Brian Lees

Advances in digital terrain analysis

Advances in Digital Terrain Analysis: The TADTM Initiative.- Digital Representation for Terrain Analysis.- Quantification of Terrain Processes.- Models of Topography.- Multi-Scale Digital Terrain Modelling and Analysis.- A Seamless and Adaptive LOD Model of the Global Terrain Based on the QTM.- Morphological Terrain Analysis.- Landform Classification of the Loess Plateau Based on Slope Spectrum from Grid DEMs.- Segmentation-based Terrain Classification.- Terrain Segmentation and Classification using SRTM Data.- Modelling Terrain Complexity.- DEM-based Analysis of Local Relief.- Re-Scaling Lower Resolution Slope by Histogram Matching.- Hydrological Terrain Analysis.- Water in the Landscape: A Review of Contemporary Flow Routing Algorithms.- An Integrated Raster-TIN Surface Flow Algorithm.- DEM-based Modelling and Simulation of Modern Landform Evolution of Loess.- Uncertainty in Terrain Analysis.- Assessing Uncertainties in Derived Slope and Aspect from a Grid DEM.- Accuracy Assessment of DEM Slope Algorithms Related to Spatial Autocorrelation of DEM Errors.- Modelling Slope Field Uncertainty Derived From DEM in the Loess Plateau.- The Impact of Neighbourhood Size on Terrain Derivatives and Digital Soil Mapping.- The Impact of DEM Error on Predictive Vegetation Mapping.- Applications of Terrain Analysis.- Global Lineaments: Application of Digital Terrain Modelling.- Modelling Channelling and Deflection of Wind by Topography.- Spatial Correlation of Topographic Attributes in Loess Plateau.- Terrain-based Revision of an Air Temperature Model in Mountain Areas.- National Mapping of Landform Elements in Support of S-Map, A New Zealand Soils Database.

Combining Non-Parametric Models for Multisource Predictive Forest Mapping

Most models of forest type for predictive mapping cannot produce estimates of confidence in the prediction of individual pixels, even where they provide good overall accuracy. A new strategy that combines several models based on different principles not only provides estimates of prediction confidence, but also improves the mapping accuracy. In this study, the theoretical foundation of Artificial Neural Networks, Decision Trees, and Dempster-Shafer’s Evidence Theory are briefly reviewed, compared, and applied to a common data set. Two ways for integrating the results of the three models were then evaluated. One method was to separately harden the probability results of the three models, then combine them to make a single classification. In the second method, the probabilities of the three models for each pixel were simply averaged, then hardened to a single classification. Deferring the hardening to the final stage produced the best results. The 3 percent increase in overall accuracy for the second approach compared with the best individual model is encouraging. More importantly, estimates of prediction confidence were derived, based on a comparison between a combined model and the three models, something that is impossible using a single model.

An Improved Liberal Cloud-Mask for Addressing Snow/Cloud Confusion with MODIS

Abstract The utility of the daily MODIS snow products depends on the ability of the MODIS cloud-masking algorithm to differentiate between snow- and cloud-cover. Although few studies have explored the issue, snow/cloud confusion is a key issue limiting the accuracy of the MODIS snow products. Recent studies from the Southern Hemisphere suggested that snow/cloud confusion limited the utility of the MODIS snow products there. In this study, MODIS snow/cloud confusion over Australia was investigated using an improved liberal cloud-mask in conjunction with a snow-detection algorithm. The performance of the proposed cloud-mask was assessed using high-resolution ASTER imagery and in situ observations. Results indicated that the improved liberal cloud-masking algorithm reduced snow/ cloud confusion, successfully identifying snow-covered pixels that were previously identified as cloudy. The analysis further suggested that scale-related differences in imagery used in the standard MODIS cloud-masking workflow might be the source of some snow/cloud confusion previously reported.

Just another node...

Our academic community is a social network. Haggett and Chorley (1970) would have argued that it decomposes into a hierarchy, but Mark Gahegan tells me that social networks are not true hierarchies: they are multiply-connected and thus graphs. Some have suggested that social networks are hierarchies of graphs. In an upcoming paper, Mark shows some examples based on IJGIS author, keywords, and institution data. The topologies are based on geographic location, ideas and author connectedness. These provide an interesting way of looking at our activities. One can argue that the role of a university is to be one of these geographic locations at which a member of each disciplinary network can reside, connecting the students and other scholars at that location to the wider group. But each of us also exists in an aspatial network of ideas and connections with like-minded colleagues. It is the role of conferences and peer-reviewed journals to facilitate those external connections. Milgram’s (1967) view is that the importance of a node to the functioning of a social network lies in the number of connections it makes. The importance of these connections is in linking parts of the network together. Usually, this means linking the less well connected parts of the structure to the better connected parts. Increasingly, our university employers are keen to evaluate us on our status within the discipline. Sadly, status and connectedness are seen as being equivalent by those who wish to evaluate us. We are being assessed according to our utility as nodes in the network. Therefore, in our community, there is a considerable importance attached to presenting papers at conferences and in peer-reviewed journals, both of which can help build our connectedness. The fun of exchanging ideas is now accompanied by an imperative for status recognition. Having been the regional editor for the Asia-Pacific for a number of years, I am now much more aware of the fallacy that publication alone is important in building and maintaining connections within the network. Connections need to be bidirectional. Someone else needs to find the paper interesting enough to refer to it, or to build on it to the next stage. It is one of the most difficult jobs negotiating between reviewers and authors when the former point out, not always politely, that the latter’s work falls short on those criteria. It is hard, as an editor, not to share some of the authors’ pain. Yet this is precisely the role of reviewers. For an editor to determine, without advice, what to publish and what not to publish, would be very dangerous. The social network that is our discipline is very dynamic, and sometimes unpredictable. The dead hand of a dictatorial editor would dampen, if not kill, that dynamism and eliminate the unpredictability. Reviewers are very important. Therefore, having partially downgraded the role of editor above, what is left for them to do? Well, selecting appropriate reviewers is a major role and, I suspect, a skill. Some of my more heartening pieces of correspondence run along the lines, ‘I’ve seen this paper five times before, and each time I’ve made recommendation for improvement – but all the authors seem to do is send it to a different journal.’ That means that we got the right reviewer for the paper, or at least made the same judgement as four other editors. It also highlights the problem that many authors International Journal of Geographical Information Science Vol. 22, No. 1, January 2008, 1–3

Assessing a single classification accuracy measure to deal with the imprecision of location and class: Fuzzy weighted kappa versus kappa

Standard accuracy assessment in raster modelling assumes that desirable classes are exhaustively exclusive, location error either does not exist or does not affect classification, and classification results are crisp, not fuzzy Failing to take these into account leads to propagation of the error. This paper introduces a fuzzy weighted Kappa measure as a global statistic which takes both of these sources of uncertainty into account in classification accuracy assessment. Calculation of the fuzzy weighted Kappa measure involves three processes. First, a virtual confusion matrix is constructed to incorporate the fuzziness of location. Then a similarity weight matrix is derived to represent the fuzziness of the category. Combining the virtual confusion matrix and the similarity weight matrix, the fuzzy weighted Kappa is calculated using the weighted Kappa algorithms. The fuzzy weighted Kappa is shown to be more flexible than standard methods and gives results closer to an intuitive assessment of accuracy. Although calculating the fuzzy weighted Kappa involves a degree of subjectivity, this poses no major limitations on the usefulness of the measure. It is therefore proposed as an alternative classification accuracy measure for the remote sensing and GISc community.

Recent Trends in IJGISc

Journal editors are uniquely positioned to observe trends in our field first-hand because of the broad flow of submissions. As the International Journal of Geographical Information Science is a truly international forum for research publication, the local bias of regional conferences is less and one can get a global picture of trends quite quickly. What is clear is that the view from place to place of what constitutes GISc research varies quite markedly and reflects local activity and concerns. The current trends, as indicated by submissions to IJGISc, seem worthy of comment. There are also other trends of importance to our authorship that need to be flagged. I will discuss these latter, more mundane, trends first.

25 Volumes of the International Journal of Geographical Information Science

Welcome to the 25th volume of the International Journal of Geographical Information Science. You may not notice it for an issue or two, but from this issue the journal has dropped its limitation on colour figures. We retain limits on publishing colour figures for purely aesthetic reasons and we will continue to ask reviewers to comment on the need for all figures submitted. The other change is an increase in the size of the journal. This year, it is growing by 10% – rather less than in the previous 3 years. We judge that this will bring the time between acceptance and publication down to much less than a year. We hope that this will reduce more but we need to be careful not to overshoot. In the 12 months to September, 2010, Manuscript Central estimates that we received a total of 595 manuscripts, 403 of which were original manuscripts (in the previous 12 months we received 564/392 and the year before that 435/310 and, in 2007, 412/308). So the increase is steady.We could publish about 93 papers a year with the 2010 page allowance, but this now rises to about 103 papers a year with the new allowance. In 2010 we had a rough ratio of 93/403 (23%) for published/submitted. This could be seen as a rejection rate of between 70 and 80%, which is about right. So, what other changes have taken place since volume 1(1) in 1987? In 1987 the journal’s name was International Journal of Geographical Information Systems. The founding editors, Terry Coppock and Eric Anderson, explained in their Editorial Review that they felt that the systems themselves were the unifying theme that linked a wide range of users and disciplines. They intended the journal to provide a focus for this diverse group (Coppock and Anderson 1987). In 1997 the title was changed under the then editor-in-chief, Peter Fisher, to reflect the growing feeling that the ‘S’ in GIS should stand for ‘Science’ as it was a growing consensus that it was the science underlying the system that was important (Goodchild 1992, Fisher 1997). This reflected the growth of a community who identified themselves as ‘GIScientists’, but did not exclude the wider group who saw the systems as tools with a very important theoretical underpinning. Through the years there have only been three editors-in-chief, Peter Fisher being the longest serving (1994–2007), but there have been a number of regional editors. The journal ‘sees’ the world in three parts: Europe and Africa, the Americas and the Asia-Pacific region. Terry Coppock and Peter Fisher in addition to being editors-in-chief looked after the United Kingdom and Europe. Now Sytze de Bruin andMonicaWachowicz share the task of looking after Europe and Africa. Eric Anderson, Steve Guptill, Keith Clark, Marc Armstrong, Harvey Miller and Mark Gahegan looked after North America. May Yuan is now our Americas’ regional editor. And, starting at a later date, Dave Abel and then Brian Lees looked after the Asia-Pacific. In 2009, the North American editor dealt with 74 papers, the Asia-Pacific editor dealt with 72 papers and the two European editors dealt with 89 papers between them. In 2010, the North American editor dealt with about 80 papers, the AsiaPacific editor dealt with about 184 papers and the two European editors dealt with about 132 International Journal of Geographical Information Science Vol. 25, No. 1, January 2011, 1–4

The Impact of DEM Error on Predictive Vegetation Mapping

Digital elevation models (DEM) are one of the most important data sources for Land Use-Land Cover (LULC) and Predictive Vegetation Mapping (PVM). A number of indices are derived from DEMs and their use depends on the nature of the classification problem and the tool being employed. In some cases it is the practice to pre-classify these prior to modelling. This chapter examines the impact of doing this on the production of a LULC classification, and on the production of a surface, or field, prediction of a single species. Secondly, the error in classification resulting from error in the original DEM is examined to give some comparison. We show that, contrary to widespread practice, leaving the input variables in an unprocessed form is clearly better than almost any of the ‘improvements’ usually made. This applied to both classification of LULC and to the prediction of a surface, or field, representing a single species. As expected, forest type mapping is likely to be quite sensitive to some level of DEM error. We can see that the DEM error has an uneven impact on the different forest types. Importantly, when increasing the level of DEM error, we found a non-linear decrease in classification performance.

Error in Spatial Ecology (PVM)

This paper looks at the error inherent in Predictive Vegetation Mapping (PVM). I review past work on assessing and tracking error in the production of the most common form in which natural resource data is stored and visualised, the area-class map. Using the extensive and well-studied Kioloa forest dataset, an alternative method of storage and display based on datasets for each individual species is described. Whilst this alternative approach is more accurate, and can carry estimates of probability for the estimated distribution of each species, it is cumbersome. Looking again at the area-class map it is suggested that, for many users, the degree of error built-into such a representation may be an acceptable cost for the convenience of a simple presentation. Having looked at the degree of error, and attempts to minimise it, the paper concludes with the observation that some types of error are the inevitable cost of attempts to simplify representation.

Comment on 'Marine GIS: Identification of mesoscale oceanic thermal fronts'

In this paper, we aim to clear up a significant conceptual error in the use of the ‘sink method’ as presented by Valavanis et al. (2005) for oceanic thermal front detection. We argue that the features identified by the authors in their paper are mostly cyclonic or cold ring eddies in the Aegean Sea.