Search

The accuracy of a classification is fundamental to its interpretation, use and ultimately decision making. Unfortunately, the apparent accuracy assessed can differ greatly from the true accuracy. Mis-estimation of classification accuracy metrics and associated mis-interpretations are often due to variations in prevalence and the use of an imperfect reference standard. The fundamental issues underlying the problems associated with variations in prevalence and reference standard quality are revisited here for binary classifications with particular attention focused on the use of the Matthews correlation coefficient (MCC). A key attribute claimed of the MCC is that a high value can only be attained when the classification performed well on both classes in a binary classification. However, it is shown here that the apparent magnitude of a set of popular accuracy metrics used in fields such as computer science medicine and environmental science (Recall, Precision, Specificity, Negative Predictive Value, J, F1, likelihood ratios and MCC) and one key attribute (prevalence) were all influenced greatly by variations in prevalence and use of an imperfect reference standard. Simulations using realistic values for data quality in applications such as remote sensing showed each metric varied over the range of possible prevalence and at differing levels of reference standard quality. The direction and magnitude of accuracy metric mis-estimation were a function of prevalence and the size and nature of the imperfections in the reference standard. It was evident that the apparent MCC could be substantially under- or over-estimated. Additionally, a high apparent MCC arose from an unquestionably poor classification. As with some other metrics of accuracy, the utility of the MCC may be overstated and apparent values need to be interpreted with caution. Apparent accuracy and prevalence values can be mis-leading and calls for the issues to be recognised and addressed should be heeded. [ABSTRACT FROM AUTHOR]

Ground reference data are typically required to evaluate the quality of a supervised image classification analysis used to produce a thematic map from remotely sensed data. Acquiring a suitable ground data set for a rigorous assessment of classification quality can be a major challenge. An alternative approach to quality assessment is to use a model-based method such as can be achieved with a latent class analysis. Previous research has shown that the latter can provide estimates of class areal extent for a non-site specific accuracy assessment and yield estimates of producer's accuracy which are commonly used in site-specific accuracy assessment. Here, the potential for quality assessment via a latent class analysis is extended to show that an estimate of a complete confusion matrix can be predicted which allows a suite of standard accuracy measures to be generated to indicate global quality on an overall and per-class basis. In addition, information on classification uncertainty may be used to illustrate classification quality on a per-pixel basis and hence provide local information to highlight spatial variations in classification quality. Classifications of imagery from airborne and satellite-borne sensors were used to illustrate the potential of the latent class analysis with results compared against those arising from the use of a conventional ground data set. [ABSTRACT FROM AUTHOR]

The role of Earth observation (EO) data in addressing societal problems from environmental through to humanitarian should not be understated. Recent innovation in EO means provision of analysis ready data and data cubes, which allows for rapid use of EO data. This in combination with processing technologies, such as Google Earth Engine and open source algorithms/software for EO data integration and analyses, has afforded an explosion of information to answer research questions and/or inform policy making. However, there is still a need for both training and validation data within EO projects – often this can be challenging to obtain. It has been suggested that citizen science can help here to provide these data, yet there is some perceived hesitancy in using citizen science within EO projects. This paper reports on the Citizen Science 4 EO (Citizens4EO) project that aimed to obtain an in-depth understanding of researchers' and practitioners' experiences with citizen science data in EO within the UK. Through a mixed methods approach (online and in-depth surveys and a spotlight case study) it was found that although the benefits of using citizen science data in EO projects were many (and highlighted in the spotlighted "Slavery from Space" case study), there were a number of common concerns around using citizen science. These were around the mechanics of deploying citizen science and the unreliability of a potentially misinformed or undertrained citizen base. As such, comparing the results of this study with those of a similar survey undertaken in 2016, it is apparent that progress towards optimizing citizen science use in EO has been incremental but positive with evidence of the realization of the benefits of citizen science for EO (Citizens4EO). As such, we conclude by offering priority action areas to support further use of citizen science by the EO community within the UK, which ultimately should be adopted further afield. [ABSTRACT FROM AUTHOR]

This review examines high-quality research evidence that synthesises the effects of extreme heat on human health in tropical Africa. Web of Science (WoS) was used to identify research articles on the effects extreme heat, humidity, Wet-bulb Globe Temperature (WBGT), apparent temperature, wind, Heat Index, Humidex, Universal Thermal Climate Index (UTCI), heatwave, high temperature and hot climate on human health, human comfort, heat stress, heat rashes, and heat-related morbidity and mortality. A total of 5, 735 articles were initially identified, which were reduced to 100 based on a set of inclusion and exclusion criteria. The review discovered that temperatures up to 60°C have been recorded in the region and that extreme heat has many adverse effects on human health, such as worsening mental health in low-income adults, increasing the likelihood of miscarriage, and adverse effects on well-being and safety, psychological behaviour, efficiency, and social comfort of outdoor workers who spend long hours performing manual labour. Extreme heat raises the risk of death from heat-related disease, necessitating preventative measures such as adaptation methods to mitigate the adverse effects on vulnerable populations during hot weather. This study highlights the social inequalities in heat exposure and adverse health outcomes. [ABSTRACT FROM AUTHOR]

A strong relationship between night-time light (NTL) data and the areal extent of urbanized regions has been observed frequently. As urban regions have an important vertical dimension, it is hypothesized that the strength of the relationship with NTL can be increased by consideration of the volume rather than simply the area of urbanized land. Relationships between NTL and the area and volume of urbanized land were determined for a set of towns and cities in the UK, the conterminous states of the USA and countries of the European Union. Strong relationships between NTL and the area urbanized were observed, with correlation coefficients ranging from 0.9282 to 0.9446. Higher correlation coefficients were observed for the relationship between NTL and urban building volume, ranging from 0.9548 to 0.9604; The difference in the correlations obtained with volume and with area was statistically significant at the 95% level of confidence. Studies using NTL data may be strengthened by consideration of the volume rather than just area of urbanized land. [ABSTRACT FROM AUTHOR]

Super-resolution mapping (SRM) is a technique to estimate a fine spatial resolution land cover map from coarse spatial resolution fractional proportion images. SRM is often based explicitly on the use of a spatial pattern model that represents the land cover mosaic at the fine spatial resolution. Recently developed deep learning methods have considerable potential as an alternative approach for SRM, based on learning the spatial pattern of land cover from existing fine resolution data such as land cover maps. This letter proposes a deep learning-based SRM algorithm (DeepSRM). A deep convolutional neural network was first trained to estimate a fine resolution indicator image for each class from the coarse resolution fractional image, and all indicator maps were then combined to create the final fine resolution land cover map based on the maximal value strategy. The results of an experiment undertaken with simulated images show that DeepSRM was superior to conventional hard classification and a suite of popular SRM algorithms, yielding the most accurate land cover representation. Consequently, methods such as DeepSRM may help exploit the potential of remote sensing as a source of accurate land cover information. [ABSTRACT FROM AUTHOR]

The main objective of this research is to assess the impact of intra-class spectral variation on the accuracy of soft classification and super-resolution mapping. The accuracy of both analyses was negatively related to the degree of intra-class spectral variation, but the effect could be reduced through the use of spectral sub-classes. The latter is illustrated in mapping the shoreline at a sub-pixel scale from Landsat ETM+ data. Reducing the degree of intra-class spectral variation increased the accuracy of soft classification, with the correlation between predicted and actual class coverage rising from 0.87 to 0.94, and super-resolution mapping, with the RMSE in shoreline location decreasing from 41.13 m to 35.22 m. [ABSTRACT FROM AUTHOR]

Given the common trade-off between the spatial and temporal resolutions of current satellite sensors, spatial-temporal data fusion methods could be applied to produce fused remotely sensed data with synthetic fine spatial resolution (FR) and high repeat frequency. Such fused data are required to provide a comprehensive understanding of Earth's surface land cover dynamics. In this research, a novel Spatial-Temporal Fraction Map Fusion (STFMF) model is proposed to produce a series of fine-spatial-temporal-resolution land cover fraction maps by fusing coarse-spatial-fine-temporal and fine-spatial-coarse-temporal fraction maps, which may be generated from multi-scale remotely sensed images. The STFMF has two main stages. First, FR fraction change maps are generated using kernel ridge regression. Second, a FR fraction map for the date of prediction is predicted using a temporal-weighted fusion model. In comparison to two established spatial-temporal fusion methods of spatial-temporal super-resolution land cover mapping model and spatial-temporal image reflectance fusion model, STFMF holds the following characteristics and advantages: (1) it takes account of the mixed pixel problem in FR remotely sensed images; (2) it directly uses the fraction maps as input, which could be generated from a range of satellite images or other suitable data sources; (3) it focuses on the estimation of fraction changes happened through time and can predict the land cover change more accurately. Experiments using synthetic multi-scale fraction maps simulated from Google Earth images, as well as synthetic and real MODIS-Landsat images were undertaken to test the performance of the proposed STFMF approach against two benchmark spatial-temporal reflectance fusion methods: the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM) and the Flexible Spatiotemporal Data Fusion (FSDAF) model. In both visual and quantitative evaluations, STFMF was able to generate more accurate FR fraction maps and provide more spatial detail than ESTARFM and FSDAF, particularly in areas with substantial land cover changes. STFMF has great potential to produce accurate time-series fraction maps with fine-spatial-temporal-resolution that can support studies of land cover dynamics at the sub-pixel scale. [ABSTRACT FROM AUTHOR]

Land cover mapping via image classification is sometimes realized through object-based image analysis. Objects are typically constructed by partitioning imagery into spatially contiguous groups of pixels through image segmentation and used as the basic spatial unit of analysis. As it is typically desirable to know the accuracy with which the objects have been delimited prior to undertaking the classification, numerous methods have been used for accuracy assessment. This paper reviews the state-of-the-art of image segmentation accuracy assessment in land cover mapping applications. First the literature published in three major remote sensing journals during 2014–2015 is reviewed to provide an overview of the field. This revealed that qualitative assessment based on visual interpretation was a widely-used method, but a range of quantitative approaches is available. In particular, the empirical discrepancy or supervised methods that use reference data for assessment are thoroughly reviewed as they were the most frequently used approach in the literature surveyed. Supervised methods are grouped into two main categories, geometric and non-geometric, and are translated here to a common notation which enables them to be coherently and unambiguously described. Some key considerations on method selection for land cover mapping applications are provided, and some research needs are discussed. [ABSTRACT FROM AUTHOR]

Dams play a significant role in altering the spatial pattern of temperature in rivers and contribute to thermal pollution, which greatly affects the river aquatic ecosystems. Understanding the temporal and spatial variation of thermal pollution caused by dams is important to prevent or mitigate its harmful effect. Assessments based on in-situ measurements are often limited in practice because of the inaccessibility of water temperature records and the scarcity of gauges along rivers. By contrast, thermal infrared remote sensing provides an alternative approach to monitor thermal pollution downstream of dams in large rivers, because it can cover a large area and observe the same zone repeatedly. In this study, Landsat Enhanced Thematic Mapper Plus (ETM+) thermal infrared imagery were applied to assess the thermal pollution caused by two dams, the Geheyan Dam and the Gaobazhou Dam, located on the Qingjiang River, a tributary of the Yangtze River downstream of the Three Gorges Reservoir in Central China. The spatial and temporal characteristics of thermal pollution were analyzed with water temperatures estimated from 54 cloud-free Landsat ETM+ scenes acquired in the period from 2000 to 2014. The results show that water temperatures downstream of both dams are much cooler than those upstream of both dams in summer, and the water temperature remains stable along the river in winter, showing evident characteristic of the thermal pollution caused by dams. The area affected by the Geheyan Dam reaches beyond 20 km along the downstream river, and that affected by the Gaobazhou Dam extends beyond the point where the Qingjiang River enters the Yangtze River. Considering the long time series and global coverage of Landsat ETM+ imagery, the proposed technique in the current study provides a promising method for globally monitoring the thermal pollution caused by dams in large rivers. [ABSTRACT FROM AUTHOR]

Image classification for thematic mapping is a very common application in remote sensing, which is sometimes realized through object-based image analysis. In these analyses, it is common for some of the objects to be mixed in their class composition and thus violate the commonly made assumption of object purity that is implicit in a conventional object-based image analysis. Mixed objects can be a problem throughout a classification analysis, but are particularly challenging in the training stage as they can result in degraded training statistics and act to reduce mapping accuracy. In this paper the potential of using mixed objects in training object-based image classifications is evaluated. Remotely sensed data were submitted to a series of segmentation analyses from which a range of under- to over-segmented outputs were intentionally produced. Training objects were then selected from the segmentation outputs, resulting in training data sets that varied in terms of size (i.e. number of objects) and proportion of mixed objects. These training data sets were then used with an artificial neural network and a generalized linear model, which can accommodate objects of mixed composition, to produce a series of land cover maps. The use of training statistics estimated based on both pure and mixed objects often increased classification accuracy by around 25% when compared with accuracies obtained from the use of only pure objects in training. So rather than the mixed objects being a problem, they can be an asset in classification and facilitate land cover mapping from remote sensing. It is, therefore, desirable to recognize the nature of the objects and possibly accommodate mixed objects directly in training. The results obtained here may also have implications for the common practice of seeking an optimal segmentation output, and also act to challenge the widespread view that object-based classification is superior to pixel-based classification. [ABSTRACT FROM AUTHOR]

The accuracy of a map is dependent on the reference dataset used in its construction. Classification analyses used in thematic mapping can, for example, be sensitive to a range of sampling and data quality concerns. With particular focus on the latter, the effects of reference data quality on land cover classifications from airborne thematic mapper data are explored. Variations in sampling intensity and effort are highlighted in a dataset that is widely used in mapping and modelling studies; these may need accounting for in analyses. The quality of the labelling in the reference dataset was also a key variable influencing mapping accuracy. Accuracy varied with the amount and nature of mislabelled training cases with the nature of the effects varying between classifiers. The largest impacts on accuracy occurred when mislabelling involved confusion between similar classes. Accuracy was also typically negatively related to the magnitude of mislabelled cases and the support vector machine (SVM), which has been claimed to be relatively insensitive to training data error, was the most sensitive of the set of classifiers investigated, with overall classification accuracy declining by 8% (significant at 95% level of confidence) with the use of a training set containing 20% mislabelled cases. [ABSTRACT FROM AUTHOR]

Remotely sensed imagery is an attractive source of information for mapping and monitoring land cover. Fine spatial resolution imagery is typically acquired infrequently, but fine temporal resolution systems commonly provide coarse spatial resolution imagery. Sub-pixel land cover change mapping is a method that aims to use the advantages of these multiple spatial and temporal resolution sensing systems. This method produces fine spatial and temporal resolution land cover maps, by updating fine spatial resolution land cover maps using coarse spatial resolution remote sensing imagery. A critical issue for sub-pixel land cover change mapping is downscaling coarse spatial resolution fraction maps estimated by soft classification to a fine spatial resolution land cover map. The relationship between a historic fine spatial resolution map and a contemporary fine spatial resolution map to be estimated at a more recent date plays an important role in the downscaling procedure. A change strategy based on the assumption that the change for each land cover class in a coarse spatial resolution pixel is unidirectional was shown to be a promising means to describe this relationship. This paper aims to assess this change strategy by analyzing the factors that affect the accuracy of the change strategy, using six subsets of the National Land Cover Database (NLCD) of USA. The results show that the spatial resolution of coarse pixels, the time interval of the previous fine resolution land cover map and the current coarse spatial resolution images, and the thematic resolution of the used land cover class scheme have considerable influence on the accuracy of the change strategy. The accuracy of the change strategy decreases with the coarsening of spatial resolution, an increase of time interval, and an increase of thematic resolution. The results also indicate that, when the historic land cover map has a 30 m resolution, like the NLCD, the average accuracy of the change strategy is still as high as 92% when the coarse spatial resolution data used had a resolution of ~1000 m, confirming the effectiveness of the change strategy used in sub-pixel land cover change mapping for use with popular remote sensing systems. [ABSTRACT FROM AUTHOR]

Spatial records of species are commonly misidentified, which can change the predicted distribution of a species obtained from a species distribution model (SDM). Experiments were undertaken to predict the distribution of real and simulated species using MaxEnt and presence-only data "contaminated" with varying rates of misidentification error. Additionally, the difference between the niche of the target and contaminating species was varied. The results show that species misidentification errors may act to contract or expand the predicted distribution of a species while shifting the predicted distribution towards that of the contaminating species. Furthermore the magnitude of the effects was positively related to the ecological distance between the species' niches and the size of the error rates. Critically, the magnitude of the effects was substantial even when using small error rates, smaller than common average rates reported in the literature, which may go unnoticed while using a standard evaluation method, such as the area under the receiver operating characteristic curve. Finally, the effects outlined were shown to impact negatively on practical applications that use SDMs to identify priority areas, commonly selected for various purposes such as management. The results highlight that species misidentification should not be neglected in species distribution modeling. [ABSTRACT FROM AUTHOR]

Crowdsourcing has become a popular means to acquire data about the Earth and its environment inexpensively, but the data-sets obtained are typically imperfect and of unknown quality. Two common imperfections with crowdsourced data are the contributions from cheats or spammers and missing cases. The effect of the latter two imperfections on a method to evaluate the accuracy of crowdsourced data via a latent class model was explored. Using simulated and real data-sets, it was shown that the method is able to derive useful information on the accuracy of crowdsourced data even when the degree of imperfection was very high. The practical potential of this ability to obtain accuracy information within the geospatial sciences and the realm of Digital Earth applications was indicated with reference to an evaluation of building damage maps produced by multiple bodies after the 2010 earthquake in Haiti. Critically, the method allowed data-sets to be ranked in approximately the correct order of accuracy and this could help ensure that the most appropriate data-sets are used. [ABSTRACT FROM PUBLISHER]

Abstract: The remote sensing science and application communities have developed increasingly reliable, consistent, and robust approaches for capturing land dynamics to meet a range of information needs. Statistically robust and transparent approaches for assessing accuracy and estimating area of change are critical to ensure the integrity of land change information. We provide practitioners with a set of “good practice” recommendations for designing and implementing an accuracy assessment of a change map and estimating area based on the reference sample data. The good practice recommendations address the three major components: sampling design, response design and analysis. The primary good practice recommendations for assessing accuracy and estimating area are: (i) implement a probability sampling design that is chosen to achieve the priority objectives of accuracy and area estimation while also satisfying practical constraints such as cost and available sources of reference data; (ii) implement a response design protocol that is based on reference data sources that provide sufficient spatial and temporal representation to accurately label each unit in the sample (i.e., the “reference classification” will be considerably more accurate than the map classification being evaluated); (iii) implement an analysis that is consistent with the sampling design and response design protocols; (iv) summarize the accuracy assessment by reporting the estimated error matrix in terms of proportion of area and estimates of overall accuracy, user's accuracy (or commission error), and producer's accuracy (or omission error); (v) estimate area of classes (e.g., types of change such as wetland loss or types of persistence such as stable forest) based on the reference classification of the sample units; (vi) quantify uncertainty by reporting confidence intervals for accuracy and area parameters; (vii) evaluate variability and potential error in the reference classification; and (viii) document deviations from good practice that may substantially affect the results. An example application is provided to illustrate the recommended process. [Copyright &y& Elsevier]

The author focuses on publications related to remote sensing. He mentions that as of December 2013 there are 27 journals related to remote-sensing in the online subscription-based scientific citation indexing service Web of Science. He also mentions the Journal Citation Reports, an annual publication related to academic journals in the sciences and social sciences and also talks about significant increase in research papers on remote sensing.

Area estimation is a common application of remote sensing especially in relation to studies of land cover change. The use of an imperfect, nongold-standard, reference is shown to be a source of substantial error in estimates of change area. The relationships between the accuracy of land cover classifications, both real and perceived, together with area of change are explored over the full range of change abundance. The magnitude of mis-estimation varies with the abundance of change and the quality of the data sets used but may be large. For scenarios based on realistic values of data set quality for the common situation in which change is rare, ≤1%, change area was overestimated by up to two orders of magnitude because of reference data error. The results highlight that the ability of remote sensing to provide accurate information on change area is limited by reference data error and is, in part, a function of change abundance. [ABSTRACT FROM AUTHOR]

Abstract: The area of land use or land cover change obtained directly from a map may differ greatly from the true area of change because of map classification error. An error-adjusted estimator of area can be easily produced once an accuracy assessment has been performed and an error matrix constructed. The estimator presented is a stratified estimator which is applicable to data acquired using popular sampling designs such as stratified random, simple random and systematic (the stratified estimator is often labeled a poststratified estimator for the latter two designs). A confidence interval for the area of land change should also be provided to quantify the uncertainty of the change area estimate. The uncertainty of the change area estimate, as expressed via the confidence interval, can then subsequently be incorporated into an uncertainty analysis for applications using land change area as an input (e.g., a carbon flux model). Accuracy assessments published for land change studies should report the information required to produce the stratified estimator of change area and to construct confidence intervals. However, an evaluation of land change articles published between 2005 and 2010 in two remote sensing journals revealed that accuracy assessments often fail to include this key information. We recommend that land change maps should be accompanied by an accuracy assessment that includes a clear description of the sampling design (including sample size and, if relevant, details of stratification), an error matrix, the area or proportion of area of each category according to the map, and descriptive accuracy measures such as user''s, producer''s and overall accuracy. Furthermore, mapped areas should be adjusted to eliminate bias attributable to map classification error and these error-adjusted area estimates should be accompanied by confidence intervals to quantify the sampling variability of the estimated area. Using data from the published literature, we illustrate how to produce error-adjusted point estimates and confidence intervals of land change areas. A simple analysis of uncertainty based on the confidence bounds for land change area is applied to a carbon flux model to illustrate numerically that variability in the land change area estimate can have a dramatic effect on model outputs. [Copyright &y& Elsevier]

Abstract: The classification of remotely sensed images such as aerial photographs or satellite sensor images for deriving ecosystem-related maps (e.g., land cover, land use, vegetation, soil) is generally based on clustering of spatial entities within a spectral space. In most cases, Boolean logic is applied in order to map landscape patterns. One major concern is that this implies an ability to divide the gradual variability of the Earth''s surface into a finite number of discrete non-overlapping classes, which are considered to be exhaustively defined and mutually exclusive. This type of approach is often inappropriate given the continuous nature of many ecosystem properties. Moreover, the standard data processing and image classification methods used will involve the loss of information as the continuous quantitative spectral information is degraded into a set of discrete classes. This leads to uncertainty in the products resulting from the use of remote sensing tools. It follows that any estimated ecosystem property has an associated error and/or uncertainty of unknown magnitude, and that the statistical quantification of uncertainty should be a core part of scientific research using remote sensing. In this paper we will review recent attempts to take explicitly into account uncertainty when mapping ecosystems. [Copyright &y& Elsevier]

Abstract: The potential of super-resolution mapping (SRM) techniques for the representation of lakes was evaluated using both an established and a newly proposed method. Both super-resolution mapping techniques were typically able to provide representations that were visually and quantitatively more realistic than standard hard classifications. The new technique was able to represent more small lakes than the established technique. The results also demonstrate the value of using a time series of images as input to the super-resolution analysis, enabling researchers to usefully exploit the typically fine temporal resolution of coarse spatial resolution sensors for land cover mapping. [Copyright &y& Elsevier]

To explore the impacts of imperfect reference data on the accuracy of species distribution model predictions. The main focus is on impacts of the quality of reference data (labelling accuracy) and, to a lesser degree, data quantity (sample size) on species presence-absence modelling. The paper challenges the common assumption that some popular measures of model accuracy and model predictions are prevalence independent. It highlights how imperfect reference data may impact on a study and the actions that may be taken to address problems. The theoretical independence of prevalence of popular accuracy measures, such as sensitivity, specificity, true skills statistics (TSS) and area under the receiver operating characteristic curve (AUC), is unlikely to occur in practice due to reference data error; all of these measures of accuracy, together with estimates of species occurrence, showed prevalence dependency arising through the use of a non-gold-standard reference. The number of cases used also had implications for the ability of a study to meet its objectives. Means to reduce the negative effects of imperfect reference data in study design and interpretation are suggested. [ABSTRACT FROM AUTHOR]

Abstract: The ground data used as a reference in the validation of land cover change products are often not an ideal gold standard but degraded by error. The effects of ground reference data error on the accuracy of land cover change detection and the accuracy of estimates of the extent of change were evaluated. Twelve data sets were simulated to allow the exploration of the impacts of a spectrum of ground data imperfections on the estimation of the producer''s and user''s accuracy of change as well as of change extent. Simulated data were used since this ensured that the actual properties of the data were known and to exclude effects due to other sources of ground reference data error; although the impacts of simulated reference data error on two real confusion matrices are also illustrated. The imperfections evaluated ranged from the inclusion of small amounts of known error into the ground reference data through to the extreme situation in which ground data were absent. The results show that even small amounts of error in the ground reference data can introduce large error into studies of land cover change by remote sensing and reinforce the desire to avoid the expression ground truth as this might imply that the data are a gold standard reference. The effect of reference data imperfections was dependent on the degree of association between the errors in the cross-tabulated data sets. For example, in the scenarios investigated, a 10% error in the reference data set introduced an underestimation of the producer''s accuracy of 18.5% if the errors were independent but an over-estimation of the producer''s accuracy of 12.3% if the errors were correlated. The magnitude of the mis-estimation of the producer''s accuracy was also a function of the amount of change and greatest at low levels of change. The amount of land cover change estimated also varied greatly as a function of ground reference data error. Some possible methods to reduce or even remove the impacts of ground reference data error were illustrated. These ranged from simple algebraic means to estimate the actual values of accuracy and change extent if the imperfections were known through to a latent class analysis that allowed the assessment of classification accuracy and estimation of change extent without the use of ground reference data if the underlying model is defined appropriately. [Copyright &y& Elsevier]

Support vector machines (SVM) are attractive for the classification of remotely sensed data with some claims that the method is insensitive to the dimensionality of the data and, therefore, does not require a dimensionality-reduction analysis in preprocessing. Here, a series of classification analyses with two hyperspectral sensor data sets reveals that the accuracy of a classification by an SVM does vary as a function of the number of features used. Critically, it is shown that the accuracy of a classification may decline significantly (at 0.05 level of statistical significance) with the addition of features, particularly if a small training sample is used. This highlights a dependence of the accuracy of classification by an SVM on the dimensionality of the data and, therefore, the potential value of undertaking a feature-selection analysis prior to classification. Additionally, it is demonstrated that, even when a large training sample is available, feature selection may still be useful. For example, the accuracy derived from the use of a small number of features may be non-inferior (at 0.05 level of significance) to that derived from the use of a larger feature set providing potential advantages in relation to issues such as data storage and computational processing costs. Feature selection may, therefore, be a valuable analysis to include in preprocessing operations for classification by an SVM. [ABSTRACT FROM AUTHOR]

Abstract: The comparison of classification accuracy statements has generally been based upon tests of difference or inequality when other scenarios and approaches may be more appropriate. Procedures for evaluating two scenarios with interest focused on the similarity in accuracy values, non-inferiority and equivalence, are outlined following a discussion of tests of difference (inequality). It is also suggested that the confidence interval of the difference in classification accuracy may be used as well as or instead of conventional hypothesis testing to reveal more information about the disparity in the classification accuracy values compared. [Copyright &y& Elsevier]

The Earth is undergoing an accelerated rate of native ecosystem conversion and degradation and there is increased interest in measuring and modelling biodiversity from space. Biogeographers have a long-standing interest in measuring patterns of species occurrence and distributional movements and an interest in modelling species distributions and patterns of diversity. Much progress has been made in identifying plant species from space using high-resolution satellites (QuickBird, IKONOS), while the measurement of species movements has become commonplace with the ARGOS satellite tracking system which has been used to track the movements of thousands of individual animals. There have been significant advances in land-cover classifications by combining data from multi-passive and active sensors, and new classification techniques. Species distribution modelling has been growing at a striking rate and the incorporation of spaceborne data on climate, topography, land cover, and vegetation structure has great potential to improve models. There have been signifi cant advances in modelling species richness, alpha diversity, and beta diversity using multisensors to quantify land-cover classifications and landscape metrics, measures of productivity, and measures of heterogeneity. Remote sensing of nature reserves can provide natural resources managers with near real-time data within and around reserves that can be used to support conservation efforts anywhere in the world. Future research should focus on incorporating recent spaceborne sensors, more extensive integration of available spaceborne imagery, and the collection and dissemination of high-quality field data. This will improve our understanding of the distribution of life on earth. [ABSTRACT FROM AUTHOR]

The coastal zone is under considerable pressure from development and is subject to change. Consequently, shoreline monitoring has grown in importance. Remotely sensed imagery from satellite sensors has been used as an alternative to conventional methods, such as those based on the interpretation of aerial photography and ground-based surveying, for monitoring shoreline position. However, the accuracy of shoreline mapping from satellite sensor imagery has been limited because of the relatively coarse spatial resolution (>10 m) of the sensors commonly used. Because of major practical and financial constraints, very fine spatial resolution (>5 m) data are often impractical for mapping large stretches of shoreline, so refinement of image analysis methods are needed to extract the desired subpixel-scale information from relatively coarse spatial resolution imagery. In this paper, the potential to map the shoreline at a subpixel scale from a soft classification of relatively coarse spatial resolution satellite sensor imagery was evaluated. Unlike conventional approaches, the methods used allowed the shoreline to be mapped within image pixels and have the potential to yield an accurate and realistic prediction of shoreline location. The approach involved the use of a soft image classification to estimate the subpixel-scale thematic composition of image pixels, which were then located geographically through postclassification analysis. Specifically, a contouring and geostatistical method based on a twopoint histogram was used to position geographically the shoreline within image pixels. The approach was applied to differently shaped shoreline extracts in imagery at two spatial resolutions. The most accurate prediction of the shoreline position from images with 16- and 32-m spatial resolutions were typically for a simple linear stretch of coast for which the smallest root mean square error values were ≤1.20 m. The shoreline predictions satisfied the map accuracy standards specified for large-scale maps. [ABSTRACT FROM AUTHOR]

The article discusses the comparison of maps, which is considered as a basic application in geographic information system. It is emphasized that map comparisons have concentrated on different map characteristics. The four main reasons to compare maps are discussed, as well as the analytical unit and method of comparison. The map comparison is based on a cross-tabulation of the labels allocated to a series of pixels chosen from the mapped region. It has been noted that some researchers have adopted information theory-based approaches to accommodate different classes of maps.

Despite a growing interest in species distribution modelling, relatively little attention has been paid to spatial autocorrelation and non-stationarity. Both spatial autocorrelation (the tendency for adjacent locations to be more similar than distant ones) and non-stationarity (the variation in modelled relationships over space) are likely to be common properties of ecological systems. This paper focuses on non-stationarity and uses two local techniques, geographically weighted regression (GWR) and varying coefficient modelling (VCM), to assess its impact on model predictions. We extend two published studies, one on the presence–absence of calandra larks in Spain and the other on bird species richness in Britain, to compare GWR and VCM with the more usual global generalized linear modelling (GLM) and generalized additive modelling (GAM). For the calandra lark data, GWR and VCM produced better-fitting models than GLM or GAM. VCM in particular gave significantly reduced spatial autocorrelation in the model residuals. GWR showed that individual predictors became stationary at different spatial scales, indicating that distributions are influenced by ecological processes operating over multiple scales. VCM was able to predict occurrence accurately on independent data from the same geographical area as the training data but not beyond, whereas the GAM produced good results on all areas. Individual predictions from the local methods often differed substantially from the global models. For the species richness data, VCM and GWR produced far better predictions than ordinary regression. Our analyses suggest that modellers interpolating data to produce maps for practical actions (e.g. conservation) should consider local methods, whereas they should not be used for extrapolation to new areas. We argue that local methods are complementary to global methods, revealing details of habitat associations and data properties which global methods average out and miss. [ABSTRACT FROM AUTHOR]

Remote sensing classification has the potential to provide important information, such as tree species distribution maps, to ecologists, at a range of spatial and temporal scales. However, standard classification procedures often fail to provide the high accuracies required for many ecological applications. Previously, a modified remote sensing classification technique was used to provide very high classification accuracies for one or two classes (e.g. species) of interest. The aim of this paper was to demonstrate that the output from the method can be suitable for spatial ecological analyses, and to provide a generic simulation framework for assessing the adequacy of any given remote sensing classification for such analyses. Marked point pattern analysis (MPPA) was applied to tree species distribution data obtained for sycamore Acer pseudoplatanus and ash Fraxinus excelsior from a 400 ha ancient semi-natural woodland in southern England using the modified remote sensing classification method to test several hypotheses of ecological interest relating to the spatial distribution and interaction of these species. Monte Carlo simulation methods were then used to evaluate the data and data quality requirements of the MPPA to check that the classified tree species maps for sycamore and ash were adequate. Using the combined method the spatial distributions for sycamore and ash were found to be aggregated and inter-dependent at a range of spatial scales. Together, the remote sensing classification and simulation approaches provide the basis for exploiting more fully the potential of remote sensing to provide information of value to ecologists. [ABSTRACT FROM AUTHOR]

Abstract: The design of the training stage of a supervised classification should account for the properties of the classifier to be used. Consideration of the way the classifier operates may enable the training stage to be designed in a manner which ensures that the aim of the classification is satisfied with the use of a small, inexpensive, training set. It may, therefore, be possible to reduce the training set size requirements from that generally expected with the use of standard heuristics. Substantial reductions in training set size may be possible if interest is focused on a single class. This is illustrated for mapping cotton in north-western India by support vector machine type classifiers. Four approaches to reducing training set size were used: intelligent selection of the most informative training samples, selective class exclusion, acceptance of imprecise descriptions for spectrally distinct classes and the adoption of a one-class classifier. All four approaches were able to reduce the training set size required considerably below that suggested by conventional widely used heuristics without significant impact on the accuracy with which the class of interest was classified. For example, reductions in training set size of ∼90% from that suggested by a conventional heuristic are reported with the accuracy of cotton classification remaining nearly constant at ∼95% and ∼97% from the user''s and producer''s perspectives respectively. [Copyright &y& Elsevier]

Abstract: The accuracy of a supervised image classification is a function of the training data used in its generation. It is, therefore, critical that the training stage of a supervised classification is designed to provide the necessary information. Guidance on the design of the training stage of a classification typically calls for the use of a large sample of randomly selected pure pixels in order to characterise the classes. Such guidance is generally made without regard to the specific nature of the application in-hand, including the classifier to be used. The design of the training stage should really be based on the classifier to be used since individual training cases can vary in value as can any one training set to a range of classifiers. It is argued here that the training stage can be designed on the basis of the way the classifier operates and with emphasis on the desire to separate the classes rather than describe them. An approach to the training of a support vector machine (SVM) classifier that is the opposite of that generally promoted for training set design is suggested. This approach uses a small sample of mixed spectral responses drawn from purposefully selected locations (geographical boundaries) in training. The approach is based on mixed pixels which are normally masked-out of analyses as undesirable and problematic. A sample of such data should, however, be easier and cheaper to acquire than that suggested by conventional approaches. This new approach to training set design was evaluated against conventional approaches with a set of classifications of agricultural crops from satellite sensor data. The main result was that classifications derived from the use of the mixed spectral responses and the conventional approach did not differ significantly, with the overall accuracy of classifications generally ∼92%. [Copyright &y& Elsevier]

Abstract: The understanding and management of biodiversity is often limited by a lack of data. Remote sensing has considerable potential as a source of data on biodiversity at spatial and temporal scales appropriate for biodiversity management. To-date, most remote sensing studies have focused on only one aspect of biodiversity, species richness, and have generally used conventional image analysis techniques that may not fully exploit the data''s information content. Here, we report on a study that aimed to estimate biodiversity more fully from remotely sensed data with the aid of neural networks. Two neural network models, feedforward networks to estimate basic indices of biodiversity and Kohonen networks to provide information on species composition, were used. Biodiversity indices of species richness and evenness derived from the remotely sensed data were strongly correlated with those derived from field survey. For example, the predicted tree species richness was significantly correlated with that observed in the field (r =0.69, significant at the 95% level of confidence). In addition, there was a high degree of correspondence (∼83%) between the partitioning of the outputs from Kohonen networks applied to tree species and remotely sensed data sets that indicated the potential to map species composition. Combining the outputs of the two sets of neural network based analyses enabled a map of biodiversity to be produced. [Copyright &y& Elsevier]

Conventional approaches to training a supervised image classification aim to fully describe all of the classes spectrally. To achieve a complete description of each class in feature space, a large training set is typically required. It is not, however, always necessary to have training statistics that provide a complete and representative description of the classes, especially if using nonparametric classifiers. For classification by a support vector machine, only the training samples that are support vectors, which lie on part of the edge of the class distribution in feature space, are required; all other training samples provide no contribution to the classification analysis. If regions likely to furnish support vectors can be identified in advance of the classification, it may be possible to intelligently select useful training samples. The ability to target useful training samples may allow accurate classification from small training sets. This potential for intelligent training sample collection was explored for the classification of agricultural crops from multispectral satellite sensor data. With a conventional approach to training, only a quarter of the training samples acquired actually made a positive contribution to the analysis and allowed the crops to be classified to a high accuracy (92.5%). The majority of the training set, therefore, was unnecessary as it made no contribution to the analysis. Using ancillary information on soil type, however, it would be possible to constrain the training sample acquisition process. By limiting training sample acquisition only to regions with a specific soil type, it was possible to use a small training set to classify the data without loss of accuracy. Thus, a small number of intelligently selected training samples may be used to classify a data set as accurately as a larger training set derived in a conventional manner. The results illustrate the potential to direct training data acquisition strategies to target the most useful training samples to allow efficient and accurate image classification. [Copyright &y& Elsevier]

This article aims to test for and explore spatial nonstationarity in the relationship between avian species richness and a set of explanatory variables to further the understanding of species diversity variation. Sub-Saharan Africa. Geographically weighted regression was used to study the relationship between species richness of the endemic avifauna of sub-Saharan Africa and a set of perceived environmental determinants, comprising the variables of temperature, precipitation and normalized difference vegetation index. The relationships between species richness and the explanatory variables were found to be significantly spatially variable and scale-dependent. At local scales > 90% of the variation was explained, but this declined at coarser scales, with the greatest sensitivity to scale variation evident for narrow ranging species. The complex spatial pattern in regression model parameter estimates also gave rise to a spatial variation in scale effects. Relationships between environmental variables are generally assumed to be spatially stationary and conventional, global, regression techniques are therefore used in their modelling. This assumption was not satisfied in this study, with the relationships varying significantly in space. In such circumstances the average impression provided by a global model may not accurately represent conditions locally. Spatial nonstationarity in the relationship has important implications, especially for studies of species diversity patterns and their scaling. [ABSTRACT FROM AUTHOR]

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. [Copyright &y& Elsevier]

Support vector machines (SVMs) have considerable potential as classifiers of remotely sensed data. A constraint on their application in remote sensing has been their binary nature, requiring multiclass classifications to be based upon a large number of binary analyses. Here, an approach for multiclass classification of airborne sensor data by a single SVM analysis is evaluated against a series of classifiers that are widely used in remote sensing, with particular regard to the effect of training set size on classification accuracy. In addition to the SVM, the same datasets were classified using a discriminant analysis, decision tree, and multilayer perceptron neural network. The accuracy statements of the classifications derived from the different classifiers were compared in a statistically rigorous fashion that accommodated for the related nature of the samples used in the analyses. For each classification technique, accuracy was positively related with the size of the training set. In general, the most accurate classifications were derived from the SVM approach, and with the largest training set the SVM classification was significantly (p < 0.05) more accurate (93.75%) than that derived from the discriminant analysis (90.00%) and decision tree algorithms (90.31%). Although each classifier could yield a very accurate classification, >90% correct, the classifiers differed in the ability to correctly label individual cases and so may be suitable candidates for an ensemble-based approach to classification. [ABSTRACT FROM AUTHOR]

There has been much debate over the meaning and content of Geographic Information System. Although a geographical component is always present in GIS, the degree to which it has been recognized and accommodated within studies has varied considerably. This report focuses on some of the recently published research literature in which the geographical component, notably in relation to some issues of scale and spatial nonstationarity, has been stressed. The report will focus on two overlapping areas, first the increasing recognition of the spatial domain in statistical analyses and secondly the increasing recognition of spatial issues in remote sensing.

Aim Conservation activities have increasingly focused on issues at the level of the landscape but are constrained by limited data and knowledge relating to biodiversity at this scale. Satellite remote sensing has considerable, but under-exploited, potential as a source of information on biodiversity at the landscape level. Remote sensing has generally been used to assess biodiversity indirectly, using approaches that often fail to fully exploit the information content of the imagery and typically only with regard to the species richness component of biodiversity. The aim of this paper was to assess the potential of remote sensing as a source of information on the richness, evenness and composition of tree species in a tropical rain forest. Location The test site was a c. 225 km[SUP2] region centred on the Danum Valley Field Centre, Borneo. This test site contained regions of undisturbed and differentially logged rain forest. Methods Data on tree biodiversity had been acquired for fifty-two sample plots by standard field survey methods and were used to derive summary indices of biodiversity for seedlings, saplings and mature trees. Differences between logged and unlogged sites were evaluated by comparison of the indices and species accumulation curves. A Landsat Thematic Mapper (TM) image of the site acquired close to the date of the field survey was obtained and rigorously pre-processed. Feedforward neural networks were used to derive predictions of biodiversity indices from the imagery. A Kohonen self organizing map neural network was used to ordinate the field data to derive classes of forest defined by relative similarity in species composition. The separability of the defined classes in the Landsat TM image was evaluated with a discriminant analysis. Results Analyses of the field data revealed considerable variation in the biodiversity of seedlings, saplings and trees at the site, associated, in part, with differences in logging activities. This variation... [ABSTRACT FROM AUTHOR]

The full realization of the potential of remote sensing as a source of environmental information requires an ability to generalize in space and time. Here, the ability to generalize in space was investigated through an analysis of the transferability of predictive relations for the estimation of tropical forest biomass from Landsat TM data between sites in Brazil, Malaysia and Thailand. The data sets for each test site were acquired and processed in a similar fashion to facilitate the analyses. Three types of predictive relation, based on vegetation indices, multiple regression and feedforward neural networks, were developed for biomass estimation at each site. For each site, the strongest relationships between the biomass predicted and that measured from field survey was obtained with a neural network developed specifically for the site (r>0.71, significant at the 99% level of confidence). However, with each type of approach problems in transferring a relation to another site were observed. In particular, it was apparent that the accuracy of prediction, as indicated by the correlation coefficient between predicted and measured biomass, declined when a relation was transferred to a site other than that upon which it was developed. Part of this problem lies with the observed variation in the relative contribution of the different spectral wavebands to predictive relations for biomass estimation between sites. It was, for example, apparent that the spectral composition of the vegetation indices most strongly related to biomass differed greatly between the sites. Consequently, the relationship between predicted and measured biomass derived from vegetation indices differed markedly in both strength and direction between sites. Although the incorporation of test site location information into an analysis resulted in an increase in the strength of the relationship between predicted and actual biomass, considerable further research is required on the problems associated with transferring predictive relations. [Copyright &y& Elsevier]

Focuses on developments in the field of geographical information science. Significance of knowledge discovery and data mining developments on geographical science; Relevance of the issue of uncertainty on geographical information science.

The production of thematic maps, such as those depicting land cover, using an image classification is one of the most common applications of remote sensing. Considerable research has been directed at the various components of the mapping process, including the assessment of accuracy. This paper briefly reviews the background and methods of classification accuracy assessment that are commonly used and recommended in the research literature. It is, however, evident that the research community does not universally adopt the approaches that are often recommended to it, perhaps a reflection of the problems associated with accuracy assessment, and typically fails to achieve the accuracy targets commonly specified. The community often tends to use, unquestioningly, techniques based on the confusion matrix for which the correct application and interpretation requires the satisfaction of often untenable assumptions (e.g., perfect coregistration of data sets) and the provision of rarely conveyed information (e.g., sampling design for ground data acquisition). Eight broad problem areas that currently limit the ability to appropriately assess, document, and use the accuracy of thematic maps derived from remote sensing are explored. The implications of these problems are that it is unlikely that a single standardized method of accuracy assessment and reporting can be identified, but some possible directions for future research that may facilitate accuracy assessment are highlighted. [ABSTRACT FROM AUTHOR]

Neural networks are attractive tools for the derivation of thematic maps from remotely sensed data. Most attention has focused on the multilayer perceptron (MLP) network but other network types are available and have different properties that may sometimes be more appropriate for some applications. Here a MLP, radial basis function (RBF) and probabilistic neural network (PNN) were used to classify remotely sensed data of an agricultural site. The accuracy of these classifications ranged from 86.25–91.25%. The accuracy of the PNN classification could be increased through the incorporation of prior probabilities of class membership but the accuracy of each classification could also be degraded by the presence of an untrained class. Post-classification analyses, however, could be used to identify potentially misclassified cases, including those belonging to an untrained class, to increase accuracy. The effect of the post-classification analysis on the accuracy of the classification derived from each of the three network types investigated differed and it is suggested that network type be selected carefully to meet the requirements of the application in-hand. [ABSTRACT FROM AUTHOR]

Abstract The biomass and biomass dynamics of forests are major uncertainties in our understanding of tropical environments. Remote sensing is often the only practical means of acquiring information on forest biomass but has not always been used successfully. Here the conventional approaches to the estimation of forest biomass from remotely sensed data were evaluated relative to techniques based on the application of artificial neural networks. Together these approaches were used to estimate and map the biomass of tropical forests in north-eastern Borneo from Landsat TM data. The neural networks were found to be particularly suited to the application. A basic multi-layer perceptron network, for example, provided estimates of biomass that were strongly correlated with those measured in the field (r = 0.80). Moreover, these estimates were more strongly correlated with biomass than those derived from 230 conventional vegetation indices, including the widely used normalized difference vegetation index (NDVI). [ABSTRACT FROM AUTHOR]

Thematic maps are often derived from remotely sensed imagery via a supervised image classification analysis. The training and testing stages of a supervised image classification may proceed ignorant of the presence of some classes in the region to be mapped. This violates the assumption of an exhaustively defined set of classes that is often made in classification analyses. In such circumstances, the overall accuracy of a thematic map produced by the application of a trained classifier will be less than the accuracy of the classification of the test set by the same classifier. This situation arises because the cases of an untrained class can normally only be commissioned into the set of trained classes. Simple mathematical relationships between classification and map accuracy are shown for assessments of overall, user's and producer's accuracy. For example, it is shown that in a simple scenario the accuracy of a thematic map is less than that of a classification, scaling as a function of the abundance of the untrained class(es). Impacts on other estimates made from thematic maps, such as class areal extent, are also briefly discussed. When using a thematic map, care is needed in interpreting and using classification accuracy assessments as sometimes they may not reflect properties of the map well. • Supervised classifications often wrongly assume classes are exhaustively defined. • Classification and map accuracy can be different. • Presence of an untrained class can reduce overall accuracy. • Relationships between map and classification accuracy (overall & per-class) defined. • General or class-specific impacts also observed area estimates. [ABSTRACT FROM AUTHOR]

Abstract. Remote sensing is the only feasible means of mapping and monitoring land cover at regional to global scales. Unfortunately the maps are generally derived through the use of a conventional 'hard' classification algorithm and depict classes separated by sharp boundaries. Such approaches and representations are often inappropriate particularly when the land cover being represented may be considered to be fuzzy. The definition of boundaries between classes can therefore be difficult from remotely sensed data, particularly for continuous land cover classes which are separated by a fuzzy boundary which may also vary spatially in time. In this paper a neural network was used to derive fuzzy classifications of land cover along a transect crossing the transition from moist semi-deciduous forest to savanna in West Africa in February and December 1990. The fuzzy classifications revealed both sharp and gradual boundaries between classes located along the transect. In particular, the fuzzy classifications enabled the definition of important boundary properties, such as width and temporal displacement. [ABSTRACT FROM AUTHOR]

Training patterns vary in their importance in image classification. Consequently, the selection and refinement of training sets can have a major impact on classification accuracy. For classification by a neural network, training patterns that lie close to the location of decision boundaries in feature space may aid the derivation of an accurate classification. The role of such border training patterns and their identification is discussed in relation to a series of crop classifications from airborne Thematic Mapper data. It is shown that a neural network trained with a set of border patterns may have a lower accuracy of learning but a significantly higher accuracy of generalization than one trained with a set of patterns drawn from the cores of the classes. Unfortunately, conventional training pattern selection and refinement procedures tend to favour core training patterns. For classification by a neural network, procedures which encourage the inclusion of border training patterns should be adopted as this may facilitate the production of an accurate classification. [ABSTRACT FROM AUTHOR]

The kappa coefficient is not an index of accuracy, indeed it is not an index of overall agreement but one of agreement beyond chance. Chance agreement is, however, irrelevant in an accuracy assessment and is anyway inappropriately modelled in the calculation of a kappa coefficient for typical remote sensing applications. The magnitude of a kappa coefficient is also difficult to interpret. Values that span the full range of widely used interpretation scales, indicating a level of agreement that equates to that estimated to arise from chance alone all the way through to almost perfect agreement, can be obtained from classifications that satisfy demanding accuracy targets (e.g. for a classification with overall accuracy of 95% the range of possible values of the kappa coefficient is −0.026 to 0.900). Comparisons of kappa coefficients are particularly challenging if the classes vary in their abundance (i.e. prevalence) as the magnitude of a kappa coefficient reflects not only agreement in labelling but also properties of the populations under study. It is shown that all of the arguments put forward for the use of the kappa coefficient in accuracy assessment are flawed and/or irrelevant as they apply equally to other, sometimes easier to calculate, measures of accuracy. Calls for the kappa coefficient to be abandoned from accuracy assessments should finally be heeded and researchers are encouraged to provide a set of simple measures and associated outputs such as estimates of per-class accuracy and the confusion matrix when assessing and comparing classification accuracy. • Kappa is not a measure of accuracy but of agreement beyond chance and chance correction is not needed. • All arguments for the use of kappa are flawed or apply equally to other measures. • Interpreting a kappa coefficient is difficult due especially to the effects of prevalence and bias. • A very accurate classification could be associated with a very wide range of kappa values. • Kappa should not be routinely used in accuracy assessment or comparison. [ABSTRACT FROM AUTHOR]

Spatio-temporal image fusion methods have become a popular means to produce remotely sensed data sets that have both fine spatial and temporal resolution. Accurate prediction of reflectance change is difficult, especially when the change is caused by both phenological change and land cover class changes. Although several spatio-temporal fusion methods such as the Flexible Spatiotemporal DAta Fusion (FSDAF) directly derive land cover phenological change information (such as endmember change) at different dates, the direct derivation of land cover class change information is challenging. In this paper, an enhanced FSDAF that incorporates sub-pixel class fraction change information (SFSDAF) is proposed. By directly deriving the sub-pixel land cover class fraction change information the proposed method allows accurate prediction even for heterogeneous regions that undergo a land cover class change. In particular, SFSDAF directly derives fine spatial resolution endmember change and class fraction change at the date of the observed image pair and the date of prediction, which can help identify image reflectance change resulting from different sources. SFSDAF predicts a fine resolution image at the time of acquisition of coarse resolution images using only one prior coarse and fine resolution image pair, and accommodates variations in reflectance due to both natural fluctuations in class spectral response (e.g. due to phenology) and land cover class change. The method is illustrated using degraded and real images and compared against three established spatio-temporal methods. The results show that the SFSDAF produced the least blurred images and the most accurate predictions of fine resolution reflectance values, especially for regions of heterogeneous landscape and regions that undergo some land cover class change. Consequently, the SFSDAF has considerable potential in monitoring Earth surface dynamics. • SFSDAF is proposed to fuse coarse and fine spatial resolution images. • SFSDAF derives endmember change to represent phenological change like FSDAF. • SFSDAF derives sub-pixel class fraction changes to accommodate land cover change. • SFSDAF accommodates the presence of mixed pixels at the fine resolution scale. • The combined effect from endmember change and class fraction change is analyzed. [ABSTRACT FROM AUTHOR]

Accuracy assessment and land cover mapping have been inexorably linked throughout the first 50 years of publication of Remote Sensing of Environment. The earliest developers of land-cover maps recognized the importance of evaluating the quality of their maps, and the methods and reporting format of these early accuracy assessments included features that would be familiar to practitioners today. Specifically, practitioners have consistently recognized the importance of obtaining high quality reference data to which the map is compared, the need for sampling to collect these reference data, and the role of an error matrix and accuracy measures derived from the error matrix to summarize the accuracy information. Over the past half century these techniques have undergone refinements to place accuracy assessment on a more scientifically credible footing. We describe the current status of accuracy assessment that has emerged from nearly 50 years of practice and identify opportunities for future advances. The article is organized by the three major components of accuracy assessment, the sampling design, response design, and analysis, focusing on good practice methodology that contributes to a rigorous, informative, and honest assessment. The long history of research and applications underlying the current practice of accuracy assessment has advanced the field to a mature state. However, documentation of accuracy assessment methods needs to be improved to enhance reproducibility and transparency, and improved methods are required to address new challenges created by advanced technology that has expanded the capacity to map land cover extensively in space and intensively in time. • Quality reference data, sampling, and error matrix remain core elements of protocol. • Map relevant analyses must reflect reality of study area mapped. • Probability sampling and design-based inference are critical to statistical rigor. • Reproducibility requires better documentation of accuracy assessment methods. • Imperfect reference data merit greater consideration in accuracy assessment. [ABSTRACT FROM AUTHOR]