Image classification used in mapping land cover form remotely sensed data are frequently described as being 'hard' of 'soft' yet in reality such a simple distinction is not observed and a continuum of classification softness can be defined. Using airborne sensors or imagery of two test sites in South Wales, classifications at different points along this continuum with a feedforward neural network are illustrated. It is shown that soft classification can provide a better and more accurate representation of both discrete and continuous land cover classes, resolving in particular problems associated with mixed pixels. Classifications produced at different positions along the continuum of classification softness, however, differed markedly in the representation of land cover distribution and accuracy, highlighting the need to recognize the existence of the continuum and its implications for land cover mapping from remotely sensed data. The results also highlight that the use of a soft or fuzzy classifier is only a partial solution to the mixed pixel problem; a full solution requires refinement of the training and testing stages and methods for this are discussed. Despite an ability to accommodate for the effects of mixed pixels on each of the three stages of supervised image classifications, other factors can degraded classification quality. One important issue is the presence of untrained classes. It is hon, however, that the effect of untrained classes can be reduced with the use of additional information on the typicality of class membership that can be derived form some soft classifications.
'Sub-Pixel Methods in Remote Sensing' published in 'Remote Sensing Image Analysis: Including The Spatial Domain'
No abstract is provided for this article.
Flash floods are a common, but poorly understood feature of arid environments. Much of the uncertainty associated with flash flooding events is associated with a lack of accurate environmental data. In addition to limiting the understanding of hydrological processes, this situation handicaps human use and development in such regions, necessitating the use of modelling approaches for environmental prediction. Here, a hydrological model driven mainly by information on land cover distribution (derived by satellite remote sensing) and soil properties (derived from field measurement) was used to predict sites at risk from large peak flows associated with flash flooding in a wadi located in the Eastern Desert of Egypt. The land cover map was derived from a maximum likelihood classification of a Landsat TM image and had an estimated accuracy of 89.5%. The soils of the classes depicted in this map differed markedly in terms of texture and permeability, with the field based estimates of infiltration capacity ranging from 0.07 cm h−1 for desert pavement through 14.01 cm h−1 for unconsolidated wadi bed deposits. Using the derived information within the hydrological modelling system, the discharge from the wadi and its sub-basins was predicted for an assumed severe storm scenario. The outputs of the model indicated two locations within the wadi where a very large peak discharge (>115 m3 s−1) could be expected. These sites corresponded to those that suffered flood damage in a recent storm event. The results indicate the potential to drive an integrated hydrological model from limited data to derive important and useful hydrological information in a region where data are scarce.
Superresolution mapping (SRM) is a commonly used method to cope with the problem of mixed pixels when predicting the spatial distribution within low-resolution pixels. Central to the popular SRM method is the spatial pattern model, which is utilized to represent the land cover spatial distribution within mixed pixels. The use of an inappropriate spatial pattern model limits such SRM analyses. Alternative approaches, such as deep-learning-based algorithms, which learn the spatial pattern from training data through a convolutional neural network, have been shown to have considerable potential. Deep learning methods, however, are limited by issues such as the way the fraction images are utilized. Here, a novel SRM model based on a generative adversarial network (GAN), GAN-SRM, is proposed that uses an end-to-end network to address the main limitations of existing SRM methods. The potential of the proposed GAN-SRM model was assessed using four land cover subsets and compared to hard classification and several popular SRM methods. The experimental results show that of the set of methods explored, the GAN-SRM model was able to generate the most accurate high-resolution land cover maps.
A contour-based pixel swapping method for super-resolution mapping, which combines contouring and pixel swapping super-resolution mapping approaches, that seeks to exploit the positive features of contouring and pixel swapping to produce a method that is more accurate than each alone is proposed. The accuracy of super-resolution mapping with the individual and combined techniques is explored. When combined, the error with which objects of varying shape were represented was typically greatly reduced relative to that observed from the application of the methods individually. For example, the root mean square error in mapping the boundary of an aeroplane represented in relatively fine spatial resolution imagery decreased from 14.43m with contouring and 2.95m with pixel swapping to 2.18m when the approaches were combined.
A decade ago, Volunteered Geographical Information (VGI) was identified as a new source of information that would blur the traditional boundary between producers and the consumers of data (Goodchil...
Accommodating for the differences between grasses following the C3 and C4 photosynthetic pathways in environmental research often requires information on their spatial distribution and relative abundance. Multi-temporal remote sensing may indicate the latter because these grasses have asynchronous phenologies. The relationship between remotely sensed variables and grassland composition, defined by C3(%), was explored with attention focused on two key issues associated with studies of large areas from multi-temporal datasets: the compositing period used and spatial generalizability of a selected relationship. MERIS Terrestrial Chlorophyll Index (MTCI) composites of the Great Plains were generated using compositing periods of 5, 7, 10 and 14 days. The results of a regression analysis indicated that a relationship between MTCI data and grassland composition may be formulated for the State of South Dakota with R 2 ∼0.6. The strength of the relationship was, generally, strongest for short compositing periods. The transferability of the relationship to other regions was, however, limited by its significant non-stationarity indicating a challenge for large area studies. Acknowledgements This work was supported by the NERC QUEST programme (NE/C516187/1, QUERCC). The GWR analyses were undertaken with the GWR3.0 software. MERIS data were obtained from ESA and 8-day MTCI composites were obtained from the NERC EODC. We are grateful to the referees for their helpful comments on the article.
Geographical information (GI) science is rapidly developing and interfaces with numerous disciplines. It is, therefore, not surprising that as well as making new developments, perhaps by transferring knowledge and methods from other fields of study into the arena, old issues are revisited, especially to clarify or perhaps recast important foundations. This report will focus on a few topics that have received recent attention in GI science. This includes the topics of knowledge discovery and data mining that are drawing on developments outside geography but that may significantly help the advancement of geographical science. First, however, attention is focused on a perennial issue of uncertainty.
The accuracy of a conventional supervised classification is in part a function of the training set used, notably impacted by the quantity and quality of the training cases. Since it can be costly to acquire a large number of high quality training cases, recent research has focused on methods that allow accurate classification from small training sets. Previous work has shown the potential of support vector machine (SVM) based classifiers. Here, the potential of the relevance vector machine (RVM) and sparse multinominal logistic regression (SMLR) approaches is evaluated relative to SVM classification. With both airborne and spaceborne multispectral data sets, the RVM and SMLR were able to derive classifications of similar accuracy to the SVM but required considerably fewer training cases. For example, from a training set comprising 600 cases acquired with a conventional stratified random sampling design from an airborne thematic mapper (ATM) data set, the RVM produced the most accurate classification, 93.75%, and needed only 7.33% of the available training cases. In comparison, the SVM yielded a classification that had an accuracy of 92.50% and needed 4.5 times more useful training cases. Similarly, with a Landsat ETM+ (Littleport, Cambridgeshire, UK) data set, the SVM required 4.0 times more useful training cases than the RVM. For each data set, however, the classifications derived by each classifier were of similar magnitude, differing by no more than 1.25%. Finally, for both the ATM and ETM+ (Littleport) data sets, the useful training cases by SVM and RVM had distinct and potentially predictable characteristics. Support vectors were generally atypical but lay in the boundary region between classes in feature space while the relevance vectors were atypical but anti-boundary in nature. The SMLR also tended to mostly, but not always, use extreme cases that lay away from class boundary. The results, therefore, suggest a potential to design classifier-specific intelligent training data acquisition activities for accurate classification from small training sets, especially with the SVM and RVM.
No abstract is provided for this article.
Data on the distribution of vegetation in space and time are required in a range of studies. Such data are, however, typically unavailable or are of poor quality (Williams 1994; DeFries and Townshend 1994). Often the only practicable means of acquiring data on...
Past studies focused on the relationships between land cover and urban temperature have commonly assumed stationarity and used conventional (global) regression analysis. In this study, geographically weighted regression (GWR) was used to test the spatial stationarity of the relationships between a set of land cover types (built-up, water, paddy field, and other vegetation) and the surface temperature in TaoYuan, Taiwan. By adopting the GWR approach, significant spatial non-stationarity of these relationships was observed and the strength of these relationships was markedly higher than from a conventional regression analysis. The differences have large impacts. If the regression models were used to derive an estimate of the urban heat island intensity for TaoYuan this would equate to 2.63°C and 3.17°C for the global and GWR models, respectively. This result showed that the urban heat island was underestimated by global model and this, therefore, increased potential to underestimate the risk of ill-health and discomfort for urban populations. The mapped parameters derived from GWR analyses provided useful information for planning temperature mitigation and adaptation strategies especially for the very young and elderly that are particularly sensitive to temperature.
