I hinted at the potential of GIS in karst geomorphological research in my undergraduate thesis, where I performed spatial analysis of karst terrain to determine the proportion of hills and depressions in the area, and classify it accordingly. This opened up the door to three-dimensional analysis of karst terrain. Since then, I've used optimal kriging interpolation to more precisely model the karst terrain on a more detailed level. 

General and specific geomorphometric techniques have been applied, where the entire landscape and individual features are measured separately, their patterns determined, and controls on the landscape identified based on a number of parameters.  Different methods of classifying the terrain have already been conceptualized, including aspect as a measure of surface roughness.  Other means of determining roughness include looking at the overall surface area of the terrain, as well as conducting neighbourhood analysis of the range and standard deviation of elevation.  Slope patterns have also been determined, and reveal the influence geology has on the landscape, whether structural controls or lithology is responsible for a particular landscape pattern. The orientations, areas and densities, among other things, of karst features can also be calculated easily.  I've also  determined the highest and lowest points of features in the terrain (the summits of hills and the sinks of depressions) in order to look at the spatial distribution of these features in cockpit karst terrain vis a vis non-cockpit karst terrain. Nearest-neighbour and spatial autocorrelation analyses have been carried out to look at the point patterns of the karst features within the study area.  Additionally, the sink and summit points can also be individually interpolated to create sink and summit surfaces. The difference between these two surfaces can be determined using map algebra in order to create a local relief surface for each area.

Specific morphometric calculations, measuring individual features in the landscape rather than the entire surface, were also carried out. In addition to looking at the patterns and distributions of sinks and summits, I also looked at the hills present in the landscape. While cockpit karst landscapes are famous for their enclosed, star-shaped depressions, these depressions were difficult to delimit using many different methods, including looking at significant breaks in slope, calculating the area that was less than the mean of a sink-normalized surface, and using edge detection filters on the processed DEM. Each time, the hills were isolated; the depressions being more connected than enclosed than the hills. The degree of isolation of hills and depressions relative to each other could be determined using proximity indices, commonly applied in landscape ecological studies. It would appear that the result of these analyses show that hills are far more isolated than the "enclosed" depressions, which were far more interconnected, a fact seemingly verified in the field. It seems then that the very nature of the "Cockpit" Country is in fact an illusion; the landscape is clearly not dominated by enclosed cockpit depressions, but instead by isolated hills.

In the end, the hills were digitized individually using a combination of imagery, TINs, aspect grids, and a pair of shaded relief grids which functioned as a sort of pseudo-stereopair when delimiting the hills. From the resulting polygons, hill morphometry could be calculated, including a new concept aimed at analyzing hill complexity by comparing the hill's 3D complexity with its planar complexity. Looking at hills is certainly useful when comparing cockpit karst landscapes with tower karst landscapes where hills can be objectively compared in both terrain.

However, by considering surface water flow and the slope of the terrain, areas where water would accummulate (depressions) and areas where water would runoff quickly (hills) could be identified. For the purposes of this research, an impermeable DEM surface was assumed; in reality, karst limestone conditions do not permit surface runoff except during periods of heavy rainfall. As such, both hills and depressions could be identified from the DEM. Morphometric analysis could be performed on individual grid patches to glean further information from the landforms.  The advantage of this method over manual digitizing of hills or identification of features greater than or less than the mean elevation of a sink-normalized surface, where the bases of hills and the rims of depressions all have the same height, is that this method is far more objective and less ambiguous; hills and depressions are determined by the slope and catchment area characteristics of the DEM. Hills have steep slopes and low catchment areas, while depressions have gentler slopes and larger areas draining into them.  These techniques are more commomly applied to fluvial terrain, but can be used as proxies to delimit karst landforms in this case.  Furthermore, this method confirmed existing understading of cockpit karst terrain, that the depressions themselves are isolated, surrounded by residual hills, which dominate the landscape. Small 'saddles' between hills which act as corridors between depressions  were in fact part of one hill.In the polje/tower karst areas at Nassau Valley and the Queen of Spain's Valley, hills there are far more isolated, with large interconnected depressions separating them.

Virtual reality simulations are also accomplished using the GIS. 3D scenes, complete with a draped aerial photograph or satellite image, along with interactive roaming, have been included in my GIS effort. Samples are shown on the right.

1-metre resolution IKONOS satellite imagery was obtained courtesy of Spatial Innovision Ltd. in Kingston. I managed to georeference these images myself and incorporate them into my analysis. The images were draped over the TINs created of each area. Only in the cases of Nassau Valley, Queen of Spain's Valley and Quickstep did the images cover the entire area; images of the rest covered roughly 75% of the area. Unfortunately cloud cover, especially at Windsor, hid some of the features on the ground, but the morphology of the terrain was still evident after draping on to the TIN surface. 

The ruggedness of the cockpit karst terrain is very evident, and stands in marked contrast to the poljes at Nassau Valley and Queen of Spain's Valley. Nevertheless, the poljes are karst features, worthy of inclusion in the study of the diversity of karst landforms in and around the Cockpit Country, which is the essence of my research.

•  IBM Thinkpad: Windows ME, Pentium 3 Processor, 64MB RAM, 10GB Hard Drive 
•   Sony Vaio: Windows XP, Pentium 4 Processor, 512MB RAM, 40GB Hard Drive 

•  ArcView GIS 3.2 
• ArcGIS 8.2 
• PC/ArcINFO 

• ESRI Spatial Analyst (for both ArcView GIS and ArcGIS packages) 
• 3D Analyst  (for both ArcView GIS and ArcGIS packages) 
• Geostatistical Analyst (ArcGIS) 
• custom  Avenue scripts and extensions (available from the ESRI website 

• First URISA Caribbean GIS Conference, Montego Bay, Jamaica, 2001  
• Fourth British Caribbean Geography Conference, San Ignacio, Belize, 2002 
• International Conference on Karst Hydrology and Ecosystems, Bowling Green, Kentucky, USA, 2003 
• British Geomorphological Research Group Annual Conference, Oxford, United Kingdom, 2003 

3D Virtual Tour of the Windsor Study Area
Created by draping a TIN of the area with a georeferenced high-resolution IKONOS image of the same area

     
Static Views of the Surface of Barbecue Bottom, showing a TIN with aerial photographs draped over the surface. The glade is prominently shown here. 

---

     
Barbecue Bottom, Trelawny                                    Elderslie, St Elizabeth 
     
            Nassau Valley, St Elizabeth                      Queen of Spain's Valley, Trelawny/St James
     
Quickstep, St Elizabeth                                                Windsor, Trelawny 

---

GIS analysis involved the creation of continuous surfaces from topographic maps and GPS spot heights.  Imagery may be draped on the surface to make the terrain more photo-realistic. The animation above shows the progression from vector lines and points to a continuous surface, represented first by a wire mesh model and a TIN. Finally, an IKONOS image of a part of Barbecue Bottom is draped on the surface.

---

The Location of the Cockpit Country, Jamaica. 

More information on the Cockpit Country.

Elevation data may be interpolated to create a DEM of the study areas. Slope and Aspect grids can be derived from this DEM.

The top (green) profile shows the actual profile of the Barbecue Bottom area as determined from the DEM. The lower (brown) profile shows the topographically normalized profile along the same transect. This was accomplished by subtracting the interpolated sink surface from the original DEM.

Terrain generally exhibit scaling characteristics, showing increasing vertical relief with increasing horizontal scale. In the karst areas under study, this pattern is observed. However, there is a major scaling break at around 100m horizontal scale, where the slope of the graph becomes more gentle. 100m reflects the average spacing between hill summit and depression sinks (as determined using nearest-neighbour analysis), and it can be assumed that this scaling break reflects the scale of these features. In all cases, the horizontal scale of the landscape show pretty much the same pattern, though the vertical relief is different; the cockpit karst areas show greater relief development, while the non-cockpit karst areas show less relief development. In all cases, these lines can be described by a Power Law y = cx^m , where y is vertical relief, x is horizontal scale, m describes the slope of the line and c describes the amplitude. Both the slope and amplitude can be used to describe the terrain. The characteristics of the terrain before and after the scaling break were analysed and compared.

This was carried out using ArcView 3.2 and the Spatial Analyst extension, using the Neighbourhood Analysis function. However, the graph displayed, created in Microsoft Excel, is the only thing on this page not created entirely within ArcView. 

Based on the scaling characteristics of the landscapes, the cockpit karst areas of each study area can be identified based on their vertical relief at the 100m horizontal scale. Cockpit karst areas are found in all six study areas, though to varying extents. In Barbecue Bottom, cockpit karst dominates the landscape, though it does not occupy 100% of the terrain; elongated glades and depressions are also present and are not considered cockpit karst as the vertical relief within these areas is too gentle to be considered cockpit karst. In the Nassau Valley, cockpit karst occurs peripherally, surrounding the central polje region. Both maps are overlain with a 40% transparency on a geological map, showing the correlation between the occurence of cockpit karst and the White Limestone. Though this cannot be seen clearly on this website, it is clear that alluvium, shown above in pink, does not support cockpit karst.

Sinks and summits may be identified on the surface using the DEM. The patterns of sinks and summits in the terrain can be analysed. Also, their combined pattern can also be examined. Individual hills and polygons may also be identified. Their spatial distribution can be analysed by conducting proximity analyses commonly used in landscape ecological applications. The planar shape and 3-Dimensional form of delimited features may also be examined. This leads to the development of an index of measurement that compares surface complexity with planar complexity, which can be used to classify hills. 
 
 

Hills and depressions could be identified by determining the Compound Topographic Index, or the Topographic Wetness Index, which looks at the catchment area and the slope of the terrain to determine areas of high and low saturation. For the purposes of this research, areas with low wetness values represented hills and areas of high wetness values represent depressions.  Attributes of each landform patch could then be determined.  This method is far more objective than manual digitizing of hills and has the added benefit of adequately delimiting depressions as well.

     
Hill and Depression Patches overlaid on IKONOS image

---

Idealized Karst Landscape Animation

Ideal karst landscape, with uniform distribution of hills and depressions in a hexagonal pattern in the landscape. The animation is showing the terrain at different levels of relief, ranging from 100m to 10m. The landscape is essentially assumed to be a homogeneous, horizontally bedded limestone mass, dissected by 3 major fracture traces equally spaced apart from each other, each aligned 120^o from each other.

A new way to look at the morphology of individual hills in the landscape would be to compare its 3D complexity (surface area/flat area - SA:FA) with its planar complexity (the shape index - perimeter of the hill planar outline, or footprint, vs the perimeter of a perfect circle of the same area as the hill being examined). Many researchers have studied both the 2D and 3D complexity of features, but have never combined the two to create a single morphometric index. Simple hills, like the one above, have a high SAFASI, with very low planar complexity (in this case, 1, since the shape index of the hill footprint is equal to 1 since the footprint is a perfect circle). Since conical hills are a characteristic feature of cockpit karst landscapes, this measure is a very useful way of classifying such terrain based on its features.