R., Dietrich, W. E. (1988). Where do channels begin? Nature, vol.
336. p. 232-234.
Montgomery, D. R., & Dietrich, W. E. (1989). Source Areas, Drainage Density, and Channel Initiation. Water Resources Research, vol. 25, No. 8. p. 1907- 1918. USA, University of California.
17 Mosley, M. P., McKerchar, A. I. (1993). Streamflow, Chapter 8. Handbook of
Hydrology. Editor in Chief: Maidment, D. R. McGraw-Hill, Inc. United States of America.
Nihm, T. (2012). Variability of Intermittent Headwater streams in Boreal Landscape – influence of different discharge conditions. Department of Earth Science, Uppsala University.
Pornprommin, A., Izumi, N. (2010). Inception of stream incision by seepage erosion.
Journal of geophysical research, vol. 155. American Geophysical Union.
Qin, C-Z., Zhu, A-X., Pei. T., Li, B-L., Scholten, T., Behrens, T., Zhou, C-H. (2011).
An approach to computing topographic wetness index on maximum downslope gradient. Precision Agric 12, p. 32-43. DOI 10.1007/s11119- 009-9152-y
SGU, Svergies Geologiska Undersökning, (2014), Berggrundskarta 1:250 000, Lantmäteriet.
Wemple, B. C., Jones J. A., Grant G. E. (1996). Channel Network Extension By Logging Roads In Two Basins, Western Cascades, Oregon. Water Resources Bulletin, American Water Resources Association, vol. 32, no.6.
Zhang, W., Montgomery, D.R. (1994). Digital elevation model grid size, landscape representation, and hydrologic simulations. Water resources research, vol. 30. No. 4. p. 1019-1028. American Physical Union.
Internet sources
Svergies Meteorologiska och Hydrologiska Institut, (2014), Normal uppmätt årsmedelnederbörd 1961-1990,
http://www.smhi.se/klimatdata/meteorologi/nederbord/1.4160 2014-05-12
18
Appendix Appendix 1
Appendix 1.1 Boxplots displaying the mean slope for ditches (left) and the natural streams (right).
Appendix 1.2 Boxplots displaying the TWI for ditches (left) and natural streams (right).
Appendix 1.3 Boxplots displaying the logarithm of the maximum catchment area for ditches (left) and natural streams (right).
19
Appendix 2
Appendix 2.1 Redefined table of GPS-measurements in May 2013. GPS-coordinates are in WSG 84.
ID Latitud Longitud Elevation Renamed Description Site No 1 64,24398 19,79073 223,81 012 Man-made ditch 1 2 64,24662 19,79445 230,3386 022 Man-made ditch 2 3 64,24638 19,79125 229,2354 032 Man-made ditch 3 4 64,24274 19,79079 208,559 042 Man-made ditch 4 5 64,23888 19,79097 197,7595 052 Natural stream 5 6 64,23891 19,79076 196,1488 062 Natural stream 6 7 64,24225 19,78672 196,741 072 Natural stream 7 8 64,24527 19,77278 235,596 082 Man-made ditch 8 9 64,24525 19,77288 234,6996 092 Man-made ditch 9 10 64,24774 19,782 234,6183 102 Man-made ditch 10 11 64,24774 19,782 234,6183 112 Man-made ditch 11 13 64,20282 19,83581 206,3843 122 Man-made ditch 49 14 64,25129 19,80894 253,9643 132 Man-made ditch 35 15 64,25738 19,80015 297,8635 142 Man-made ditch 30 16 64,24999 19,80805 246,9303 152 Man-made ditch 36 17 64,25845 19,79947 302,7006 162 Man-made ditch 31 18 64,2587 19,80013 306,4234 172 Man-made ditch 32 19 64,25658 19,80723 301,9086 182 Man-made ditch 33 20 64,25599 19,80697 292,9517 192 Man-made ditch 26 21 64,25618 19,80547 307,8863 202 Man-made ditch 27 22 64,25506 19,80724 279,8982 212 Man-made ditch 28 23 64,25202 19,80919 267,1014 222 Man-made ditch 29 24 64,25128 19,8098 248,861 232 Man-made ditch 34 25 64,21952 19,75886 187,2344 242 Natural stream 45 26 64,21796 19,75526 220,341 252 Natural stream 46 27 64,21774 19,75572 225,5955 262 Natural stream 47 28 64,2182 19,75464 212,2351 272 Natural stream 48 29 64,25861 19,771 293,1956 282 Man-made ditch 18 30 64,25672 19,77107 291,0794 292 Man-made ditch 19 31 64,25553 19,77454 263,5634 302 Man-made ditch 20 32 64,25297 19,77494 259,4669 312 Man-made ditch 21 33 64,25325 19,77632 265,4725 322 Road ditch 22 34 64,25974 19,78383 266,0755 332 Road ditch 23 35 64,26015 19,78057 277,5352 342 Natural stream 24 36 64,26195 19,77469 293,5665 352 Man-made ditch 25 37 64,24993 19,77275 247,5235 362 Man-made ditch 12 38 64,25163 19,7749 251,7206 372 Man-made ditch 13 39 64,2511 19,77654 244,3093 382 Man-made ditch 14 40 64,2511 19,77654 244,3093 392 Man-made ditch 15 41 64,25536 19,77714 272,6787 402 Man-made ditch 16 42 64,25534 19,7772 270,7505 412 Man-made ditch 17
20
43 64,24908 19,80714 233,975 422 Man-made ditch 37 44 64,24833 19,80533 242,0448 432 Man-made ditch 38 45 64,2498 19,8035 244,3828 442 Man-made ditch 39 46 64,2499 19,80352 243,3575 452 Natural stream 40 47 64,25238 19,80306 269,1515 462 Man-made ditch 41 48 64,25372 19,8046 270,5672 472 Natural stream 42 49 64,25476 19,80506 284,2332 482 Man-made ditch 43 50 64,25647 19,80509 299,5201 492 Man-made ditch 44
21 Appendix 2.2 Redefined table of notes concerning data collected in May 2012, by Tum Nihm (2012)
Date Site
No Head
Water table depth (m)
Water width (m)
Channel depth (m)
Channel
width (m) Descriptions Formation
16/5/2
012 1 0,06 0,3 0,6 Ditch, no flow further upward Seepage from saturated flow
2 0,08 0,3 0,8 Ditch discontinuous further upward with water grass Seepage from saturated flow
3 0,1 0,3 0,4 0,6 Ditch continuous flow, slow flow Seepage from saturated flow
4 0,09 Not channelized, falling tree over. Natural flow Seepage erosion from under the tree roots
5 0,1 Not channelized. Natural flow with many small hollows Seepage erosion from under the tree roots
6 0,13 Flood plain or semi-wetland. Not channelized Saturated overland flow
7 0,12 0,4 0,4 0,8 Ditch with continuous flow Seepage from saturated flow
8 0,13 0,35 0,4 0,8 Ditch with continuous flow Seepage from saturated flow
9 0,11 0,6 0,7 1,2 Ditch sharing head with site 10 at the highest dividing elevation Seepage from saturated flow
10 0,11 0,6 0,7 1,2 Ditch sharing head with site 9 at the highest dividing elevation Seepage from saturated flow
17/5/2
012 11 0,05 0,6 0,3 0,7 Continuous flow, ditch
Seepage from saturated flow with the presence of grass
12 0,08 0,6 1 Continuous flow, ditch with snow cover
Seepage from saturated flow and snow melting contribution
13
Ditch covered by snow the whole stream length. Sharing head
with site 14
14
Ditch covered by snow the whole stream length, sharing head
with site 13
22
15 0,08 0,5 0,45 1,2 Covered by snow, ditch Seepage from saturated flow
16 0,02 0,35 0,3 0,8
Covered by snow, ditch with continuous flow, snow melting
contribute Seepage from saturated flow
17 0,04 0,45 0,4 1 ditch continuous flow Seepage from saturated flow
18 0,03 0,3 0,25 0,5 ditch continuous flow Seepage from saturated flow
19 0,06 0,3 0,4 0,6 Ditch continuous flow, snow melting contribute Seepage from saturated flow
20 0,05 0,4 0,85 1,7 Ditch no flow upward Seepage from saturated flow
21 0,05 0,8 Not channelized, widen road ditch Saturate surface flow
22 0,06 Widen road ditch, covered by snow at the end Saturated flow
23 0,1 Not channelized, dead grass, natural and continuous flow
24 0,12 0,4 0,4 Ending with a tree, snow covered, nearby a hillslope Seepage from saturated flow
25 Ditch
18/5/1
2 26 0,06 0,3 0,7 1,2
Discontinuous upwards of the stream head, presence of channel
erosion Seepage from saturated head
head
27 0,05 0,25 0,2 0,5
Ditch with discontinuous flow upwards with water grass,
presence of surface water Seepage from saturated head
head
28 0,05 0,3 0,35 0,4 Ditch with discontinuous flow upwards Seepage erosion
29 0,05 0,4 0,4 0,6 Ditch with continuous flow until the end Seepage from saturated head
30 0,06 0,45 0,57 1,05 Debris, killed grass, organic matter, at the base of hillslope ditch Seepage from saturated head
31 0,04 0,04 0,8 0,3 Presence of surface water upwards of the head, ditch Seepage erosion from under the tree roots
32 0,08 0,4 0,4 1 Sharing head with 33, covered by snow
