Group D
An investigation into factors affecting a flood forecasting model in a glaciated region.
Hannah O’Reilly, Rachael Niblett, Pamela Murray, Richard Jenkins,
Nasir Mahmood, Edward Krussman, Tom Hickson and Kathryn McGrath.
Contents
By page number
3. Abstract
3. Aims
4. Objectives
4. Hypotheses
4-5. Introduction and background
6-7. Methods
7-15. Results
16-17. Discussion
17. Limitations
17. Conclusion
18. References
Abstract
The Odenwinkelkees glacier, Austriais a place of scientific interest as it provides ample opportunities to study glacier behaviour such as ablation rates and proglacial stream behaviour. Information obtained about the characteristics of the glacier and the proglacial stream are vital in understanding the factors leading to flooding and the prediction of flooding. Measurements taken at 500m intervals and 50m distance from the stream showed a positive relationship between infiltration and slope and a significant relationship between vegetation cover and infiltration rate. Using a DTM model and flow accumulation model obtained from SalzburgUniversity we were able to see that as tree-line height at altitude increases discharge will decrease. The overall accuracy between the data we collected, paper maps and the GIS data sets was fairly good and our field measurements were at a higher resolution than that of the GIS thus improving the accuracy of our study.
Aims
The aim of this project is to identify the factors that affect flood forecasting downstream. We will do this by collecting the following data from various locations on either side of the proglacial stream running out of the Odenwinkelkees glacier, Austria:
- Infiltration rates
- Aspect
- Vegetation type and percentage cover
- Location
- Elevation
- Slope
We will then relate the collected data to melt rate results gathered over the whole week by two other groups, secondary long term local meteorological data and geological data. We will also map inputs in to the main river using GIS (Geographical Information System) software. This will hopefully allow us to produce a flood forecasting model data and calculate flood levels from given amounts of rainfall.
Objectives
- Our first objective is to collect all the data from the chosen sites in the field.
- The second objective is to collect local meteorological data and melt rate data and input this into our model
- The third objectiveis to input our data into a flood forecasting model.
- The forth objective is to map the location of the sample sites and the inputs into the main river using GIS and comparing GIS, GPS and OS map data in order to test the reliability of the data put into our model.
- The fifth objective is to investigate a relationship between altitude of the treeline and discharge.
- The sixth objective is to investigate a relationship between slope angle and infiltration rates.
Hypotheses
Null Hypothesis 1: An increasing tree-line height at altitude will not decrease the discharge.
Alternate Hypothesis 1: With increasing tree-line height at altitude discharge will decrease.
Null Hypothesis 2: There is no statistical relationship between slope angle and infiltration rates.
Alternate Hypothesis 2: There is a statistical relationship between slope angle and infiltration rates.
Null Hypothesis 3: There is no significant difference between GIS, GPS and OS map data.
Alternate Hypothesis 3: There is a significant difference between GIS, GPS and OS map data.
Null Hypothesis 4: There is no statistical relationship between vegetation and infiltration rates.
Alternate Hypothesis 4: There is a statistical relationship between vegetation and infiltration rates.
Introduction and Background:Why is it important to investigate factors that affect flood forecasting in a glaciated region?
The local area is situated in a glaciated region with heavy snow in winter and snow melt in summer. These changes cause seasonal differences in melt water inputs into the river. The ability for the river and surrounding flood plain to cope with these changes has significant impacts and will depend on local geology, vegetation type and cover, soil infiltration rates, slope angle and local meteorology as well as snow melt rates.
One of the most probing questions that we are faced with today is the impacts of global climate change on melt rates and subsequently flooding of the local area.
Proglacial flooding
As with normal river catchments, proglacial catchments experience periods of flooding. However the causes of this flooding can be quite different. Glacial flood events can be triggered by individual weather systems for example summer and autumn storms. These determine not only the intensity of rainfall but also how much increased ablation is likely to occur enabling more efficient melting by transferring heat from the atmosphere (Benn and Evans 1998) Changes to the glacial drainage systems also contributes to irregular flood events. The changes can include the collapse of englacial and subglacial channels the release of stored melt water behind sediment or ice jams or on a bigger scale ice dammed lakes. Such flood events have a great geomorphological impact (Maizels 1995 cited in Benn and Evans 1998) and can have in turn an effect on nearby towns and energy systems.
Tree-line and climate change
Changes in the climate have been long suspected of controlling the tree-line; “Compared to the second half of the 19th century present mean temperatures present a warming of 1.5K for the growing period (June to September) at tree-line elevations in the Alps… A general warming of 1.5K… corresponds to 250 m of elevation”, (Paulsen et al., 2000). This increase in elevation will lead to a decrease in discharge as the tree cover will intercept the runoff and through flow of the rain water, where as above the tree line less interception will occur.
Methods
To collect all the data in the field we used the following equipment:
- GPS
- Meter rule
- 1 bottle of water
- Infiltration meter
- OS map
- Clinometer
- Compass
To store and display our data we used the following equipment.
- Microsoft Excel
- GIS Software
Fieldwork Method
Due to the known terrain of the area, a stratified random sampling technique will be used. The transect began at 2km from the snout of the proglacial stream.After this point, a transect wasundertaken on each side of the valley running parallel with the stream.Measurements weretaken every 500m at a 50m distance from the stream where possible and from this we were able to decide on further sampling siteswhich will reflect the general area the best. At each sampling site, the following is measured:
- Aspectof each area was measured using a standard compass.
- Slope was measured by placing a clinometer on a metre rule that is facing the direction of greatest slope.
- Vegetation coverage and type was estimated using a 1m2 quadrat.
- Grid reference locationand elevation data was obtained using a handheld GPS.
- Infiltration was measured by placing a plastic bottle with its bottom cut off into the soil and then recording the time taken for 200ml of water to infiltrate completely.
Secondary data
Ablation stakes are used to measure the melt rates up the glacier.
GIS Method
Using GIS software we were able to use a DTM and a flow accumulation model to create discharge levels when the same level of precipitation occurred and when the tree line was at a different altitude.
Using GRID and ArcMap, the elevation, wetness, slope, and aspect data we collected was entered into the GIS. This data was then used to compare with the GIS data that was collected by SalzburgUniversity. Using ArcMap, the GIS data for elevation (DTM), slope, wetness index, and aspect was displayed, with the 12 site locations displayed on top. By clicking on each site, the data we collected in the field is displayed along with the GIS data, enabling us to compare the two data sets. This allowed us to analyse the data and spot any relationships or differences between the two.
Results
Table 1: Field Results Table
Site Number / Aspect / Slope (º) / Elevation (m) / GPS / Soil Moisture (1-5)1 / NW311º / 15º / 2050 / N47.13343: E012.263725 / 3
2 / NW291º / 1º / 2072 / N47.13142: E012.63426 / 3
3 / N/A / 0º / 2109 / N47.13142: E012.63485 / 2
4 / NW300º / 1º / 2112 / N47.13212: E012.63573 / 2
5 / NE032º / 10º / 2141 / N47.13212: E012.63897 / 2
6 / W280º / 30º / 2172 / N47.12358: E012.64020 / 2
7 / E083º / 20º / 2105 / N47.12265: E012.63752 / 2
8 / E090º / 4º / 2088 / N47.12674: E012.63314 / 0
9 / E080º / 6º / 2068 / N47.13310: E012.63430 / 2
10 / SE119º / 0º / 2085 / N47.12936: E012.63247 / 2
11 / S180º / 6º / 2063 / N47.13408: E012.63409 / 2
13 / S170º / 9º / 2239 / N47.13622: E012.63162 / 3
14 / E065º / 1º / 2265 / N47.13735: E012.63082 / 2
Site Number / Vegetation Cover / Total Vegetation Percentage Cover
1 / Grass, Rhoddendendum / 100%
2 / Grass, Bushes / 50%
3 / None / 0%
4 / Grass / 100%
5 / Moss / 100%
6 / Grass / 50%
7 / None / 0%
8 / None / 0%
9 / Daisies and shrubs / 90%
10 / Mosses and Gentinia / 80%
11 / Grass / 100%
13 / Grass / 100%
14 / Grass / 100%
Site Number / Infiltration (seconds) / Infiltration Velocity (cm/s)
1 / 25.055 / 0.1016
2 / 13 / 0.1958
3 / 19.86 / 0.1282
4 / 11.625 / 0.219
5 / 7 / 0.3637
6 / 4.25 / 0.5991
7 / 4.54 / 0.561
8 / 3.64 / 0.6995
9 / 38.03 / 0.0669
10 / 31.32 / 0.0813
11 / 3.5 / 0.7274
13 / 9.08 / 0.2804
14 / 33.57 / 0.0758
Using our data in table 2, we statistically tested the relationship between slope angle and infiltration velocity. Figure 1 and figure 2 display our results which will be explored in the discussion section.
Slope (º) / Infiltration Velocity (cm/s)15 / 0.1016
1 / 0.1958
0 / 0.1282
1 / 0.219
10 / 0.3637
30 / 0.5991
20 / 0.561
6 / 0.0669
0 / 0.0813
9 / 0.2804
1 / 0.0758
Table 2: Slope and Infiltration
Fig 1: Graph showing relationship between slope angle infiltration velocity.
Fig 2: Minitab regression results
Regression Analysis: C1 versus C2
The regression equation is
C1 = 0.109 + 0.0158 C2
Predictor Coef SE Coef T P
Constant 0.10918 0.04792 2.28 0.049
C2 0.015825 0.003804 4.16 0.002
S = 0.117795 R-Sq = 65.8% R-Sq(adj) = 62.0%
Analysis of Variance
Source DF SS MS F P
Regression 1 0.24011 0.24011 17.30 0.002
Residual Error 9 0.12488 0.01388
Total 10 0.36499
Unusual Observations
Obs C2 C1 Fit SE Fit Residual St Resid
1 15.0 0.1016 0.3466 0.0434 -0.2450 -2.24R
6 30.0 0.5991 0.5839 0.0893 0.0152 0.20 X
R denotes an observation with a large standardized residual.
X denotes an observation whose X value gives it large influence.
Using a DTM model and flow accumulation model we were able to investigate the effects of the altitude of the tree line on discharge. Table 3 displays the data we used and figure 2 displays our results. We moved the tree line up every 100m to explore the effects and we could see that if there is a higher the tree line there is a decrease in discharge.
Altitude of treeline (m) / Discharge (cumecs)2000 / 31916
2100 / 31002
2200 / 29776
2300 / 28230
2400 / 26366
2500 / 24676
2600 / 22756
2700 / 20835
Table 3: Treeline Altitude and Discharge
Fig 3 A graph to show how discharge differs with altitude.
The relationship between total percentage cover and infiltration velocity was also investigated using data from table 1 and the results are shown in figure 3.
Fig 4: A scatterplot to show how total percentage cover against infiltration velocity
GIS Results
elevation / aspect / slope / wetnessours / map / gis / ours / gis / ours / gis / ours / gis
2050 / 2040 / 2056 / 311 / 0 / 15 / 11 / 16 / 8
2072 / 2070 / 2073 / 291 / 5 / 1 / 4 / 12 / 8
2109 / 2110 / 2103 / 0 / 5 / 0 / 5 / 8 / 12
2112 / 2115 / 2109 / 300 / 5 / 1 / 4 / 8 / 13
2141 / 2140 / 2138 / 320 / 4 / 10 / 1 / 8 / 11
2172 / 2180 / 2158 / 280 / 4 / 30 / 25 / 8 / 8
2105 / 2110 / 2149 / 90 / 1 / 20 / 16 / 8 / 7
2088 / 2080 / 2105 / 43 / 0 / 5 / 8 / 8 / 9
2085 / 2090 / 2082 / 119 / 0 / 9 / 7 / 8 / 11
2068 / 2080 / 2059 / 80 / 0 / 6 / 13 / 8 / 11
2063 / 2060 / 2080 / 180 / 2 / 6 / 12 / 8 / 8
2106 / 2100 / 2106 / 180 / 2 / 15 / 25 / 16 / 6
Table 4: GIS, Map and GPS results
Fig 5: A graph to show the slope measurements from GIS and clinometer readings
Fig 6: A graph to show the wetness measurements from GIS and field measurements.
Fig 7: A graph to show the comparisons between aspect measurements
Fig 8: A graph to show the comparisons between elevation measurements from GIS, GPS and maps.
We then statistically tested the elevation of our results, the GIS elevation results and the elevation map using an anova test and the boxplot below displays our results.
Fig 9: Boxplot of our field elevation results, GIS elevation results and the map elevation.
One-way ANOVA: elevation us, elevation gis, elevation map
Source DF SS MS F P
Factor 2 113 57 0.05 0.955
Error 33 40903 1239
Total 35 41016
S = 35.21 R-Sq = 0.28% R-Sq(adj) = 0.00%
Individual 95% CIs For Mean Based on
Pooled StDev
Level N Mean StDev ------+------+------+------+
elevation us 12 2097.6 34.5 (------*------)
elevation gis 12 2101.5 33.6 (------*------)
elevation map 12 2097.9 37.4 (------*------)
------+------+------+------+
2088 2100 2112 2124
Pooled StDev = 35.2
Boxplot of elevation us, elevation gis, elevation map
One-Sample T: slope us, slope gis
Variable N Mean StDev SE Mean 95% CI
slope us 12 9.83333 8.89160 2.56678 (4.18388, 15.48279)
slope gis 12 10.7500 7.8964 2.2795 ( 5.7328, 15.7672)
Boxplot of slope us
Boxplot of slope gis
One-Sample T: slope us, slope gis
Variable N Mean StDev SE Mean 95% CI
slope us 12 9.83333 8.89160 2.56678 (4.18388, 15.48279)
slope gis 12 10.7500 7.8964 2.2795 ( 5.7328, 15.7672)
Boxplot of slope us
Boxplot of slope gis
Two-Sample T-Test and CI: slope us, slope gis
Two-sample T for slope us vs slope gis
N Mean StDev SE Mean
slope us 12 9.83 8.89 2.6
slope gis 12 10.75 7.90 2.3
Difference = mu (slope us) - mu (slope gis)
Estimate for difference: -0.916667
95% CI for difference: (-8.055688, 6.222354)
T-Test of difference = 0 (vs not =): T-Value = -0.27 P-Value = 0.792 DF = 21
Fig 4: Minitab display of our elevation results, the GIS results and the map elevation
1
Discussion
From the graph we can see there is a trend for infiltration rate to increase with increasing slope. From this we performed a regression analysis to determine how much of the variance in infiltration could be explained by changes in slope. The result we gained showed that 65% of the data could be explained by slope. The regression was shown to be robust through testing for autocorrelation and normality within the residuals. The p value of the regression of 0.002 meant this was a significant result.
This shows that there is a relationship of causality between slope and infiltration. This positive correlation could be due to the condition of the ground at the start of the rainfall. With steeper slopes the effect of gravity means through flow removes water from the soil very quickly leaving an unsaturated soil. Thus as new rain falls it is able to infiltrate rapidly into the soil. It could also be said however that slope would reduce infiltration as water is more likely to be removed via overland flow on steeper slopes although our results support the ideas of the positive correlation.
The relationship between infiltration velocity and percentage cover was statistically tested using Pearson’s product moment correlation coefficient. The p value was 0.068 which was below the 95% confidence level. This shows that there was no significant relationship between these two factors. However, if there was a significant relationship, vegetation can decrease infiltration by providing a layer in the soil for interception or it could increase because of the rooting system.
The results for the GIS section, displayed in figures 3-6 and in table 4, showed that the elevation data that was collected in the field had very similar readings. The GPS and GIS data were compared with an OS map which showed that the GPS was more accurate than the GIS data, as it closely matched the map data. The aspect data was very different to the GIS data, which may be due to the variation in terrain and the fact that the results recorded in the field were at a much higher resolution than the GIS. The wetness index was also very different to the GIS data, which is probably a result of the weather conditions over the past few days. Again, the slope results were similar to each other, but the field data was at a higher resolution, with the GIS data accurate to 25m and GPS to 7m, therefore the field data being more accurate. Slope and elevation was statistically tested by carrying out anova tests which showed no significance. This is because the p values at 0.955 and 0.792 were below the 95% confidence level.
Limitations
We came across the following limitations:
Vegetation cover estimates are subjective as is vegetation identification.
Infiltration rates were difficult to obtain in some areas because the soil was already fully saturated from a previous storm.
We were unable to take an equal amount of samples on both sides of the river because we were limited by the size of the area, health and safety.
Time constraints meant that we were unable to take multiple measurements at each site hence we were restricted to 2 readings at each site.
Time limitations also hindered the development of a complex flood forecasting model.
The GPS readings had an estimated 30 meter accuracy reading so a map would have been more accurate.
We also used 2 different models of GPS which may have had a significant impact on our results.
Conclusion
In conclusion we found that the overall accuracy between the data we collected, paper maps and the GIS data sets was fairly good and our field measurements were at a higher resolution than that of the GIS thus improving the accuracy of our study, hence showing the importance of going into the field for more accurate investigation. With regards to the slope angle and infiltration we found that there was a statistical relationship as slope angle increased the water infiltrated into the soil more quickly. From using the flow accumulation model we have found that as altitude increases there is a decline in discharge. If we were to further this investigation the results would have been used in a flood forecasting model.
References
BENN AND EVANS,1998,Glaciers and Glaciation, Arnold
J PAULSEN, U.M WEBER, and CH KÖRNER, 2003, Tree Growth Near Treeline: Abrupt or Gradual Reduction with altitude, Arctic, Antarctic and Alpine Research, 32, 1, 14-20.
1
1