Poster Sessions
Integrated GIS and Airborne Remote Sensing
A Tool for Coastal Conservation and Management in South Wales, UK
Sanjeevi. S
Geo-Environment and Earth Observation Group
Department of Geology, Anna University, Chennai - 600 025. India.
Email: ssanjeevi@annauniv.edu
Abstract
An attempt to understand the problems associated with coastal ecosystems and landscape units for conservation management requires, in the first instance, a well defined/documented map, a resource information system, and change detection techniques. Also, accurate habitat maps are essential pre-requisites to an improved and effective conservation-management programmes. Conventional surveying techniques and in situ measurements clearly have an important role in producing such maps, but they are time-consuming, manpower-intensive and, hence, expensive - particularly in the context of long-term monitoring programmes. For this reason, attention has increasingly been focussed on the use of airborne and satellite remote sensing data, combined with the spatial analytical capabilities of modern Geographical Information System (GIS) technology.
This paper presents an attempt to quantify vegetation and landscape changes in a coastal dune system (Kenfig National Nature Reserve, south Wales, UK) using GIS and airborne remote sensing. The loss of 'slack' habitats associated with the continuing stabilization of this dune system is a major cause for concern. These habitats support a range of plant species, including the rare fen orchid, Liparis loeselii, and other hydrophytes. A decrease in their areal extent implies a reduction in biodiversity. To quantify the overall rate and spatial dimension of these changes, a series of aerial photographs dating from 1962 to 1994 were digitally scanned and analysed in an image processing system. The resultant maps, transferred to a vector-based GIS, were used to derive a transition matrix for the dune system over this period of time. The results indicate that there has been a marked reduction in the total area of bare sand (19.6% of the dune system in 1962, but only 1.48% 1994) and a decline in both the areal extent and the number of dune slacks. Analysis of the habitat maps and hydrological data within the GIS analysis suggests that even dry slacks have the potential for further 'greening' and to support invasive species. In terms of habitat management, however, there is still scope to restore many of the slacks to their original state. It is estimated that at least 24% of the area occupied by partially and moderately vegetated slacks could be rehabilitated.
Introduction
In the general context of coastlines, issues of biodiversity and nature conservation are perhaps most pronounced for dune systems, which form approximately 20% of the area occupied by world's coastal landforms and which are especially rich in species of plants and animals. These features not only make them areas of special research interest, but have also led to some of them being categorized as Protected Areas. The over-stabilization of coastal dunes - often as much a problem as erosion - leads to loss of biodiversity.
An attempt to understand the problems associated with dunal landscape and vegetation succession (and hence loss of biodiversity) for conservation management requires, in the first instance, a well defined/documented map, a resource information system, and change detection techniques. Also, accurate vegetation and habitat maps are essential pre-requisites to an improved understanding of the problems associated with conservation and management programmes. They can form the basis of, and spatial dimension to, resource information systems and change detection techniques. Conventional surveying techniques and in situ measurements clearly have an important role in producing such maps, but they are time-consuming, manpower-intensive and, hence, expensive - particularly in the context of long-term monitoring programmes. For this reason, attention has increasingly been focussed on the use of airborne and satellite remote sensing data, combined with the spatial analytical capabilities of modern Geographical InformationSystem (GIS) technology.
Hartog et. al. (1992), for example, combined aerial photography and GIS to derive transition matrices in a study of the succession of dune vegetation structure resulting from changes in the level of ground-water in the Amsterdam Waterworks dunes. They present ideas as to how these data could be used to analyse spatial patterns on the dune surface and to model landscape succession. Similarly, Davis et al. (1994) provide an account of the use of remotely-sensed images (Landsat-TM data and aerial photographs) and GIS technology to characterise vegetation communities in southwestern California. The authors demonstrate how a vector-based GIS, combined with remotely-sensed data, can be used to produce improved landscape-ecological maps compared to those generated using traditional mapping and manual cartographic procedures.
This paper presents an attempt to measure the path of habitat and vegetation transition in the Kenfig coastal dune system and NNR (National Nature Reserve, south Wales, UK) using remote sensing (sequential aerial photography) and GIS. The loss of 'slack' habitats associated with the continuing stabilization of this dune system has been a major cause for concern among the reserve mangers. These habitats support a range of plant species, including the rare fen orchid, Liparis loeselii, and other hydrophytes. A decrease in their areal extent implies a reduction in biodiversity. Landscape transition ccurring at Kenfig need to be quantified before proceeding to formulate and implement any management plans. Also, identification of the landscape units undergoing change will improve our understanding of the interactions between the geomorphological and biological agents and processes.
Study Area
Kenfig dunes and NNR (Figure 1), with an areal extent of about 650 hectares, contains 575 species of flora of which 550 are native to the UK. Of these, 50% are Welsh flora, 23 of which are nationally scarce. Presently, however, Kenfig is undergoing a reduction of biodiversity due to the over-stabilization of the dune system and as a result : (i) there is acceleration of succession towards a dune-heathland climax status by expansion of vegetated areas at the expense of mobile sand areas; (ii) sand mobility has almost stopped; and (iii) surviving pockets of the original dune habitat are fragmented/isolated (Davies 1995).
The significance of this loss of biodiversity is best highlighted with respect to a number of plant species: First, Kenfig NNR is the British stronghold of the rare and declining fen orchid, Liparis loeselii, supporting more than 95% of the total British population. This species, threatened throughout its European range, is the only higher plant occurring in Britain to be listed as a priority species in Annex - II of the original EC
Figure 1. Map showing the location of the Kenfig National Nature Reserve, UK, chosen for the present study.
|
Habitats and Species Directive. It was also listed as requiring protection by the Bern Convention 1992, and is on Schedule 8 of the Wildlife and Countryside Act, 1992. Regrettably, it is still continuously declining in Britain. An estimated 10,000 plants were present at the site during its best recorded year in 1989, while in 1992 it is believed that fewer than 200 flowering spikes were present (Hurford 1994). At Kenfig, Liparis loeselii occurs in wet dune slacks which are under invasion by Phragmites australis (reed) and Hippophae rhamnoides (sea buckthorn) (Jones 1992, Hurford 1994). Second, Ammophilia arenaria (marram grass) in the foredune areas is declining due to the increase of other biomass. Third, Rumex rupestris (shore dock), and Lomosella subulata (welsh mud wort), both pioneer species, are now almost lost.
Jones (1993) recognised and distinguished 5 principal slack types at Kenfig Burrows. This classification is different from that of Ranwell and Boar (1986) which discriminates between primary slacks (formed at actively prograding sites) and secondary slacks (formed by erosive processes). Jones (1993) comments that the latter category conceals much subtle variation in the dunes and slacks. Consequently his expanded classification identifies 5 slack types as follows:
- Large complexes (>100,000 m2) oriented west-east and bounded to south and north by elongated dune ridges.
- Small narrow parabolic slacks (<8000 m2) occurring immediately landward of the haul road.
- Small parabolic slacks (<13,000 m2) which may be elongated from west-east, although many have no discernable long axis. The characteristic complex of high dunes which generally occur at the eastern end of parabolic slacks may be absent altogether, and type C slacks are generally bound on all sides by dunes up to 20.0 m in height.
- Slacks less than 5000 m2 in size occurring within dune ridges which are invariably aligned along an east-west axis. These 'ridge slacks' are bounded on all sides by high (5-15 m) dunes.
- Extremely common type and characterised by a more or less irregular outline with little differentiation into a long and short axis. Usually surrounded by low dunes ( up to 5.0 m above the slack floor).
Jones and Etherington (1992) mention the replacement of the open, herb-rich vegetation by less diverse areas of coarse grassland and thick bushy Salix repens (creeping willow), mainly due to advent of myxomatosis in 1954. The cumulative effect has been the reversal of the geomorphic process of dune mobility and erosion, in other words, dune stabilization. Vegetation changes at Kenfig Burrows have been especially rapid since the cessation of an effective grazing regime during the mid 1970's. Bushy growth forms of Salix repens (creeping willow) now dominate the vegetation of many dune slacks, and there has been a progressive reduction in the number of slacks supporting population of rare and uncommon hydrophytes including Carex serotina, Baldellia ranunculoides (lesser water-plantain) and Anagallis tenella (bog pimpernel).
Quantifying Landscape/Habitat Succession
The methodology (Figure 2) adopted in this study is similar to that used by Hartog et al. (1992). Thus, the aerial photographs (Table 1) were visually interpreted based on the keys developed by Hartog et al. (1992) and the field identification keys developed by Hurford (1997)(pers.commn.). Clearly, there are limitations to the ability to identify and map the many types of vegetation species and communities in Kenfig NNR using panchromatic or colour aerial photographs, the most important ones being the spatial scale and spectral limitations.
As mentioned earlier, the focus of studies of ecological diversity and nature conservation has shifted from species to habitats, with this change in emphasis accelerating in recent years. This general observation is also true for Kenfig NNR, where loss of habitat has been the major cause for concern among the nature reserve managers. Consequently the aim here has been to map the major landscape/habitat types based on a visual assessment of vegetation density and structure in aerial photographs.
Figure 2. Schematic representation of the methodology adopted in this study.
|
To quantify the transition that had occurred at Kenfig NNR between 1962 and 1994, the changes in the areal extent of various landscape units (or broad habitat categories) have been interpreted from aerial photographs acquired during 1962, 1971,
1991 and 1994 (Table 1). Taking into account the nature of the study, the scale of photographs used, and the structure of the vegetation at Kenfig NNR, a simplified classification of habitats was produced (Table 2). The habitat maps generated from the aerial photographs were digitised using a semi-automatic approach involving scanning, conversion to grids and then semi-automatic digitising. The vector coverages generated in this way were subsequently geo-referenced, co-registered and rectified with reference to control points derived from UK Ordnance Survey maps of 1:10000 scale. Each parcel of a given habitat is thus a closed polygon in a coverage. The vector coverage for a given year comprises a number of different layers, where each layer contains polygons representing the parcels of a given habitat. The attributes of each polygon in any given coverage include its area, perimeter, habitat type and hydro-type.
Table 1. Details of aerial photographs used in the study.
| Year |
Type |
Scale |
| 1962 |
B/W Panchromatic |
1:10,000 |
| 1971 |
B/W Panchromatic |
1:5,000 |
| 1991 |
Colour |
1:10,000 |
| 1994 |
Colour |
1:25,000 |
Table 2. Habitat classification system and their equivalent NVC types
(Dargie 1995) used for aerial photo interpretation
| Habitat Classification (present study) |
Corresponding NVC types |
| Open sand (OS) |
Mobile dunes SD4-SD6 |
| Partially vegetated dunes (PVD) |
Semi fixed dunes SD7 |
Moderately vegetated dunes (MVD) |
Dune grassland SD8, SD9 |
| Densely vegetated dunes (DVD) |
|
Partially vegetated slacks (PVS)Moderately
vegetated slacks (MVS)Densely vegetated
slacks (DVS) |
Dune slacks SD14-SD15 |
| Woodland/Scrub (W) |
Scrub and Woodland SD18 and W |
Vector GIS and Dune-Landscape Transition Analysis
Overlays were produced of the individual habitat maps for each year considered in this study. The resultant change vectors present a vivid picture describing the type of successional changes that took place between 1962 and 1994. The changes are reflected not only in terms of an increase or decrease in the areal extent of individual habitats, but also in terms of partial or complete change of each habitat parcel into another stage in the successional sequence. To understand the nature of succession, each habitat was compiled as a separate layer (i.e., vector coverage) and overlays were constructed between corresponding layers from different years. A total of 156 vector coverages, representing all the landscape types for the four years in consideration, were generated.
Figure 3 is an example of a vector coverage (1962) containing the eight habitat/ landscape types. Computation of areal extent of each of the habitats/landscape for each of the years studied confirm that there has been a general 'greening' of the dune system since 1962. In other words, there has not only been a substantial decrease in the areal extent of open sand/mobile dune, but also a reduction in the generation of dune slacks and an increase in the biomass on existing slacks. The most likely explanation of this is that there has been a decrease in the mobility of the dunes, as a consequence of over-stabilisation.
A quantification of the change brought about in the dunal system over the years, described qualitatively above, is shown in Figure 4. This figure shows that about 20% (approximately 82 hectares) of the study area within the NNR was covered by mobile sand during the early sixties, but that this had reduced to a mere 1.5% (6.5 hectares) by 1994. Similarly, the areal extent of the partially-vegetated sand/dunes (PVD) reduced from 26% in 1962 to 2.7% by 1994. On the other hand, the area of moderately vegetated dunes (MVD) increased from 18.7% to 56.8% over the same period of time. In the case of the dune slacks, the partially vegetated slacks have reduced in extent from ~12% to 1.6%: this is a result of their transition into more densely vegetated slacks (0.24% in 1962 to 5.65% in 1994). The slacks also have a tendency to transform into woodlands, the area of which has increased eight-fold.
Figure 3. Example of a digital vector coverage of the landscape / habitat map of Kenfig NNR derived from aerial photographs acquired in 1962. (The entire area of the nature reserve has not been considered).
|
Table 3 provides an overview of the changes in the spatial structure of vegetation in the dune and slack habitats. A more detailed analysis of the successional changes for specific habitats, inferred from Table 3, is as follows. In 1994, the open sand (OS) cover had reduced to 5.36% of its areal extent in 1962. This was due to growth of vegetation on the dunes. Thus, 8.8% of the open sand habitat had been transformed into partially vegetated dune habitat, 64.5 % transformed into moderately vegetated dune, and 0.6 % to densely vegetated dune by 1994. While this general trend indicates the greening and stabilisation of the dunes, there is also evidence of some erosion/dune mobility which has resulted in the conversion of 1.17% of the partially vegetated dunes into open sand. Analyses of the change vectors between 1971 and 1991 show that the geomorphological process of erosion/dune mobility seems to have ceased sometime after 1971.
Table 3.
Transition matrix, generated using a GIS, showing habitat changes for part of
Kenfig NNR between 1962 and 1994 (see Table 2 for explanation of the habitat
codes; elements of the matrix indicate the percentage of areal extent since
1962).
| 1962 |
1994 |
| OS |
PVD |
MVD |
DVD |
PVS |
MVS |
DVS |
WL/S |
| |
Area(ha) |
6.17 |
11.20 |
189.80 |
49.50 |
6.76 |
92.09 |
23.55 |
33.07 |
| OS |
76.64 |
5.36 |
8.80 |
64.50 |
0.62 |
1.58 |
14.25 |
1.18 |
2.63 |
| PVD |
104.48 |
1.17 |
1.98 |
73.50 |
0.40 |
0.89 |
15.02 |
0.15 |
7.41 |
| MVD |
75.20 |
0.90 |
1.15 |
55.10 |
35.20 |
1.65 |
6.10 |
- |
0.20 |
| DVD |
3.70 |
4.10 |
- |
- |
93.80 |
- |
- |
- |
2.10 |
| PVS |
47.63 |
0.001 |
1.21 |
24.70 |
2.51 |
0.95 |
41.60 |
23.20 |
5.98 |
| MVS |
83.03 |
- |
0.52 |
- |
- |
- |
29.90 |
55.60 |
15.00 |
| DVS |
0.98 |
- |
- |
- |
- |
- |
- |
50.86 |
49.20 |
| WL/S |
3.98 |
- |
- |
- |
- |
- |
- |
- |
100.00 |
A similar analysis for the dune slacks shows that significant erosional/dune mobility processes were in operation during 1962 and for some years later, resulting in certain areas of the slacks being converted to dunal habitats. By 1994, however, 24.7% of the partially-vegetated slacks (PVS) had been converted into moderately vegetated slacks (MVS). This conversion was principally along the periphery of the slacks. Comparison of vegetation growth on the dunes and in the slacks indicates that the dunes were greening at a faster rate than the slacks. In the dunal habitat, the areal extent of PVD conversion to MVD is higher than any other type of habitat conversion. The rate of succession towards densely vegetated dunes (DVD) and scrubland is not as effective as the succession towards moderately vegetated dunes (MVD). Field observation showed that MVD habitats could easily be restored to OS status by mowing or removing the vegetation cover. Restoration of DVD into OS is more difficult as it involves intensive, repeated mowing and complete removal of seabuck thorn (including the roots) by manual cutting and injection of weedicide into the roots.
The interest of the reserve managers / ecologists of the NNR lies more with the slacks (as they are the habitats that support the rare species such as Liparis loeselii) that are being invaded by Hippophae rhamnoides and Salix repens. In this context, 71.5 % of the area under PVS in 1962 had been transformed to MVS, DVS (densely vegetated slacks) or WL/S (woodland / scrub) by 1994. The areas of PVS that existed in 1994 were the ones created due to erosion during or soon after 1962. Similarly, 70.6% of MVS has been transformed to DVS and WL/S, while only 29.9% has remained as MVS. This study has brought to light the fact that there was some kind of erosional and dune mobility process going on after 1962, as a small percentage of vegetated dunes and slacks have been converted into open sand or less vegetated landscapes.
Figure 4. Transition diagram showing changes in the type and the areal
extent of landscape/ habitat units from 1962 to 1994.
|
GIS analysis using slack hydrology as an attribute
GIS analysis using overlays (vector coverages) of the habitat maps of different years has helped to quantify the direction, amount and rate of transition/succession as described above. As the hydrological regime plays an important role in determining vegetation growth and the subsequent successional trend in the dunal system, an attempt was made to demarcate slack areas that are undergoing rapid transition due to hydrological factors. Based on their hydrological status, the dune slacks at Kenfig NNR were classified into one of four types by Jones (1993) as:
Type 1 slacks are typified by very shallow winter-flooding. During periods of flooding the water table displays a subdued response even to prolonged periods of heavy rainfall,
Type 2 slacks are similar to the above, but flood to a greater depth. The depth of flooding rarely exceeds 30cm, however, and for most dip-wells in the group lies between 10cm and 25cm;
Type 3 are characterised by deep winter-flooding; and
Type 4 slacks are characterised by extremely deep winter-flooding.
A vector (point) coverage consisting of 192 dip-well locations (Jones 1993) - together with their corresponding hydrological characteristics (hydro-type/attribute) - was combined with the habitat change vector maps. This indicated that: (i) slack types 2 and 3 are the most vulnerable to transition to denser vegetation types; (ii) faster increase in vegetation cover occurred in areas originally characterised by slack types 2 and 3; and (iii) after becoming moderately vegetated, hydrological factors have little impact on the transition into dense slack and woodland.
Limitations of this study
The most difficult (and perhaps error-prone) part of studies such as this is rectification of the aerial photographs to some reference map projection. Although the utmost care was taken in geo-referencing the photographs and rectifying them, the scale of the photographs was different for each date, so that distortion and subsequent error in matching may have resulted in a small percentage of error creeping in the geometry of the vectors and eventually in the change statistics. A second consideration is that colour photographs, had they been available for all the years under consideration, would have provided more reliable information on vegetation type and density.
Conclusions
The general conclusion that can be drawn from the integration of Geographic Information System (GIS) and aerial photographs and used in this study is that, while the processes of erosion and dune mobility were active at Kenfig NNR at the start of the period considered (1962-1994), the dunes have become increasingly stabilised and there has been a cessation of dune-slack generation. The transition matrix, that was a result of integration of GIS and aerial photography, helped to realise that the greening of the dunes has been slower than that of the slacks, and that at least 25% of the slacks could be restored to the conditions that support the survival of rare and pioneer species given appropriate management practices.
References
- Davies, A. J. D., 1995, Greening of Kenfig Burrows. BSc Dissertation. Swansea, University of Wales Swansea.
- Davis, F. W., Stine, P. A., and Stoms, D. M., 1994, Distribution and conservation status of coastal sage scrub in southern California. Journal of Vegetation Science, 5, 743-756.
- Hartog, M., van der Meulen, F., and Jongejans, J., 1992, Dune landscape development and changing groundwater regime: Quantitative landscape succession with help of a GIS. 119-127. In Coastal dunes. Geomorphology, Ecology and Management for Conservation. Balkema. Rotterdam
- Hurford, C., 1994, A survey to monitor the fen orchid Liparis loeselii in dune slack ND6 at Kenfig NNR, October 1992. A report by the Countryside Council for Wales, UK.
- Jones, P.S., and Etherington, J.R., 1989, Ecological and physiological studies of sand dune slack vegetation, Kenfig Pool and Dunes Local Nature Reserve, Mid-Glamorgan, Wales, U.K. in Perspectives in coastal dune management ed. van der Muelen. F., Jungeris. P.D., and Visser. J.H. SPB Academic Publishing, The Hague. 297-303.
- Jones, P.S., and Etherington, J. R., 1992, Autoecological studies non the rare orchid Liparis loeselii and their application to the management of dune slack ecosystems in South Wales. In Coastal Dunes, edited by R. W. G. Carter, T. G. F. Curtis and M. J. Sheehy-Skeffington. (Rotterdam: Balkema), 299-312.
- Jones, P.S., 1993, Ecological and hydrological studies of dune slack vegetation at Kenfig NNR, Mid-Glamorgan. Ph.D. Thesis. Cardiff, University College of Wales.
