- Email: [email protected]
A GIS-based methodology for predicting walking track stability
Journal of Environmental Management 115 (2013) 295e299
Contents lists available at SciVerse ScienceDirect
Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman
A GIS-based methodology for predicting walking track stability Martin Hawes 1, Grant Dixon*, Roger Ling Parks and Wildlife Service, Department of Primary Industries, Parks, Water and Environment, GPO Box 1715, Hobart, Tasmania 7001, Australia
a r t i c l e i n f o
a b s t r a c t
Article history: Received 14 March 2011 Received in revised form 19 November 2012 Accepted 22 November 2012 Available online 11 January 2013
To manage extensive walking track (trail) systems effectively, managers need information about the condition, stability and likely rates of deterioration of tracks. This information may be impractical to obtain from ground inspections, particularly if the track systems of interest encompass hundreds or even thousands of kilometres of tracks. Two trials were undertaken in Tasmania, Australia to assess the practicality of using a GIS-based methodology to predict track ‘types’, types being classes of environmental and track-orientation variables that are associated with characteristic rates of widening and erosion as tracks develop. In the first trial, type values previously measured at 500 18 m long monitoring sites located across a wide range of environments were compared with those predicted for 50e75 m long track segments that included or overlapped the sites. In the second trial, the type values of 300 75 m track segments distributed across five tracks were measured in the field and predicted using a refined version of the methodology. The reliability of the methodology was slightly improved in the second trial, in which 50% of the predictions were accurate and 38% were out by one category. Predictions of the statistical distribution of types were prone to bias due to local conditions on individual tracks, but agreed closely with the measured distribution across the entire data set. The methodology was used to assess track types across the 1700 km track system managed by the Tasmanian Parks and Wildlife Service, as a basis for identifying and prioritising management responses including track stabilisation works. It is likely that with further refinement and with better GIS information, the methodology could reliably predict the stability of individual tracks. Crown Copyright Ó 2012 Published by Elsevier Ltd. All rights reserved.
Keywords: Walking track (trail) stability modelling Extensive track systems Recreation impact GIS Tasmania Australia
1. Introduction 1.1. Track deterioration and the role of information in track management Walking tracks are generally subject to ongoing physical deterioration unless they are artificially stabilised or located on stable or self-maintaining substrates (Cole et al., 1987; Marion and Leung, 2004). The main forms of deterioration are increasing depth (i.e. trenching and gullying), track widening, quagmire development and track braiding or duplication (Cole, 2004; Marion et al., 1993). These developments can adversely affect the recreational experience of walkers (Lynn and Brown, 2003) and are generally contrary to the management objectives of national parks and similar reserves. Impacts are of particular concern when they are likely to be irreversible, as is generally true for soil erosion (Liddle, 1997; Jewell and Hammit, 2000).
* Corresponding author. Tel.: þ61 3 6233 2705; fax: þ61 3 6223 8308. E-mail address: [email protected] (G. Dixon). 1 469 Abels Bay Rd, Abels Bay, Tasmania 7112, Australia.
To effectively manage track systems that include substantial amounts of unimproved track, managers need information about the location and extent of track sections where impacts are likely to be occurring. On the basis of such information, managers can weigh appropriate management responses and assign priorities to responses such as redirecting usage or physically stabilising tracks. 1.2. Techniques for assessing the condition of extensive track systems A range of techniques have been trialled for recording track conditions in the field (Hawes et al., 2006; Hill and Pickering, 2009; Marion and Leung, 2001; Wimpey and Marion, 2010). These techniques require walking the tracks under consideration, and hence may be impractical if resources are limited particularly if the track system of interest extends to hundreds or even thousands of kilometres. Track conditions depend primarily on environmental and siting variables (Coleman, 1981; Leung and Marion, 1996). Given the increasing availability of GIS information, it may be possible to generate such data and (in combination with data on track age and usage levels, where relevant) infer likely track stability, track
0301-4797/$ e see front matter Crown Copyright Ó 2012 Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jenvman.2012.11.027
296
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
condition and rates of deterioration. However, no studies have attempted this to date. Vegetation mapping data have been used in conjunction with information from trampling trials as a basis for mapping the sensitivity of environments to trampling (Gneiser, 2000; Whinam et al., 2003). However, such studies cannot be used to predict the condition or rates of change of established tracks. Chiou et al. (2010) used GIS-derived track slope data to assess expected walking times and energy expenditure on a track network in central Taiwan. Dixon et al. (2004) described a technique for classifying sections of unimproved tracks in the Tasmanian Wilderness World Heritage Area (TWWHA) by ‘type’. Each type is defined in terms of a range of environmental and siting variables and is associated with a characteristic pattern of track deterioration over time. For example, Type 1 sites are defined by low track gradients, good drainage and stable (root rich) substrates, and are generally stable. By contrast, Type 4 tracks are steep, unstable and often associated with water flow, and tend to be subject to substantial widening and track erosion over time, particularly when subject to high usage. 1.3. Study area The Tasmanian Parks and Wildlife Service is responsible for managing over 1700 km of walking tracks on national parks and other reserved land covering approximately 37% of the land area of the state of Tasmania, Australia. The 2.5 million hectare reserve system, which includes the 1.4 million hectare TWWHA, encompasses environments ranging from alpine moorland and cooltemperate rainforest to dry woodland and coastal scrubland. Much of the reserve system is characterised by steep terrain, high rainfall, poor drainage and fine-textured soils (including organic soils), all of which are factors that contribute to track instability. In alpine and moorland environments, track stability is also adversely affected by the absence of woody roots in the substrate. Walking tracks in the reserve system range from wheelchairstandard nature trails to rough unimproved tracks. Many of the tracks have developed from informal walkers’ routes and are hence often poorly sited in terms of gradients, drainage and the resilience of the environments they traverse. Despite extensive investment in track stabilisation in recent decades, more than 80% of the track system remains in an unimproved condition including hundreds of kilometres of boggy and erosion-prone tracks (Parks and Wildlife Service, 2011). 2. Methodology 2.1. Track typing The track typing system described by Dixon et al. (2004) was developed using data obtained over a 16-year period from more than 500 fixed ‘clustered transect’ monitoring sites located across a wide range of environments in the TWWHA. The system classifies track segments into six types using categorically defined environmental and track-siting variables, principally track gradient, drainage and substrate stability (see Fig. 1). Although seasonal and year-to-year fluctuations in rainfall can influence rates of track deterioration over these timeframes, the typing model does not take this into account. Unimproved track segments with higher type number tend to be wider and deeper than those with lower type number when subject to the same levels and duration of trampling impacts. Track segments with ‘Boggy’ drainage were defined as a separate type, and are typically wide but shallow after substantial use unless
Fig. 1. Defining criteria for track ‘type’. Drainage variables are ‘Normal’, ‘Water flow’ and ‘Boggy’. Substrate is classified as either R (indicating the presence of significant quantities of woody roots) or O (‘other’; comprising either organic or mineral soils or mobile fragments, all with a low density of woody roots). From Dixon et al. (2004).
artificially stabilised. A sixth type ‘Non erodible’ was not represented in the current study. 2.2. Overview of the trials In the first trial a GIS record of the 1700 km track network managed by the Parks and Wildlife Service was divided into 50e 75 m segments (see 2.4). The methodology described in the following sections was used to predict the type of each segment, based on GIS data such as vegetation maps and a digital elevation model (DEM). The types previously measured at fixed 18 m monitoring sites (see 2.1) were compared with those predicted for the nearest 50e75 m segment. As the monitoring sites did not coincide exactly with the segments, comparisons were only made at sites that were within 25 m of the nearest segment. As a result, 80 of the 580 monitoring sites were excluded from the analysis. In the second trial a number of modifications were made to the predictive methodology in an attempt to improve its accuracy and reliability. Field trips were undertaken to measure the types of 300 75 m segments evenly spaced along five tracks in the TWWHA. The tracks were selected because of their accessibility and because they represented a broad range of settings and environment types. The measured segments were then virtually generated and the types predicted from GIS data. 2.3. The GIS data set The trials used pre-existing GIS data on walking track locations, terrain elevation, drainage and vegetation type. All data were available, and were modified where necessary, in MapInfo format. The track location data were derived from on-ground GPS surveys. In the first trial the data were pre-existing, tracks being recorded as polylines comprising straight segments typically around 20 m long. In the second trial most of the track-location data
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
were collected during the field inspections, with segment lengths in the range 0e15 m. Terrain elevation and gradient data were available in a digital elevation model (DEM) with a 25 m resolution and vertical accuracy of 5 m. For the second trial, an interpolated version of this DEM was used with an effective resolution of 10 m. Vegetation type data were available at a 1:25,000 scale, with a resolution as fine as 10 m. The data, which were based on aerial surveys ranging from 10 to 2 years old, identified 158 ecological mapping communities (Harris and Kitchener, 2005). In the course of this study some inaccuracies and inconsistencies were noticed in this data, such as the misclassification of some alpine areas known to be free of vegetation. 2.4. Generating virtual track segments In both trials, types were assessed for track segments of length 50e75 m. This was judged to be the shortest length for which meaningful gradient and drainage data could be generated using the 25 m resolution DEM. For the first trial a MapBasic program was written to convert each of the polylines in the MapInfo tracks layer into new polylines with segments in the 50e75 m length range, using the existing nodes as often as possible. In the second trial, 75 m virtual segments centred on the field measurement points were generated from the track log. (These segments generally comprised several sub-segments and were generally not straight.) 2.5. Estimating track and terrain gradients
297
Segments were predicted to be ‘Boggy’ if more than 50% of their length lay within a ‘wetland’ or ‘Potentially boggy’ vegetation type on terrain having gradient less than 5 . ‘Wetland’ had been mapped photogrammetrically based on visual information such as the presence of surface water and vegetation typically associated with swamps. The ‘wetland’ condition was dropped in the second trial as significant inaccuracies were observed in the GIS layer. 2.8. Identifying ‘Woody roots’ segments Vegetation types were predicted to be associated or not associated with the presence of woody roots in the substrate, based on information about the height, density and morphology of the dominant species in each vegetation type (again derived from Harris and Kitchener, 2005, and from the authors’ personal knowledge). Generally speaking, vegetation types with dominant species over 1 m high (or 0.5 m in alpine areas) were considered to be associated with woody roots. 2.9. ‘Dry’ and ‘stable’ vegetation types A small number of vegetation types were known to be exceptionally resilient to trampling impacts (Whinam and Chilcott, 1999) or to be associated with well-drained, drier soils (Harris and Kitchener, 2005) and hence less susceptible to bogginess or water flow. In calculating type, the presence of these vegetation types reduced the type value by one unit (e.g. reclassified Type 3 as Type 2).
The track gradient of each segment was calculated by using the DEM to estimate the difference in elevation h of the two ends of the segment. The gradient was calculated as atan (h/d), where d was the segment length. The terrain gradient corresponding to each track segment was calculated by applying the Line Inspection tool in Vertical Mapper (a MapInfo add-on) to calculate the mean grid-slope value across the segment.
2.10. Calculating types
2.6. Identifying track segments subject to water flow
3.1. First trial
In the first trial, track segments were predicted to be subject to water flow along the track if they were close to the fall line and located more than 100 m downhill from the nearest upslope watershed. The location of watersheds was determined by applying several functions within the StreamBuilder application (also a MapInfo add-on), based on a minimum basin size of 100 ha. In the second trial, flow paths were constructed through the centroids of the 10 m (interpolated) DEM cells using the functions StreamBuilder and Vertical Mapper. The catchment area of each segment was estimated by identifying the flow paths that intersected the segment, counting the centroids that were located on these flow paths above the intersection points, and multiplying the count by the DEM cell area (100 m2). Segments were predicted to be subject to water flow if their catchment area exceeded a threshold that was defined in terms of the track gradient, vegetation type and the track’s orientation relative to the fall line.
In the first trial each of the major components of type (i.e. track gradient category, bogginess and water flow) was correctly predicted for 70e90% of the segments. Track type was correctly predicted for 44% of the segments, and the predicted and measured types were ‘neighbouring’ for a further 44%. (Types can be regarded as ‘neighbouring’ if they differ by one category of one component variable.) Fig. 2 shows the statistical distribution of predicted and measured track types across all 500 track segment/monitoring site pairs. The methodology over-predicts the percentage of type 1 segments and under-predicts the percentages of types 2, 3 and 4, hence underestimating the overall instability of the segments. In 2010 this version of the methodology was used to predict track types across the 1700 km track network managed by the Tasmanian Parks and Wildlife Service (Parks and Wildlife Service, 2011). The resulting data were used as a guide for identifying potentially unstable tracks and track sections, and for identifying and prioritising management responses including track stabilisation works across the track system.
2.7. Identifying ‘Boggy’ segments The 158 vegetation types in the Vegetation layer were manually classified as either ‘Potentially boggy’ or ‘Non boggy’ based on the authors’ personal knowledge of the environments typically associated with these vegetation types, supplemented with information from Harris and Kitchener (2005), which is essentially a handbook to the Vegetation layer (see 2.3).
Track segments were assigned the type ‘Boggy’ if the corresponding drainage was ‘Boggy’. All other segments were assigned types using the criteria illustrated in Fig. 1, with the type adjusted for ‘dry’ or ‘stable’ vegetation as noted in Section 2.9. 3. Results
3.2. Second trial In the second trial each of the major components of type was correctly predicted for 75e85% of the segments. Track type was correctly predicted for 50% of the segments, and the predicted and
298
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
Fig. 2. Predicted and measured distribution of track types for 500 segment/site pairs in the Tasmanian Wilderness World Heritage Area (Trial 1).
measured types were ‘neighbouring’ for a further 38%. Among the latter, 8% of the total involved discrepancies that could be explained by factors noted at the time of the field measurements e e.g. the drainage of some segments was noted to be borderline between two categories. Fig. 3 shows the predicted and measured type distributions for all 300 segments assessed during the second trial. Most of the predicted percentages are accurate within a few points, the largest error occurring in the predicted occurrences of ‘Boggy’ and Type 2 segments. This discrepancy was largely due to a drainage anomaly on the Lake Rhona Track (see below). Predictions of the statistical distribution of types along particular walking tracks tended to be inaccurate owing to local conditions that either could not be predicted from the available GIS data, or were not accounted for by the track-typing model. For example, Fig. 4 shows the predicted and measured type distribution for the 89 segments on the Lake Rhona Track, much of which traverses low-altitude moorland. The model predicted far more ‘Boggy’ segments than were observed because large numbers of flat segments with low vegetation proved to be well drained. This was probably due the presence of coarse gravel subsoils, which could not be predicted from the GIS data.
Fig. 3. Predicted and measured distribution of track types for 300 segments on five tracks in the Tasmanian Wilderness World Heritage Area (Trial 2).
Fig. 4. Predicted and measured distribution of track types for 89 segments on the Lake Rhona Track, Tasmanian Wilderness World Heritage Area (Trial 2).
Fig. 5 shows the predicted and measured type distribution for the 103 segments on the Southern Ranges traverse, in the far south of the TWWHA. Here the methodology tended to over-predict the occurrence of (highly unstable) Type 4 segments, partly because of observed inaccuracies in the GIS vegetation layer. 4. Discussion and conclusion The foregoing findings indicate that the methodology described in this paper can reliably predict the statistical distribution of track types across a large data set (i.e. hundreds of sample segments). It therefore provides a useful basis for predicting the likely stability of extensive track systems. For example, the prediction that 15% of a track system is likely to be Type 4 implies that this percentage is likely to be subject to rapid erosion and widening as tracks mature. This information could prompt managers to undertake an on-ground assessment of track conditions, which in turn could trigger management responses such as realigning or physically stabilising tracks. In its current form, the methodology is an unreliable guide to the type of individual track segments or to the statistical
Fig. 5. Predicted and measured distribution of track types for 103 segments on the Southern Ranges Traverse, Tasmanian Wilderness World Heritage Area (Trial 2).
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
distribution of types across individual tracks. The fact that the component variables of track type are predicted with 75e85% accuracy suggests that this unreliability cannot be attributed to a single component. Rather it results from the accumulated inaccuracy of the components, much of which can be attributed to the limitations of the GIS data as well as to local factors that the methodology does not take into account (such as subsoil composition and track cross-profiles, both of which can influence local drainage). As the methodology is further refined and as more accurate GIS data become available, it should be possible to reliably predict the statistical distribution of track types across individual tracks such as those sampled in the current study. In particular, the use of LIDAR-based elevation data (which is not yet widely available in Tasmania) will facilitate more accurate gradient and catchment estimates, and may allow the determination of track cross-profiles. Other improvements that are likely to enhance the reliability of the methodology include the availability of soil type data and more accurate vegetation data. References Chiou, C.R., Tsai, W.L., Leung, Y.F., 2010. A GIS-dynamic segmentation approach to planning travel routes on forest trail networks in Central Taiwan. Landscape and Urban Planning 97 (4), 221e228. Cole, D.N., 2004. Impacts of hiking and camping on soils and vegetation. In: Buckley, R. (Ed.), Environmental Impact of Tourism. CABI Publishing, Wallingford, UK, pp. 41e60. Cole, D.N., Peterson, M.E., Lucas, R.C., 1987. Managing Wilderness Recreation Use: Common Problems and Potential Solutions. General Technical Report INT-230. USDA Forest Service, Intermountain Research Station, Ogden, UT. Coleman, R., 1981. Footpath erosion in the English Lake District. Applied Geography 1, 121e131. Dixon, G., Hawes, M., McPherson, G., 2004. Monitoring and modelling walking track impacts in the Tasmanian Wilderness World Heritage Area, Australia. Journal of Environmental Management 71 (4), 303e318. Gneiser, C., 2000. Ecological Consequences of Recreation on Subarctic-Alpine Tundra: Experimental Assessment and Predictive Modelling as Planning Tools
299
for Sustainable Visitor Management in Protected Areas. Unpublished Ph.D. dissertation, University of Calgary, Alberta, Canada, 486 pp. Harris, S., Kitchener, A., 2005. From Forest to Fjaeldmark: Descriptions of Tasmania’s Vegetation. Department of Primary Industries, Water and Environment, Printing Authority of Tasmania, Hobart. Hawes, M., Candy, S., Dixon, G., 2006. A method for surveying the condition of extensive walking track systems. Landscape and Urban Planning 78, 275e287. Hill, W., Pickering, C., 2009. Manual for Assessing Walking Tracks in Protected Areas. Technical Report, Sustainable Tourism CRC, Griffith University, Gold Coast, Australia. Jewell, M.C., Hammit, W.E., 2000. Assessing soil erosion on trails: a comparison of techniques. In: Cole, D.N., McCool, S.F., Borrie, W.T., O’Loughlin, J. (Eds.), Proceedings: Wilderness Science in a Time of Change. May 23e27, 1999, Missoula, MT, vol. 5. USDA Forest Service, Rocky Mountain Research Station, Ogden, UT, pp. 133e140. Proceedings RMRS-P-15-Vol-5. Leung, Y.-F., Marion, J.L., 1996. Trail degradation as influenced by environmental factors: a state-of-knowledge review. Journal of Soil and Water Conservation 5, 130e136. Liddle, M., 1997. Recreation Ecology: the Ecological Impact of Outdoor Recreation and Ecotourism. Chapman & Hall, London. Lynn, N.A., Brown, R.D., 2003. Effects of recreational impacts on hiking experiences in natural areas. Landscape and Urban Planning 64, 77e87. Marion, J.L., Leung, Y.-F., 2001. Trail resource impacts and an examination of alternative assessment techniques. Journal of Park and Recreation Administration 19 (3), 17e37. Marion, J.L., Leung, Y.-F., 2004. Environmentally sustainable trail management. In: Buckley, R. (Ed.), Environmental Impact of Tourism. CABI Publishing, Wallingford, UK, pp. 229e244. Marion, J.L., Roggenbuck, J.W., Manning, R.E., 1993. Problems and Practices in Backcountry Recreation Management: a Survey of National Park Service Managers. Natural Resources Report NPS/NRVT/NRR-93/12. USDI National Park Service, Denver, CO. Parks and Wildlife Service, 2011. Walking Track Management Strategy for Tasmania’s National Parks and Reserves 2011e2020. Dept. Primary Industries, Parks, Water & Environment, Hobart. Whinam, J., Chilcott, N., 1999. Impacts of trampling on alpine environments in central Tasmania. Journal of Environmental Management 57, 205e220. Whinam, J., Chilcott, N., Ling, R., Wyatt, P., 2003. A method to calculate environmental sensitivity to walker trampling in the Tasmanian Wilderness World Heritage Area. In: Buckley, R., Pickering, C., Weaver, D.B. (Eds.), Nature-based Tourism, Environment and Land Management. Australian Academy of Science, Canberra, pp. 151e165. Wimpey, J.F., Marion, J.L., 2010. The influence of use, environmental and managerial factors on the width of recreational trails. Journal of Environmental Management 91, 2028e2037.
Contents lists available at SciVerse ScienceDirect
Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman
A GIS-based methodology for predicting walking track stability Martin Hawes 1, Grant Dixon*, Roger Ling Parks and Wildlife Service, Department of Primary Industries, Parks, Water and Environment, GPO Box 1715, Hobart, Tasmania 7001, Australia
a r t i c l e i n f o
a b s t r a c t
Article history: Received 14 March 2011 Received in revised form 19 November 2012 Accepted 22 November 2012 Available online 11 January 2013
To manage extensive walking track (trail) systems effectively, managers need information about the condition, stability and likely rates of deterioration of tracks. This information may be impractical to obtain from ground inspections, particularly if the track systems of interest encompass hundreds or even thousands of kilometres of tracks. Two trials were undertaken in Tasmania, Australia to assess the practicality of using a GIS-based methodology to predict track ‘types’, types being classes of environmental and track-orientation variables that are associated with characteristic rates of widening and erosion as tracks develop. In the first trial, type values previously measured at 500 18 m long monitoring sites located across a wide range of environments were compared with those predicted for 50e75 m long track segments that included or overlapped the sites. In the second trial, the type values of 300 75 m track segments distributed across five tracks were measured in the field and predicted using a refined version of the methodology. The reliability of the methodology was slightly improved in the second trial, in which 50% of the predictions were accurate and 38% were out by one category. Predictions of the statistical distribution of types were prone to bias due to local conditions on individual tracks, but agreed closely with the measured distribution across the entire data set. The methodology was used to assess track types across the 1700 km track system managed by the Tasmanian Parks and Wildlife Service, as a basis for identifying and prioritising management responses including track stabilisation works. It is likely that with further refinement and with better GIS information, the methodology could reliably predict the stability of individual tracks. Crown Copyright Ó 2012 Published by Elsevier Ltd. All rights reserved.
Keywords: Walking track (trail) stability modelling Extensive track systems Recreation impact GIS Tasmania Australia
1. Introduction 1.1. Track deterioration and the role of information in track management Walking tracks are generally subject to ongoing physical deterioration unless they are artificially stabilised or located on stable or self-maintaining substrates (Cole et al., 1987; Marion and Leung, 2004). The main forms of deterioration are increasing depth (i.e. trenching and gullying), track widening, quagmire development and track braiding or duplication (Cole, 2004; Marion et al., 1993). These developments can adversely affect the recreational experience of walkers (Lynn and Brown, 2003) and are generally contrary to the management objectives of national parks and similar reserves. Impacts are of particular concern when they are likely to be irreversible, as is generally true for soil erosion (Liddle, 1997; Jewell and Hammit, 2000).
* Corresponding author. Tel.: þ61 3 6233 2705; fax: þ61 3 6223 8308. E-mail address: [email protected] (G. Dixon). 1 469 Abels Bay Rd, Abels Bay, Tasmania 7112, Australia.
To effectively manage track systems that include substantial amounts of unimproved track, managers need information about the location and extent of track sections where impacts are likely to be occurring. On the basis of such information, managers can weigh appropriate management responses and assign priorities to responses such as redirecting usage or physically stabilising tracks. 1.2. Techniques for assessing the condition of extensive track systems A range of techniques have been trialled for recording track conditions in the field (Hawes et al., 2006; Hill and Pickering, 2009; Marion and Leung, 2001; Wimpey and Marion, 2010). These techniques require walking the tracks under consideration, and hence may be impractical if resources are limited particularly if the track system of interest extends to hundreds or even thousands of kilometres. Track conditions depend primarily on environmental and siting variables (Coleman, 1981; Leung and Marion, 1996). Given the increasing availability of GIS information, it may be possible to generate such data and (in combination with data on track age and usage levels, where relevant) infer likely track stability, track
0301-4797/$ e see front matter Crown Copyright Ó 2012 Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jenvman.2012.11.027
296
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
condition and rates of deterioration. However, no studies have attempted this to date. Vegetation mapping data have been used in conjunction with information from trampling trials as a basis for mapping the sensitivity of environments to trampling (Gneiser, 2000; Whinam et al., 2003). However, such studies cannot be used to predict the condition or rates of change of established tracks. Chiou et al. (2010) used GIS-derived track slope data to assess expected walking times and energy expenditure on a track network in central Taiwan. Dixon et al. (2004) described a technique for classifying sections of unimproved tracks in the Tasmanian Wilderness World Heritage Area (TWWHA) by ‘type’. Each type is defined in terms of a range of environmental and siting variables and is associated with a characteristic pattern of track deterioration over time. For example, Type 1 sites are defined by low track gradients, good drainage and stable (root rich) substrates, and are generally stable. By contrast, Type 4 tracks are steep, unstable and often associated with water flow, and tend to be subject to substantial widening and track erosion over time, particularly when subject to high usage. 1.3. Study area The Tasmanian Parks and Wildlife Service is responsible for managing over 1700 km of walking tracks on national parks and other reserved land covering approximately 37% of the land area of the state of Tasmania, Australia. The 2.5 million hectare reserve system, which includes the 1.4 million hectare TWWHA, encompasses environments ranging from alpine moorland and cooltemperate rainforest to dry woodland and coastal scrubland. Much of the reserve system is characterised by steep terrain, high rainfall, poor drainage and fine-textured soils (including organic soils), all of which are factors that contribute to track instability. In alpine and moorland environments, track stability is also adversely affected by the absence of woody roots in the substrate. Walking tracks in the reserve system range from wheelchairstandard nature trails to rough unimproved tracks. Many of the tracks have developed from informal walkers’ routes and are hence often poorly sited in terms of gradients, drainage and the resilience of the environments they traverse. Despite extensive investment in track stabilisation in recent decades, more than 80% of the track system remains in an unimproved condition including hundreds of kilometres of boggy and erosion-prone tracks (Parks and Wildlife Service, 2011). 2. Methodology 2.1. Track typing The track typing system described by Dixon et al. (2004) was developed using data obtained over a 16-year period from more than 500 fixed ‘clustered transect’ monitoring sites located across a wide range of environments in the TWWHA. The system classifies track segments into six types using categorically defined environmental and track-siting variables, principally track gradient, drainage and substrate stability (see Fig. 1). Although seasonal and year-to-year fluctuations in rainfall can influence rates of track deterioration over these timeframes, the typing model does not take this into account. Unimproved track segments with higher type number tend to be wider and deeper than those with lower type number when subject to the same levels and duration of trampling impacts. Track segments with ‘Boggy’ drainage were defined as a separate type, and are typically wide but shallow after substantial use unless
Fig. 1. Defining criteria for track ‘type’. Drainage variables are ‘Normal’, ‘Water flow’ and ‘Boggy’. Substrate is classified as either R (indicating the presence of significant quantities of woody roots) or O (‘other’; comprising either organic or mineral soils or mobile fragments, all with a low density of woody roots). From Dixon et al. (2004).
artificially stabilised. A sixth type ‘Non erodible’ was not represented in the current study. 2.2. Overview of the trials In the first trial a GIS record of the 1700 km track network managed by the Parks and Wildlife Service was divided into 50e 75 m segments (see 2.4). The methodology described in the following sections was used to predict the type of each segment, based on GIS data such as vegetation maps and a digital elevation model (DEM). The types previously measured at fixed 18 m monitoring sites (see 2.1) were compared with those predicted for the nearest 50e75 m segment. As the monitoring sites did not coincide exactly with the segments, comparisons were only made at sites that were within 25 m of the nearest segment. As a result, 80 of the 580 monitoring sites were excluded from the analysis. In the second trial a number of modifications were made to the predictive methodology in an attempt to improve its accuracy and reliability. Field trips were undertaken to measure the types of 300 75 m segments evenly spaced along five tracks in the TWWHA. The tracks were selected because of their accessibility and because they represented a broad range of settings and environment types. The measured segments were then virtually generated and the types predicted from GIS data. 2.3. The GIS data set The trials used pre-existing GIS data on walking track locations, terrain elevation, drainage and vegetation type. All data were available, and were modified where necessary, in MapInfo format. The track location data were derived from on-ground GPS surveys. In the first trial the data were pre-existing, tracks being recorded as polylines comprising straight segments typically around 20 m long. In the second trial most of the track-location data
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
were collected during the field inspections, with segment lengths in the range 0e15 m. Terrain elevation and gradient data were available in a digital elevation model (DEM) with a 25 m resolution and vertical accuracy of 5 m. For the second trial, an interpolated version of this DEM was used with an effective resolution of 10 m. Vegetation type data were available at a 1:25,000 scale, with a resolution as fine as 10 m. The data, which were based on aerial surveys ranging from 10 to 2 years old, identified 158 ecological mapping communities (Harris and Kitchener, 2005). In the course of this study some inaccuracies and inconsistencies were noticed in this data, such as the misclassification of some alpine areas known to be free of vegetation. 2.4. Generating virtual track segments In both trials, types were assessed for track segments of length 50e75 m. This was judged to be the shortest length for which meaningful gradient and drainage data could be generated using the 25 m resolution DEM. For the first trial a MapBasic program was written to convert each of the polylines in the MapInfo tracks layer into new polylines with segments in the 50e75 m length range, using the existing nodes as often as possible. In the second trial, 75 m virtual segments centred on the field measurement points were generated from the track log. (These segments generally comprised several sub-segments and were generally not straight.) 2.5. Estimating track and terrain gradients
297
Segments were predicted to be ‘Boggy’ if more than 50% of their length lay within a ‘wetland’ or ‘Potentially boggy’ vegetation type on terrain having gradient less than 5 . ‘Wetland’ had been mapped photogrammetrically based on visual information such as the presence of surface water and vegetation typically associated with swamps. The ‘wetland’ condition was dropped in the second trial as significant inaccuracies were observed in the GIS layer. 2.8. Identifying ‘Woody roots’ segments Vegetation types were predicted to be associated or not associated with the presence of woody roots in the substrate, based on information about the height, density and morphology of the dominant species in each vegetation type (again derived from Harris and Kitchener, 2005, and from the authors’ personal knowledge). Generally speaking, vegetation types with dominant species over 1 m high (or 0.5 m in alpine areas) were considered to be associated with woody roots. 2.9. ‘Dry’ and ‘stable’ vegetation types A small number of vegetation types were known to be exceptionally resilient to trampling impacts (Whinam and Chilcott, 1999) or to be associated with well-drained, drier soils (Harris and Kitchener, 2005) and hence less susceptible to bogginess or water flow. In calculating type, the presence of these vegetation types reduced the type value by one unit (e.g. reclassified Type 3 as Type 2).
The track gradient of each segment was calculated by using the DEM to estimate the difference in elevation h of the two ends of the segment. The gradient was calculated as atan (h/d), where d was the segment length. The terrain gradient corresponding to each track segment was calculated by applying the Line Inspection tool in Vertical Mapper (a MapInfo add-on) to calculate the mean grid-slope value across the segment.
2.10. Calculating types
2.6. Identifying track segments subject to water flow
3.1. First trial
In the first trial, track segments were predicted to be subject to water flow along the track if they were close to the fall line and located more than 100 m downhill from the nearest upslope watershed. The location of watersheds was determined by applying several functions within the StreamBuilder application (also a MapInfo add-on), based on a minimum basin size of 100 ha. In the second trial, flow paths were constructed through the centroids of the 10 m (interpolated) DEM cells using the functions StreamBuilder and Vertical Mapper. The catchment area of each segment was estimated by identifying the flow paths that intersected the segment, counting the centroids that were located on these flow paths above the intersection points, and multiplying the count by the DEM cell area (100 m2). Segments were predicted to be subject to water flow if their catchment area exceeded a threshold that was defined in terms of the track gradient, vegetation type and the track’s orientation relative to the fall line.
In the first trial each of the major components of type (i.e. track gradient category, bogginess and water flow) was correctly predicted for 70e90% of the segments. Track type was correctly predicted for 44% of the segments, and the predicted and measured types were ‘neighbouring’ for a further 44%. (Types can be regarded as ‘neighbouring’ if they differ by one category of one component variable.) Fig. 2 shows the statistical distribution of predicted and measured track types across all 500 track segment/monitoring site pairs. The methodology over-predicts the percentage of type 1 segments and under-predicts the percentages of types 2, 3 and 4, hence underestimating the overall instability of the segments. In 2010 this version of the methodology was used to predict track types across the 1700 km track network managed by the Tasmanian Parks and Wildlife Service (Parks and Wildlife Service, 2011). The resulting data were used as a guide for identifying potentially unstable tracks and track sections, and for identifying and prioritising management responses including track stabilisation works across the track system.
2.7. Identifying ‘Boggy’ segments The 158 vegetation types in the Vegetation layer were manually classified as either ‘Potentially boggy’ or ‘Non boggy’ based on the authors’ personal knowledge of the environments typically associated with these vegetation types, supplemented with information from Harris and Kitchener (2005), which is essentially a handbook to the Vegetation layer (see 2.3).
Track segments were assigned the type ‘Boggy’ if the corresponding drainage was ‘Boggy’. All other segments were assigned types using the criteria illustrated in Fig. 1, with the type adjusted for ‘dry’ or ‘stable’ vegetation as noted in Section 2.9. 3. Results
3.2. Second trial In the second trial each of the major components of type was correctly predicted for 75e85% of the segments. Track type was correctly predicted for 50% of the segments, and the predicted and
298
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
Fig. 2. Predicted and measured distribution of track types for 500 segment/site pairs in the Tasmanian Wilderness World Heritage Area (Trial 1).
measured types were ‘neighbouring’ for a further 38%. Among the latter, 8% of the total involved discrepancies that could be explained by factors noted at the time of the field measurements e e.g. the drainage of some segments was noted to be borderline between two categories. Fig. 3 shows the predicted and measured type distributions for all 300 segments assessed during the second trial. Most of the predicted percentages are accurate within a few points, the largest error occurring in the predicted occurrences of ‘Boggy’ and Type 2 segments. This discrepancy was largely due to a drainage anomaly on the Lake Rhona Track (see below). Predictions of the statistical distribution of types along particular walking tracks tended to be inaccurate owing to local conditions that either could not be predicted from the available GIS data, or were not accounted for by the track-typing model. For example, Fig. 4 shows the predicted and measured type distribution for the 89 segments on the Lake Rhona Track, much of which traverses low-altitude moorland. The model predicted far more ‘Boggy’ segments than were observed because large numbers of flat segments with low vegetation proved to be well drained. This was probably due the presence of coarse gravel subsoils, which could not be predicted from the GIS data.
Fig. 3. Predicted and measured distribution of track types for 300 segments on five tracks in the Tasmanian Wilderness World Heritage Area (Trial 2).
Fig. 4. Predicted and measured distribution of track types for 89 segments on the Lake Rhona Track, Tasmanian Wilderness World Heritage Area (Trial 2).
Fig. 5 shows the predicted and measured type distribution for the 103 segments on the Southern Ranges traverse, in the far south of the TWWHA. Here the methodology tended to over-predict the occurrence of (highly unstable) Type 4 segments, partly because of observed inaccuracies in the GIS vegetation layer. 4. Discussion and conclusion The foregoing findings indicate that the methodology described in this paper can reliably predict the statistical distribution of track types across a large data set (i.e. hundreds of sample segments). It therefore provides a useful basis for predicting the likely stability of extensive track systems. For example, the prediction that 15% of a track system is likely to be Type 4 implies that this percentage is likely to be subject to rapid erosion and widening as tracks mature. This information could prompt managers to undertake an on-ground assessment of track conditions, which in turn could trigger management responses such as realigning or physically stabilising tracks. In its current form, the methodology is an unreliable guide to the type of individual track segments or to the statistical
Fig. 5. Predicted and measured distribution of track types for 103 segments on the Southern Ranges Traverse, Tasmanian Wilderness World Heritage Area (Trial 2).
M. Hawes et al. / Journal of Environmental Management 115 (2013) 295e299
distribution of types across individual tracks. The fact that the component variables of track type are predicted with 75e85% accuracy suggests that this unreliability cannot be attributed to a single component. Rather it results from the accumulated inaccuracy of the components, much of which can be attributed to the limitations of the GIS data as well as to local factors that the methodology does not take into account (such as subsoil composition and track cross-profiles, both of which can influence local drainage). As the methodology is further refined and as more accurate GIS data become available, it should be possible to reliably predict the statistical distribution of track types across individual tracks such as those sampled in the current study. In particular, the use of LIDAR-based elevation data (which is not yet widely available in Tasmania) will facilitate more accurate gradient and catchment estimates, and may allow the determination of track cross-profiles. Other improvements that are likely to enhance the reliability of the methodology include the availability of soil type data and more accurate vegetation data. References Chiou, C.R., Tsai, W.L., Leung, Y.F., 2010. A GIS-dynamic segmentation approach to planning travel routes on forest trail networks in Central Taiwan. Landscape and Urban Planning 97 (4), 221e228. Cole, D.N., 2004. Impacts of hiking and camping on soils and vegetation. In: Buckley, R. (Ed.), Environmental Impact of Tourism. CABI Publishing, Wallingford, UK, pp. 41e60. Cole, D.N., Peterson, M.E., Lucas, R.C., 1987. Managing Wilderness Recreation Use: Common Problems and Potential Solutions. General Technical Report INT-230. USDA Forest Service, Intermountain Research Station, Ogden, UT. Coleman, R., 1981. Footpath erosion in the English Lake District. Applied Geography 1, 121e131. Dixon, G., Hawes, M., McPherson, G., 2004. Monitoring and modelling walking track impacts in the Tasmanian Wilderness World Heritage Area, Australia. Journal of Environmental Management 71 (4), 303e318. Gneiser, C., 2000. Ecological Consequences of Recreation on Subarctic-Alpine Tundra: Experimental Assessment and Predictive Modelling as Planning Tools
299
for Sustainable Visitor Management in Protected Areas. Unpublished Ph.D. dissertation, University of Calgary, Alberta, Canada, 486 pp. Harris, S., Kitchener, A., 2005. From Forest to Fjaeldmark: Descriptions of Tasmania’s Vegetation. Department of Primary Industries, Water and Environment, Printing Authority of Tasmania, Hobart. Hawes, M., Candy, S., Dixon, G., 2006. A method for surveying the condition of extensive walking track systems. Landscape and Urban Planning 78, 275e287. Hill, W., Pickering, C., 2009. Manual for Assessing Walking Tracks in Protected Areas. Technical Report, Sustainable Tourism CRC, Griffith University, Gold Coast, Australia. Jewell, M.C., Hammit, W.E., 2000. Assessing soil erosion on trails: a comparison of techniques. In: Cole, D.N., McCool, S.F., Borrie, W.T., O’Loughlin, J. (Eds.), Proceedings: Wilderness Science in a Time of Change. May 23e27, 1999, Missoula, MT, vol. 5. USDA Forest Service, Rocky Mountain Research Station, Ogden, UT, pp. 133e140. Proceedings RMRS-P-15-Vol-5. Leung, Y.-F., Marion, J.L., 1996. Trail degradation as influenced by environmental factors: a state-of-knowledge review. Journal of Soil and Water Conservation 5, 130e136. Liddle, M., 1997. Recreation Ecology: the Ecological Impact of Outdoor Recreation and Ecotourism. Chapman & Hall, London. Lynn, N.A., Brown, R.D., 2003. Effects of recreational impacts on hiking experiences in natural areas. Landscape and Urban Planning 64, 77e87. Marion, J.L., Leung, Y.-F., 2001. Trail resource impacts and an examination of alternative assessment techniques. Journal of Park and Recreation Administration 19 (3), 17e37. Marion, J.L., Leung, Y.-F., 2004. Environmentally sustainable trail management. In: Buckley, R. (Ed.), Environmental Impact of Tourism. CABI Publishing, Wallingford, UK, pp. 229e244. Marion, J.L., Roggenbuck, J.W., Manning, R.E., 1993. Problems and Practices in Backcountry Recreation Management: a Survey of National Park Service Managers. Natural Resources Report NPS/NRVT/NRR-93/12. USDI National Park Service, Denver, CO. Parks and Wildlife Service, 2011. Walking Track Management Strategy for Tasmania’s National Parks and Reserves 2011e2020. Dept. Primary Industries, Parks, Water & Environment, Hobart. Whinam, J., Chilcott, N., 1999. Impacts of trampling on alpine environments in central Tasmania. Journal of Environmental Management 57, 205e220. Whinam, J., Chilcott, N., Ling, R., Wyatt, P., 2003. A method to calculate environmental sensitivity to walker trampling in the Tasmanian Wilderness World Heritage Area. In: Buckley, R., Pickering, C., Weaver, D.B. (Eds.), Nature-based Tourism, Environment and Land Management. Australian Academy of Science, Canberra, pp. 151e165. Wimpey, J.F., Marion, J.L., 2010. The influence of use, environmental and managerial factors on the width of recreational trails. Journal of Environmental Management 91, 2028e2037.