Development of a potential index map for the initiation of small-scale blowouts in Meijendel

(1)

Development of a potential index map for the

initiation of small-scale blowouts in Meijendel

By Floris Veloo

Internship at Dunea

Part of the Msc Earth Science Environmental track (18 EC)

Daily supervisors: HGJM van der Hagen & K. Rood Uva Examiner: dhr. dr. K.F. Rijsdijk

Date: 03-06-2020 UvAnetID: 10628150

(2)

Thematic summary

As a result of nitrogen deposition and the decline of the rabbit population, vegetation succession is accelerating while the biodiversity is in decline in the dunes of Meijendel. The nature managers of Dunea seek ways to increase biodiversity, counteract vegetation succession and maintain the mosaic landscape. The distribution of relatively calcareous sand by blowouts over the topsoil could help to reset succession, create opportunities for pioneer species and help in maintaining the preferred mosaic landscape. With the use of sand dynamics from blowouts and accurate implementation not only nature management costs can be minimized, but also the goals of the interventions can easier be achieved. Therefore the aim of this research is to develop a potential index for the initiation of small-scale blowouts in Meijendel. Four factors are selected as model parameters due to their importance for the initiation of small-scale blowouts: slope, aspect, elevation and vegetation. The factors are indexed, reclassified and summed in order to create a potential index. In order to understand and test the potential index, the results are analysed with the use of the input data. The accuracy of the potential index is tested on actual blowouts in the field and by using orthophotos since 1938. The analysis showed that the vegetation and aspect indices have the largest influence on the patterns of the potential index, while the elevation and slope indices influence the potential hotspots the most. Furthermore, except for the aspect, blowouts score higher at every index than would be expected by just looking at the class distribution of the indices. Moreover, more than 50% of the selected hotspots are within 1 gridcell away from a blowout in each of the analysed orthophotos. Therefore, the potential index could predict potential locations for the initiation of blowouts, but the accuracy of the model needs to be improved.

(3)

1. Introduction

1.1 The relevance of mapping small-scale blowouts in Meijendel

Dunea is a drinking water company, which provides drinking water to inhabitants in the western part of Zuid-Holland. Dunea uses the dunes for surface water infiltration. The dunes filter the water, which can later be pumped up and used as drinking water. Besides, Dunea has an active role in the (nature) management of the dunes of Solleveld, Berkheide and Meijendel. The dunes provide various ecosystem services, are a crucial natural defence line against coastal flooding and are part of the Natura2000 network. Therefore,

management aims to conserve and protect the dune ecosystem (Breedveld et al. 2016; van Engeldorp Gastelaars B. & Rood K. 2010).

As a result of nitrogen deposition and the decline of the rabbit population, vegetation succession is accelerating while the biodiversity is in decline in the dunes of Meijendel. Because biodiversity is the main indication of the resilience of an ecosystem, nature managers seek ways to increase biodiversity, counteract vegetation succession and

maintain the mosaic landscape. The initiation of small-scale blowouts could potentially help to increase the biodiversity and could help to reach the conservation goals set by Natura 2000 (Aggenbach et al. 2018).

Moreover, soil development follows vegetation succession. As succession continues, organic material in the soil accumulates and the acidity increases. A soil containing a

relatively high organic matter content, can hold more water for longer periods of time. As the soil develops, soil conditions change resulting in new and better environmental conditions for plant species with a higher biomass. Distribution of relatively calcareous sand over the topsoil could reset succession, create opportunities for pioneer species and help in maintaining the preferred mosaic landscape (Aggenbach et al. 2018).

From a natural perspective, small-scale blowouts distribute relatively calcareous sand over the surrounding areas and increase the pH of the topsoil. With the use of sand dynamics from blowouts and accurate implementation not only nature management costs can be minimized, but also the goals of the interventions can easier be achieved. Therefore the aim of this research is to develop a potential index for the initiation of small-scale blowouts in Meijendel.

(5)

1.2 Research Questions

In order to develop the potential index for the initiation of small-scale blowouts the following research question was formulated:

Which parameters should be selected for the calculation of a potential map for small-scale blowouts in Meijendel?

In order to assess and validate the potential index for the initiation of small-scale blowouts the following research questions were formulated:

a. To what extent is the potential index capable of actually predicting blowouts in the field?

b. To what extent are blowouts linked to the potential index score?

1.3 Importance of several factors for the initiation of small-scale blowouts.

The following factors are selected as model parameters due to their importance for the initiation of small-scale blowouts (Aggenbach et al. 2018): slope, aspect, elevation and vegetation. Hereby, the focus is on the initiation process rather than the maintenance. Water erosion has a large influence on the initiation of small-scale blowouts. As a result of the radiation of the sun, the topsoil and the vegetation can dehydrate. The vegetation dies, while the topsoil becomes water repellent. As a result, the precipitation can not penetrate the soil and runoff occurs. The topsoil is eroded by runoff resulting in the exposure of the sand beneath.

Dehydration of the vegetation and topsoil during the summer months has more impact on the southern slope as this hillside is longer exposed to the sun. This effect is larger on southern slopes than on northern slopes making the aspect factor a good indication for the potential of water erosion.

Furthermore, water erosion has in general more impact on steeper slopes. The steepness of a slope determines the velocity of the runoff and therefore the potential of the water to erode soil particles and transport them downhill. Because the slope factor is a good indication for the occurrence of water erosion, it is included as model parameter.

When the topsoil is removed and the sand is visible at the surface, the wind can start transporting the sand to the surrounding area. The wind is an important factor for the growth of blowouts and in the distribution of sand over the surrounding area. In Meijendel, the dominating wind direction is from the southwest. However, a long-lasting supply of saturated southwestern winds can be as important for the development of a blowout as dry eastern winds or storms. Because local wind properties are too dynamic for calculation, the elevation is used as a wind speed indicator. The exposure to wind determines the probability that a blowout develops once initiated. Locations that lie relatively high in the landscape are often more exposed to wind. In this research, the relative elevation is simplified through the use of the actual elevation.

(6)

2. Method

2.1 General overview

Index maps can be used to discover spatial patterns and trends in the occurrence of natural features in different areas. Natural features are often the result of a combination of several factors. Combining the main factors that determine the presence of a natural feature, the likelihood of a location to contain such a feature can be mapped. Such a map can give a first indication for further searching.

In this research, several factors were selected that determine the potential of a location for the natural initiation of small-scale blowouts. The factors were selected based on their relevance for the initiation of small-scale blowouts and on the datasets available. These main factors were used as model parameters and were indexed from which a potential index is calculated. The calculation of the potential index (PI) is based on 4 parameters that determine the likelihood of the initiation of blowouts; elevation (Ei), slope (Si), aspect (Ai) and vegetation (Vi). A formula is developed in which each parameter is of equal importance (weight = 1), and is formulated as (1):

(1) PI = Ei + Si + Ai + Vi

The model can be adjusted to different dune areas and local preferences. Also the cell size and weights of the sub-indices can be adapted if preferable. Therefore, the same method can be adopted for different circumstances and goals, while only adjusted a few parameters.

In order to make the research reproducible, transparent and constructive, a generalized workflow has been created (Figure 1). The workflow consists of four main phases; preparation, calculation of the indices, data analysis and the deliverables. During the preparation phase, the literature on blowouts is reviewed and data is preprocessed (Breedveld et al. 2016; Aggenbach et al. 2018;van Engeldorp Gastelaars & Rood, 2010). Also, fieldwork was conducted in order to localize actual blowouts in the field and to

determine the current state of the vegetation cover. The fieldwork was done to acquire field data of blowouts for the accuracy assessment as well as to get a better understanding of the current state and history of Meijendel. Then the sub-indices are calculated as well as the final index maps. Furthermore, the data analysis shows the influence of the individual sub-indices on the patterns in the final index maps. During the data analysis, the average index scores of the actual blowouts in the field are analysed based on their vegetation cover. Also, orthophotos since 1938 are used to determine the accuracy of the model in predicting the occurrence of blowouts. The deliverables will help in the understanding and improvement of the model. The workflow will be further elaborated on the next paragraphs.

(7)

Figure 1. Simplified overview of the workflow.

Calculations and analysis of the data were done with ArcGIS Pro 2.5.0. Input data, interim calculations and final deliverables can be found in the data folder. Metadata of the elevation and vegetation input data can be found in Appendix B (Table 3). During the

research, the same geographic coordinate system (GCS_Amersfoort) has been used as the input data obtained from Dunea. The data is projected in the RD_New coordinate system. The data analysis was conducted with ArcGIS as well as Excel. Original data used in the data analysis can be found in Appendix B (Tables 4 & 6). Furthermore, an ArcGIS model is made to show interim calculations and to automate the digital process. The model can be found in the toolbox within the geodatabase and an overview is given in Appendix (Figures 12-15).

2.2 Development of a study area template

An important preprocessing step is the development of a template polygon for Meijendel. This template is created in order to outline the study area within Meijendel. The study area template is used for the clipping of the calculation grid, which in turn acts as a template to clip all the input data. In order to be able to calculate a potential index, the raster cells and area extent of the sub-indices have to be of equal size. The study area template is the basis on which the workflow depends.

In order to create a study area template, the vegetation map was copied and dissolved into 1 polygon. During an edit-session, the template was modified to exclude landscape elements that were not interesting for the initiation of small-scale blowouts. The landscape element data was retrieved from the database of Dunea and includes among

(8)

others buildings, water bodies, beaches and forests. Metadata of landscape elements can be found in the ReadMe-folder of the data folder.

2.3 Creation of a calculation grid

The calculation grid overlays the input data. For each raster cell of the calculation grid, an index is calculated based on the value range of the input data. The cell size used for this grid is based on recommendations given by Hengl et al. (2006). The recommended grid size depends on the grid size of the input datasets and the aim of the research. Because the elevation data from the AHN3 has a cell size of 0.5mx0.5m, a calculation grid cell size was chosen of a factor 10 larger; 5mx5m. This cell size was also selected in order to have a sufficient amount of data points in the calculation of a grid cell. The ‘create-fishnet’-tool was used to create a grid. The calculation grid was then clipped to the template of the study area (Figure 2). In order for each grid cell to have the same amount of input data points and an equal influence on the calculation of the indices, only whole cells were selected.

Figure 2. The calculation grid with equal size cells clipped to the study area

template.

(9)

After the main factors were identified that determine the potential of a location for blowout development, the sub-index maps were created. The extent to which a factor makes a location suitable for blowout development is based on the literature review and the method of classification (Aggenbach et al. 2018). The sub-indices are classified into 5 classes. The classes are classified from ‘very low’ to ‘very high’ potential. The classes represent the sensitivity of a location to the factor its potential for the natural initiation of small-scale blowouts.

The ‘zonal-statistics’-tool was used to calculate the average slope angle and elevation for each grid cell. Then the slope angle index and the elevation index were classified according to the ‘equal interval’-method. This method results in an equal

distribution of the classes between the minimum and maximum values within the study area. Therefore, the index values cover the average linear gradient of the actual values of the two factors. In other words, the average steepness of a slope is equally distributed over the index classes.

Furthermore, the ‘zonal-statistics’-tool was also used for the calculation of the aspect index and the vegetation index. Because these factors are categorical variables, it was necessary to classify the categories prior to the index calculations. Firstly, the ‘polygon-to-raster’-tool was used to transform the vegetation data to a raster format. Then, the

‘reclassify’-tool was used to assign a sensitivity value to the categories. Lastly, the average sensitivity score was calculated per grid cell and reclassified into 5 classes with the natural breaks method (Jenks). This method of classification reduces the variance within classes, while maximizing the variance between classes. Therefore, the index values are

representing the average sensitivity score of a grid cell in comparison to the distribution of the input data. Data tables on the (pre)classification of the aspect and vegetation are given in Appendix B (Tables 7 & 8).

2.5 Calculation of the small-scale blowout potential map

The potential index formula (1) was used in the ‘raster calculator’-tool to calculate the potential index. Similarly, the landscape parameter index was calculated by excluding the vegetation index from the formula (1). Both formulas use an equal weight of 1 for all sub-indices to ensure that each sub-index has an equal influence on the final sub-indices. The result is a raster in which each cell contains the summation of the sub-index scores. The 2 output rasters are then reclassified into 5 classes as summarized in table 1. A manual method for reclassification was chosen, because it enables the user to manually change the amount of ‘very high’ potential areas. These are the areas that are interesting for analysis and the amount of areas should be adaptable to different types of analyses. In this research, the index scores of the output rasters are about equally distributed over the reclassified index, while maintaining a usable amount of ‘very high’-potential classes for fieldwork reviews.

(10)

Blowout development index

Index score without reclassification

Landscape parameters Potential Index 1. Very low 3-5 4-7 2. Low 6-7 8-10 3. Moderate 8-9 11-13 4. High 10-11 14-16 5. Very high 12-14 17-19

2.6 The analysis

To show the degree of accuracy of the research, a data analysis is carried out. The data analysis helps to show to what extent the index maps predicts actual blowout

development in the field. The data analysis also reveals how the potential map is constructed and provides a framework for future improvements to be made. The data analysis consists of the following components: model parameter analysis, blowout index score rating analysis and a model accuracy assessment for the prediction of blowouts.

2.6.1 Model parameter analysis

To analyze the influence of the individual sub-indices on the potential index and landscape parameter index the ‘band collection statistics’-tool was used to create a correlation matrix. These tables summarize the correlation coefficients and thus represent the degree of linear dependency between the indices. In order to understand the origin of the patterns, the sub-index maps are visually analysed and interpreted. Also, the distribution of the input data over the index classes is taken into account. A comparison between the landscape parameter index and potential index shows the role of vegetation on the potential index.

2.6.2 Index score ratings of small-scale blowouts

In order to show how well actual blowouts score on the classes of the indices, fieldwork was conducted. During the fieldwork, 61 small-scale blowouts were localized of which 54 fell within the borders of the calculation grid. The blowouts were ranked based on

(11)

their aeolian activity by taking the amount and type of vegetation cover into account. 31 Blowouts were identified as aeolian active, while 23 blowout were identified as vegetated.

The calculation grid was used to overlay the basemap with a transparency of about 50%. The cells that partially covered a blowout were selected from the calculation grid and dissolved to 1 polygon feature, which would later be merged with the other 54 blowouts into 1 dataset. The ‘zonal statistics’-tool was used to calculate the average index scores per blowout. The resulting 54 blowout data points per index are added to Appendix B (Table 4).

The analysis shows how the 54 blowouts are distributed over the classes of the indices. Also a distinction has been made between aeolian active blowouts and blowouts covered with vegetation. The distinction shows the difference in index scores between active and vegetated blowouts. As a result, the theoretical GIS-model can be tested and validated on actual blowouts in the field.

2.6.3 Model accuracy assessment for the prediction of blowouts

In order to show how accurate the model predicts actual blowouts in the field, a model accuracy assessment has been carried out. From the 109 ‘very high’-grid cells of the potential index (or hotspots), 30 random cells were selected and analysed with the use of orthophotos on the presence of a blowout. In this case, a blowout is defined as a natural surface that is not covered by vegetation and can be identified as the epicentrum of the surrounding distributed sand. A distinction was also made between rabbit holes and actual blowouts. Orthophotos were used of the year 1938, 1975 and 2016. These orthophotos were selected to give a coarse but rather large overview of the history of blowout development within Meijendel.

For each orthophoto and cell, the presence of a blowout within the cell was documented in Appendix B (Table 6) as well as the presence of a blowout within 1 cell in the surroundings. The analysis estimates the accuracy of the model to predict blowout development of the past 82 years from orthographic data.

3. Results

(12)

Figure 3 shows an overview of the potential index map for the initiation of small-scale blowouts in Meijendel. The index is classified into 5 categories ranging from ‘very low’ to ‘very high’ potential. Most of the potential hotspots are found on the elevated inner dunes or on the foredunes with some exceptions in the middle dunes. (old) Agricultural land scores lower on the index, because these lands lay low in the landscape and are mostly flat. Figure 4 shows about the same trends for the landscape parameter index, but the distinctions between the classes is better visible. Besides, the vegetation index has a large influence on the presence of local index hotspots. Maps of the sub-indices are added to Appendix A (Figures 7-10).

(13)

Figure 4. Landscape parameter index map of small-scale blowouts in Meijendel.

Figure 5 shows the class distribution of the classes of the sub-indices and the distribution of the classes of the final indices. Noticeable is the similar distribution of the elevation and slope index. The aspect and vegetation have a larger share in the higher index classes, while the slope and elevation do better in the middle and lower index classes. The landscape parameter and final indices consist mainly of lower index classes, while there are only a few higher classes present. The few higher classes are the result of the distribution of the index classes as well as the final classification.

(14)

Figure 5. Distribution of the index classes per index map.

Table 2 shows the band collection statistics, which summarizes the correlation coefficients between the indices. A high value would mean that the patterns of both indices are quite similar. The vegetation index pattern has the largest similarity with the potential index, while the patterns of the elevation index correspond the least with the potential index. The patterns of the elevation and slope index slightly match. However, none of the other sub-indices have a similar pattern to a sub-index, meaning that their index patterns do not match. In other words, the influence of the vegetation and aspect indices on the patterns of the potential index is mainly the result of the sub-index, while the influence of the elevation and slope indices on the patterns of the potential index are partly the result of an overlap in patterns between both indices. Therefore, very high potential locations in the map are more influenced by the combined distribution of the slope and elevation index, than the correlation coefficients would directly indicate.

Table 5 in Appendix B summarizes the band collection statistics for the landscape parameter index. All three parameters have about equal influence on the patterns of the landscape parameter index, while the elevation and slope influence hotspots the most.

Table 2. Band collection statistics of the Potential index map

Final Vegetation Aspect Elevation Slope

Final 1

Vegetation 0,57420 1

Aspect 0,47119 0,03376 1

Elevation 0,36924 -0,03063 -0,03063 1

Slope 0,41047 -0,00415 -0,00415 0,23718 1

3.2 Accuracy assessment of the potential index on blowouts

3.2.1 Distribution of indices over the small-scale blowout locations

Figure 6 shows the average index scores of 31 active and 23 vegetated blowouts in the study area. The average index scores are a representation/sample of how well actual blowouts in the field score and the influence of the distribution of the classes of the potential index. The index score for active and vegetated blowouts is for the vegetation relatively high. Therefore, the potential index is both for active and vegetated blowouts larger than the landscape parameter index. This is the result of the high vegetation index score. Except for the aspect, all blowouts score on average a higher index than the average class distribution of the sub-indices.

(15)

The average index scores of blowouts is dependent on the amount of cells needed to calculate this average index score. The average index scores of the blowouts is calculated by the number of grid cells it covers. The average amount of cells per blowout is 9,1, which is 8,4 for blowouts covered by vegetation and 9,5 for aeolian active. In Appendix B, Table 10 summarizes the index statistics for actual blowouts, while Figure 11 in Appendix C shows the amount of calculation cells used per blowout. In order for the average index score per blowout to be significant, the number of grid cells should be increased per blowout.

Figure 6. Average index scores of active and vegetated blowouts.

3.2.2. Model accuracy on the prediction of blowouts

From the 30 random selected cells of the ‘very high’ potential class in Appendix B (Table 6), 23% was found to cover a portion of a blowout in 2016, 20% in 1975 and 43% in 1938. In 2016, in 60% of the cases a blowout was located within 1 of the surrounding cells. In 1975 and 1938 this was respectively 50% and 87%. In all years combined, a blowout was always present within one adjacent cell of the hotspot (except for 1). A blowout was present within the hotspot cell in 2 out of the 3 cases.

4. Discussion

Meijendel has a rich history in which social, cultural and economic influences have shaped the landscape. Even though the development of blowouts is dependent on a

multitude of factors, the development of blowouts in Meijendel has also been dependent on the functions that the area had had in the past. Therefore, the results of the research should also be seen in light of the anthropogenic influences that shaped the landscape.

Nevertheless, this research was created to act as a framework on which to further develop the model.

(16)

The vegetation index followed by the aspect index are most correlated with the potential index. Therefore, these indices have the largest influence on the patterns in the potential index map. On average, the vegetation and aspect indices are well represented in the higher index classes, while their share in the lower index classes is moderate. On the contrary, the slope and elevation index are underrepresented in the higher index classes, while both indices have a large share in the lower index classes. The influence of the vegetation and aspect indices are better divided over the classes of the potential index, which results in a larger influence on the eventual patterns. However, because the slope and elevation index are slightly correlated to each other, the influence of these indices on the hotspot areas is rather large.

Actual blowouts in the field score slightly higher on the potential index than on the landscape parameter index. This is likely the result of the high index score of blowouts for vegetation. Blowouts score lower on the slope and elevation index. However, because the index score of a blowout represents the average index score between multiple index cells, the index score of a blowout is partly influenced by the cell distribution of the classes of the index. As a result of the larger share of the vegetation and aspect indices in the higher index classes, these index scores were expected to be higher than for the elevation and slope index. Noticeable is the difference between the index scores of the blowouts and the

average class distribution of the parameters. In comparison, blowouts score higher at every index, except the aspect, than would be expected by just looking at the class distribution of the index.

Furthermore, it should be kept in mind that this analysis is oriented at already developed blowouts instead of the initiation phase of a blowout. The potential index was developed to map the potential hotspot areas for the initiation phase. Therefore, blowouts in a later state of development should score differently than potential blowout locations would have scored in the initiation phase.

In the field, blowouts that are covered by vegetation score slightly higher on every index than active blowouts with the exception of the aspect. An explanation could be found in the vegetation index, because the difference between active and vegetated blowouts should be most visible here. However, the vegetation index of active and vegetated blowouts does not differ as much as would be expected. During the fieldwork, a blowout would be identified as vegetated even if the vegetation cover was low. Therefore, the difference in vegetation cover between active and vegetated blowouts is small. However, the difference in vegetation cover between the classes of the vegetation index is in comparison rather large. For example, bare soil would be classified into the same index class as certain lower vegetation types. As a result, active and vegetated blowouts score similar on the vegetation index and thus score similar on the potential index.

Potential index hotspots could predict potential locations for blowout development. In more than half of the hotspots a blowout was present within one adjacent cell in each period that the orthophotos were made. Most blowouts were detected within an hotspot in 1938 as well as within one adjacent cell. This trend corresponds to the amount of blowouts that were present at the time. Therefore, the average likelihood for a blowout to fall within the borders of the hotspot cell was larger in 1938 than in the other years. The increase in model

accuracy in 1938 is likely to be the result of the higher presence of blowouts. Nevertheless, in all years combined, a blowout was present within one adjacent cell of the hotspot (except for 1). A blowout was present within the hotspot cell in 2 out of the 3 cases.

(17)

4.2 Recommendations for future improvements

The patterns and trends that are shown by index maps are among others dependent on the cell size of the calculation grid and of the input data. Even though, the current cell size of the index maps are large enough to show patterns in the landscape, a smaller cell size would be better capable of catching fine-scale patterns for landscape elements such as small-scale blowouts. Firstly, the cell size of 5 m2_{could be increased to 3 m}2_{. A smaller cell}

size would increase the accuracy of the model to visualize patterns. On the contrary, the statistical precision of individual cells would decrease. As a result of indexation as well as maintaining a sufficient amount of input data per grid cell, the precision per cell can be sufficient enough while still being able to visualize patterns in the landscape. Figure 11 in Appendix C shows the scarce amount of cells that were used per blowout. In this case, a smaller cell size would result in more index cells used in the calculation of the average blowout score. Another benefit of a small cell size is the amount of additional area that can be included in the study area. This could increase the amount and the accuracy of blowouts that are included in the assessment, because often a blowout was found at the borders of the study area. Also the accuracy assessments of blowouts would become more accurate. Smaller index cells are better able to outline and thus represent actual blowouts in the field. Also, the distance of ‘very high’ class cells to blowouts can be better represented by

including more grid cells.

Wind speeds have a large influence on blowout development, because the wind fuels the blowout once initiated. Due to the amount of different features in the landscape, wind speeds near the surface differ locally and from the seaside part of the dune inland. The influence of the wind is less for relatively low places in the landscape (Aggenbach et al. 2018) and for the places on the lee-side of a landscape feature. Therefore, the cells at the borders of the study area template are on average likely to be influenced more by wind obstruction than cells near the middle. Even though the distance is reduced by removing incomplete cells at the border of the calculation grid, the influence of wind plays still a large role in the development of blowouts at a local scale. Therefore, the model could be improved by including a wind speed parameter that takes into account the distance of a grid cell to a feature. This parameter should also include the relative elevation of a cell compared to the surrounding cells. As a result, the influence of the elevation is better represented in the model and potential hotspots could be less affected by the combined influence of the slope and elevation indices.

Some of the blowout locations that were visited during the fieldwork were not included in the assessment. This was because the study area template did not cover these locations. There were several reasons why these locations were not covered; certain habitat types were not included (H0000), habitat types have changed since the creation of the vegetation map and at the time did not have the correct habitat type, and certain areas belonged to subareas of Meijendel such as Berkheide.

(18)

The development of a potential index has resulted in a potential index map for the initiation of small-scale blowouts in Meijendel. The selected parameters are useful in the mapping of potential blowout locations and help to explain the potential patterns and hotspots.

The vegetation index followed by the aspect index are most correlated with the potential index. The influence of the vegetation and aspect indices are better divided over the classes of the potential index, which results in a larger influence on the eventual patterns. However, because the slope and elevation index are slightly correlated to each other, the influence of these indices on the hotspot areas is large.

In the field, blowouts that are covered by vegetation score slightly higher on every index than active blowouts with the exception of the aspect. An explanation could be found in the large range of vegetation types that the classes of the vegetation index cover and the way of identifying vegetation cover of blowouts in the field. As a result, active and vegetated blowouts score similar on the vegetation index and thus score similar on the potential index. Nevertheless, blowouts score higher at every index, except the aspect, than would be expected by just looking at the class distribution of the indices.

Potential index hotspots could predict potential locations for blowout development. However, the model is currently not accurate enough. A smaller cell size of both the input data and the calculation grid as well as further research could improve the accuracy.

The potential index gives a general overview of the patterns and hotspots for potential locations for the initiation of blowouts. However, the potential index does not necessarily represent actual potential for blowout development in the field. The methods used to establish the input data as well as the methods used in this research may not be useful in all applications of the index. Further research is needed in order to make the index applicable in the dunes.

6. References

Aggenbach C, Arens S, Fujita Y, Kooijman A, Neijmeijer T, Nijssen M, Stuyfzand P, Til M, Boxel J, Cammeraat L, 2018. Herstel grijze duinen door reactiveren

kleinschalige dynamiek. KNNV Publishing, OBN223-DK, VBNE, Driebergen.

Breedveld MJ, Stempher W, de Boer ME, 2016. Ontwerpbeheerplan bijzondere

natuurwaarden Meijendel & Berkheide. ARCADIS Nederland bv in opdracht van de provincie Zuid-Holland.

van Engeldorp Gastelaars B. & Rood K. 2010. Beheernota tussen strand en stad. Dunea, Voorburg.

(19)

Hengl T, 2006. Finding the right pixel size. Computers & Geosciences, 32, pp. 1283-1298. Janssen JAM, Bijlsma RJ, Damm T, van Heerden A, 2015. Vegetatie- en habitatkartering

duinen Meijendel 2011 met een toelichting op de habitatkaart van het Natura 2000-gebied Meijendel-Berkheide. Alterra, Van der Goes en Groot, provincie Zuid-Holland.

Schaminée JHJ, Hommel PWFM, Stortelder AHF, Weeda EJ, Westhoff V, 1995-1999.

De

Vegetatie van Nederland 1-5. Uppsala, Opulus press, Leiden.

Schipper PC, 2002. Catalogus vegetatie en vegetatietypologie (tabbladen 4 en 5). In:

Catalogie Bedrijfssturing: Natuur, Bos, Recreatie en landschap (versie 2002). Staatsbosbeheer, afdeling Terreinbeheer. Driebergen.

Appendix A.

(20)

(21)

(22)

(23)

(24)

Appendix B

Data Tables

Table 3. Metadata of the input data used to create the indices.

Dataset name Source Contains Projected

coordinate system

Resolution (cm)

Type

M_30EZ1 AHN3 Digital terrain model RD_New 50 Raster

M_30EZ2 AHN3 Digital terrain model RD_New 50 Raster

M_30GN1 AHN3 Digital terrain model RD_New 50 Raster

M_30GN2 AHN3 Digital terrain model RD_New 50 Raster

M_30DN2 AHN3 Digital terrain model RD_New 50 Raster

N2K_HK_97_Meijendel

_Berkheide Dunea Vegetation data RD_New 1:5000 Polygon

Table 4. Average index values per small-scale blowout in Meijendel. Landscape parameters include

the Slope, Aspect and DTM. The potential index includes landscape parameters and vegetation. Values are rounded to 2 decimals.

Blowout number

Number of grid cells

Plant cover Slope Aspect DTM Vegetation Landscape parameters Potential index 1 2 Vegetated 3 2,5 3,5 3,5 3 3 2 4 Vegetated 2,75 3,75 3 5 3,5 3,75 3 2 Vegetated 3 3,5 3 5 3,5 3,5 4 19 Vegetated 2,84 3,21 2,68 4,79 3,11 3,47 5 6 Vegetated 3 3,17 2,17 4,33 2,83 3,17 6 17 Vegetated 2,82 3,35 3,29 3,88 3,41 3,41 7 12 Active 2,75 3,42 3 3,83 3,25 3,33 8 14 Vegetated 2,71 2,5 3 3 2,92 2,71 9 8 Vegetated 2,5 2,880 2,88 3,5 3 2,88

(25)

10 11 Vegetated 1,81 3,73 2,09 2 2,64 2,18 11 8 Active 2,13 3,75 2 4 2,75 2,75 12 29 Active 2,10 3,28 2,45 3,69 2,69 2,90 13 19 Active 1,84 2,68 3 4 2,47 2,89 14 14 Vegetated 2,14 3,21 2,14 4 2,5 2,86 15 8 Vegetated 2,38 3,38 3,75 3,88 3,5 3,5 16 8 Vegetated 2,13 3,25 2 4,13 2,38 2,88 17 3 Vegetated 1,67 3,33 2 4 2 3 18 3 Vegetated 2 3 2 2 2,33 1,67 19 6 Vegetated 2,33 3 2,5 2 2,67 2,33 20 9 Vegetated 2,11 2 2,78 5 2,11 2,89 21 9 Vegetated 1,89 3,33 2 4 2,44 2,78 22 18 Vegetated 1,83 2 2,89 4,5 2,06 2,72 23 5 Active 1,80 3,2 3 3,6 2,6 3 24 6 Vegetated 2,5 2,83 3 3,33 2,83 2,83 25 15 Active 2,47 3,6 2,07 4,27 2,8 3,2 26 4 Active 1 2,25 3 3,5 1,75 2 27 10 Active 2,80 3 2,5 4,60 2,9 3,4 28 4 Vegetated 2,75 3,75 3 5 3,5 3,75 29 3 Active 2,33 3,67 2 4 2,67 3 30 9 Vegetated 2,44 2,67 2,11 4 2,33 2,67 31 14 Active 2,5 2,93 1,57 3,29 2,21 2,36 32 5 Active 2,2 2,8 1 2 1,6 1,6 33 18 Active 2,17 3,22 1,22 2,89 1,94 2,17 34 3 Active 2,67 3,33 1,33 5 2,33 3 35 1 Active 2 5 5 3 3 4 36 1 Active 1 4 2 3 2 2 37 5 Vegetated 2,8 3,2 2 5 2,8 3,2 38 4 Active 2,25 3,25 2 3,75 2,5 2,75 39 3 Active 2 3,67 3 5 3 3,67

(26)

40 6 Active 2,5 3,5 3 4,67 3,33 3,67 41 9 Active 2,67 3,22 3,11 4,78 3,33 3,67 42 12 Active 2,5 3,58 3 4,42 3,25 3,33 43 9 Vegetated 2,33 3,11 4 5 3,56 3,89 44 7 Active 2,43 3,14 3 5 3 3,57 45 19 Active 2,95 2,74 3 4,05 3,05 3,32 46 26 Active 2,54 3,19 1,88 4,92 2,58 3,15 47 13 Active 2,46 3,38 2 1,85 2,62 2,23 48 10 Active 2,3 3,7 2 2 2,9 2,3 49 2 Active 2 2 2 2 1,5 1,5 50 2 Active 1,5 5 2 2 3 2,5 51 12 Active 2,75 3,08 2 3,67 2,67 2,92 52 13 Active 2,38 3,46 1,92 5 2,69 3,23 53 6 Active 3,33 3,67 1,67 4,83 3,17 3,33 54 4 Active 3 3,5 2 4 3 3

Table 5. Band collection statistics of the Landscape parameters index. Landscape

parameters

Aspect Elevation Slope

Landscape parameters 1 Aspect 0,59071 1 Elevation 0,59977 -0,01443 1 Slope 0,58259 0,00156 0,25272 1

Table 6. Presence of blowouts within randomly selected grid cells with a classification of ‘Very high’.

ObjectID’s corresponds to the field of the feature class ‘Random_VH_class’.

(27)

ObjectID 2016 2016 + 1 gridcell 1975 1975 + 1 gridcell 1938 1938 + 1 gridcell

1 nee ja nee ja nee nee

2 nee ja ja ja nee ja

3 nee ja nee nee ja ja

4 ja ja nee nee ja ja

5 nee nee nee nee nee ja

6 nee ja nee nee nee nee

7 nee nee nee ja nee ja

8 ja ja nee nee nee ja

9 nee ja nee ja ja ja

11 ja ja nee nee ja ja

12 ja ja nee nee nee ja

13 ja ja nee ja nee ja

14 nee ja ja ja ja ja

16 nee nee nee ja nee ja

18 nee nee nee ja ja ja

20 nee nee nee nee ja ja

23 ja ja nee ja nee ja

24 nee nee nee nee ja ja

25 nee nee nee ja nee nee

26 nee nee nee nee nee nee

27 nee ja nee ja nee ja

28 nee nee ja ja nee ja

(28)

30 nee nee ja ja ja ja

Table 7. Sensitivity score of the Aspect’s input raster

Table 8. Classification of the vegetation sub-index map used in the Potential index map of small-scale

blowouts in Meijendel (Schaminée et al. 1995-1999).

VvN-Code Name Plant Community (Dutch Translation) Classification (1-5)

08Bb04 Typho-Phragmitetum (riet associatie) 1 09Ba04 Junco Baltici-Schoenetum Nigricantis (knopbies

associatie) 1

14Aa02 Violo-Corynephoretum (duin-buntgras

associatie) 4

14Bb02a Festuco-Galietum Veri (typicum) (duin-struisgras

associatie) 3

14Ca01 Phleo-Tortuletum Ruraliformis (duinsterretjes

associatie) 5

14Ca01a Phleo-Tortuletum Ruraliformis (typicum)

(duinsterretjes associatie) 5 14Cb01 Taraxaco-Galietum Veri (duin-paardenbloem

associatie) 2

14Cb01c Taraxaco-Galietum Veri (fragarietosum) (duin-paardenbloem associatie) 2 14DG01 Derivaatgemeenschap van Campylopus

introflexus (grijs kronkelsteeltje) (een mos) 2 14RG01 Rompgemeenschap van Carex arenaria

(zandzegge) 3

14RG03 Rompgemeenschap van Dicranum scoparium (gewoon gaffeltandmos) 2

(29)

14RG06 Rompgemeenschap van Agrostis (struisgras) 3 14RG09 Rompgemeenschap van Calamagrostis epigejos

(duinriet)

2

14RG10 Rompgemeenschap van Salix repens (Kruipwilg) 1 14RG11 Rompgemeenschap van Rosa spinosissima

(duinroosje) 1

23RG01 Rompgemeenschap van Elymo-Ammophiletum (Helmvergrassing) 2 31Ba01 Echio-Verbascetum (slangekruid associatie) 1 37Ab1/Ac3 Pruno Crataegetum (Sleedoorn en eenstijlige

meidoorn associatie) / Rhamno-Crataegetum (Wegedoorn en eenstijlige meidoorn associatie)

1

37Ac01 Hippophao-sambucetum (duindoorn en vlier

associatie) 1

37Ac02 Hippophao-Ligustretum (duindoorn en liguster

associatie) 1

37Ac03 Rhamno-Crataegetum (Wegedoorn en

eenstijlige meidoorn associatie) 1 37RG01 Rompgemeenschap van Hippophae

rhamnoides-Sonchus arvensis 1 37RG02 Rompgemeenschap van Hippophae

rhamnoides-Cladonia 1 37RG03 Rompgemeenschap van Hippophae

rhamnoides-Calamagrostis epigejos 1 37RG04 Rompgemeenschap van Ligustrum vulgare 1 39RG04 Rompgemeenschap van Urtica dioica 1 41DG03 Derivaatgemeenschap van Carex

arenaria-Calamagrostis epigejos 1 43Aa02 Fraxino-Ulmetum (Essen-Iepenbos) 1 43Aa03a Crataego-Betuletum Pubescentis

(Meidoorn-Berkenbos)

1

Table 9. SBB-Catalogus classification of the vegetation sub-index map used in the Potential index

map of small-scale blowouts in Meijendel (Schipper, 2002).

(30)

14/a Derivaatgemeenschap van Rubus caesius-Rubus fruticosus (Dauwbraam) 1 14-m Rompgemeenschap van Rosa pimpinellifolia-

[duinroosje1)] 1

23B1a Elymo-Ammophiletum typicum (helm associatie) 2

23B1b Elymo-Ammophiletum festucetosum (helm

associatie) 2

50C Droge kale grond/steen/droog, stuivend zand 5

Table 10. Summary statistics of blowouts in the field Cells per

blowout Slope Aspect Elevation Vegetation Landscape parameter Potential

Mean total 9,19 2,35 3,24 2,49 3,84 2,73 2,94 Mean active 8,43 2,30 3,36 2,35 3,76 2,66 2,89 Mean with Vegetation 9,52 2,42 3,07 2,69 3,95 2,82 3,00 Std 6,29 0,48 0,58 0,73 0,99 0,50 0,59 Max 29 3,33 5 5 5 3,56 4 Min 1 1 2 1 1,85 1,5 1,5

Appendix C

Figures

(31)

Figure 11. Distribution of the amount of calculation grid cells used in the calculation of the average

index scores per blowout.

(32)

Figure 13. Upper-part of the ArcGIS model. The vegetation data of Dunea is used to create the study

area template and vegetation input map. The calculation grid is clipped to the extent of the study area template.

Figure 14. Middle-part of the ArcGIS model. The landscape parameters and sub-indices are created

(33)

Figure 15. Bottom-part of the ArcGIS-model. The landscape parameter index and potential index are

(34)

Appendix D

Additional internship deliverables

Description of Dunea

The internship took place at Dunea. Dunea is a drinking water company, which provides drinking water to inhabitants in the western part of Zuid-Holland. Dunea uses the dunes for surface water infiltration. The dunes filter the water, which can later be pumped up and used as drinking water. Dunea has also an active role in the management of the dunes of Solleveld, Berkheide and Meijendel. Management aims to conserve and protect the dune ecosystem, because the dunes provide various ecosystem services and are a crucial natural defence line against coastal flooding. In order to keep the dunes resilient, management measures are meant to keep the preferred mosaic landscape and to increase biodiversity. As a result of the nitrogen deposition and the decline of the rabbit population, natural succession is accelerating. In order to keep the dune landscape open and maintain the mosaic landscape, natural succession is counteracted with several measures. Hereby, an vision is found between all the different stakeholders that have an interest in the dunes. For example, the dunes are used by recreants, important for the extraction of drinking water, while most of Meijendel is the property of Staatsbosbeheer. Dunea is not only responsible for the supply of fresh drinking water, but also for the management of the dune ecosystem and the ecosystem services it provides.

Description of the activities employed

The internship project is aimed at the development of a suitability map in ArcGIS for the possible testing with small-scale blowouts in Meijendel. Before experiments with

blowouts could start, a model has to be developed which maps the most optimal locations for the initiation of blowouts. The goal of the internship was to develop a framework for this model.

During the first week, I got introduced to Dunea. I experienced the daily working routine, learned how Dunea is structured and what type of activities are employed. I was introduced to the project and the study area of Meijendel. During two field days, my

supervisor showed me the practical side of dune management as well as its history and the aims for the future. I saw several examples of blowouts and learned about the factors that influence blowout development. Afterwards, I was tasked to make a plan of approach inclusive timetable during the first 2 weeks of the internship (Figure 16). Therefore, a small literature study was carried out. When the plan of approach was approved and feedback was given, the research started. When the Corona-crisis started, I discussed with my supervisors about continuing the internship. Because the project was mainly digital, I could do most of

(35)

the work from home via a VPN. The fieldwork could be done autonomously and the weekly meetings with the guidelines provided by the RIVM. Therefore, the project could be

continued. During the individual fieldwork, I visited blowouts with and without vegetation cover in Meijendel. The location was documented and the blowouts were then localized in ArcGIS to test the potential index score on. The development of the index maps was based on the fieldwork and literature study, while the maps were made in ArcGIS.

A personal reflection on what was learned during the Internship.

If I look back at the internship, I have had a positive experience. I experienced the daily working routine, learned about the departments of Dunea and what type of activities are employed. During the first 2-3 weeks, I learned a lot about the responsibilities and activities of Dunea. Unfortunately, due to the occurrence of the Corona-virus, I was sent home to further work on the project. From home, I kept contact with my supervisors on a weekly basis. They gave me advice, direction and goals to aim at during the internship project. If i had questions, I could always email Karin Rood for the digital and technical problems, while I could reach out to Harrie vd Hagen for information about the vegetation and dune

ecosystems. Both my supervisors helped to motivate me and to (re)think about the research and the possible implementation in practice. Also, I was given the time and space to think about the research and to learn how I could outline and shape the research. As a result, I felt that I was free to use my creativity to connect the theories with practice, while maintaining the overview of the research. Even though the Corona-crisis pushed me to do research in a more independent way, I did learn about the importance of cooperation between employees with different skill sets and what can be achieved by working together. The internship made me better understand the dune ecosystem, management aims and involved stakeholders. Also, it became clear to me that solutions for environmental problems do normally not directly work in practice. It is through trial and error that we can learn more, which could eventually result in a solution. However, there are often many attempts needed before the right solution is found. By combining expert knowledge from different scientific fields, an interdisciplinary problem can be easier solved. Over time, the accumulation of new insights results in the movement towards the goals that are aimed at. Furthermore, the fieldwork kept me connected to the ways nature could be managed in practice. I learned how to implement fieldwork in a research and that it is an important aspect for an Earth scientific research. I improved my writing, ArcGIS and analytical skills. Also, I increased my knowledge on vegetation, the dune ecosystems and the ecosystem services the dunes provide. I learned which problems nature managers can encounter and that different stakeholders want different things from the area. Also, I have gained some practical insights in a company that is focussed on the field of Earth Sciences. I learned about past and current management strategies and saw some examples in the field. In conclusion, I can reflect back at the internship in a positive way, but I would have wanted to experience more at working at the office.

(36)

Development of a potential index map for the initiation of small-scale blowouts in Meijendel