Accuracy Assessment Procedure Research Articles

Postfire Forest Regrowth Algorithm Using Tasseled-Cap-Retrieved Indices

Wildfires are a common disturbance factor worldwide, especially over the last decade due to global climate change. Monitoring postfire forest regrowth provides fundamental information needed to enhance the management and support of ecosystem recovery after fires. The purpose of this study is to propose an algorithm for postfire forest regrowth monitoring using tasseled-cap-derived indices. A complex approach is used for its implementation, for which a model is developed based on three components—Disturbance Index (DI), Vector of Instantaneous Condition (VIC), and Direction Angle (DA). The final product—postfire regrowth (PFIR)—allows for a quantitative assessment of the intensity of regrowth. The proposed methodology is based on the linear orthogonal transformation of multispectral satellite images—tasseled cap transformation (TCT)—that increases the degree of identification of the three main components that change during a fire—soil, vegetation, and water/moisture—and implies a higher accuracy of the assessments. The results provide a thematic raster representing the intensity of the regrowth classes, which are defined after the PFIR threshold values are determined (HRI—high regrowth intensity; MRI—moderate regrowth intensity; and LRI—low regrowth intensity). The accuracy assessment procedure is conducted using very-high-resolution (VHR) aerial and satellite data from World View (WV) sensors, as well as multispectral Sentinel 2A images. Three different forest test sites affected by fire in Bulgaria are examined. The results show that the classified thematic raster maps are distinguished by a good performance in monitoring the regrowth dynamics, with an average overall accuracy of 62.1% for all three test sites, ranging from 73.9% to 48.4% for the individual forests.

Depremin Arazi Kullanım ve Arazi Örtüsü (AKAÖ) Üzerine Etkilerinin Google Earth Engine (GEE), Sentinel-2 Görüntüleri ve Makine Öğrenmesi Kullanılarak Değerlendirilmesi: Antakya Örneği

Natural disasters, especially earthquakes, known to be the most devastating process that threating human life, ecosystems, and land properties including land use and land cover (LULC). Understanding of such changes may help for rehabilitation processes, as well as presentation of baseline to develop management strategies for further steps. Remote sensing technologies have long been used for determination of change directions and magnitudes after earthquakes while development in cloud-based platforms provided users to avoid issues in storage and processing costs, effectively. In present study, it was aimed to determine LULC changes occurred around Antakya city of Hatay after February 06, 2023 and February 20, 2023 earthquakes, which caused serious losses. Moreover changes within 5 km zone from central coordinates were also investigated by considering individual subzones with 1 km width. One of the most widely used machine learning algorithm, random forest (RF), was used classify Sentinel-2 imageries via Google Earth Engine (GEE) platform. Accuracy assessment procedures were implemented to determine reliabilities of LULC2022 and LULC2023, and accuracies were found over 0.85. Investigation of overall changes have revealed that areas of forest (F) and cultivated fields (CF) were considerably decreased while concrete (C), natural vegetation (N) and water (W) areas have increased. Dispersal of collapse buildings resulted in increase of C class not only at city level, but also within each subzone of 5 km buffer zone. Classification of Sentinel-2 imageries through RF algorithm in GEE provided rapid and reliable results for determining changes in Antakya, whereby periodically monitoring of further changes strongly suggested.

Vejetasyon İndeksleri, Ana Bileşenler Analizi ve Google Earth Engine Kullanılarak Tarımsal Alan Sınıflandırması: Söke/Aydın Örneği

Land use and land cover (LULC) classification is known to be one of the most widely used indicators of environmental change and degradation all over the world. There are various algorithms and methods for LULC classification, whereby reliability of the classification maps presents the principal concern. The study focused on evaluation of accuracies of LULC maps produced from original bands of Sentinel-2 imageries together with Normalized Difference Vegetation Index (NDVI), Green NDVI (GNDVI), and Principal Component Analysis (PCA) using Google Earth Engine (GEE) platform to identify best enhancing method for agricultural land classification. Moreover, short-term LULC changes aimed to be identified in the specified area. To achieve the aims, all available imageries acquired in the same month of different years with less than 10% cloud contamination were used to compose averaged images for May 2018 and May 2022 for generating LULC2018 and LULC2022 maps. The area has separated into seven main classes, namely, olive (O), perennial cultivation (P), non-perennial cultivation (NP), forest (F), natural vegetation (N), settled area-bare land (S), and water surface (W) via random forest algorithym. Reliabilities of LULC maps were evaluated through accuracy assessment procedures considering stratified randomized control points. Transitions between each LULC classes were identified.

A novel framework to assess all-round performances of spatiotemporal fusion models

Spatiotemporal data fusion, as a feasible and low-cost solution for producing time-series satellite images with both high spatial and temporal resolution, has undergone rapid development over the past two decades with more than one hundred spatiotemporal fusion methods developed. Accuracy assessment of fused images is crucial for users to select appropriate methods for real-world applications. However, commonly used assessment metrics do not comprehensively cover multiple aspects of spatiotemporal fused image quality, contain redundant information, and are not comparable across different study areas. To address these problems, this study proposed a novel framework to assess all-round performances of spatiotemporal fusion methods. Four accuracy metrics, including RMSE, AD, Edge, and local binary patterns (LBP), were selected as the optimal set of assessment metrics according to the assessment criteria. These metrics not only quantify the spectral and spatial information in the fused images but also greatly alleviate information redundancy and feature computational simplicity. Furthermore, inspired by Taylor diagrams, we designed an all-round performance assessment (APA) diagram to provide a visual tool for a comprehensive assessment of the performance of spatiotemporal fusion methods, supporting cross-comparison of different spatiotemporal fusion methods by considering the effects of input data and land surface characteristics. The case study in three typical sites demonstrated that the proposed framework can better differentiate the performances of six spatiotemporal fusion methods. This new framework can promote the cross-comparison of different spatiotemporal fusion methods and guide users to select suitable methods for real-world applications, as well as facilitate the establishment of a standard accuracy assessment procedure for spatiotemporal fusion methods.

Impact of Blood Glucose Monitoring System Accuracy on Clinical Decision Making for Diabetes Management.

The accuracy of blood glucose monitoring systems (BGMS) is crucial for the safe and effective management of diabetes mellitus. Despite standardization of accuracy assessment procedures and requirements, various studies have shown that the accuracy of BGMS on the market can vary considerably. This article therefore provides health care professionals and users with an intuitive illustration of the impact of BGMS accuracy on clinical decision making. Several hypothetical patient scenarios based on blood glucose (BG) levels in the low, normal, and high BG range are devised. Using data from a recent BGMS accuracy study, a method for calculating the expected range of BG readings from four examined BGMS at the selected BG levels is introduced. Based on these ranges, it is illustrated how clinical decisions and subsequent outcomes of the hypothetical patients are affected by the expected inaccuracies of the BGMS. The range of expected BGMS readings for the same true BG level can vary considerably between different BGMS. The discussion of hypothetical patient scenarios revealed that the use of some BGMS could be associated with an increased risk of adverse events such as failure to detect hypoglycemia, driving with an unsafe BG level, delay of treatment intervention in diabetes during pregnancy, or the failure to prevent diabetic ketoacidosis. This article can support both health care professionals and patients to understand the impact of BGMS accuracy in a relatable, clinical context. Furthermore, it is suggested that current accuracy requirements might be insufficient for the prevention of adverse clinical outcomes in certain circumstances.

ASSESSING THE ACCURACY OF LAND USE LAND COVER (LULC) MAPS USING CLASS PROPORTIONS IN THE REFERENCE DATA

Abstract. Traditionally the accuracy assessment of a hard raster-based land use land cover (LULC) map uses a reference data set that contains one LULC class per pixel, which is the class that has the largest area in each pixel. However, when mixed pixels exist in the reference data, this is a simplification of reality that has implications for both the accuracy assessment and subsequent applications of LULC maps, such as area estimation. This paper demonstrates how the use of class proportions in the reference data set can be used easily within regular accuracy assessment procedures and how the use of class proportions can affect the final accuracy assessment. Using the CORINE land cover map (CLC) and the more detailed Urban Atlas (UA), two accuracy assessments of the raster version of CLC were undertaken using UA as the reference and considering for each pixel: (i) the class proportions retained from the UA; and (ii) the class with the majority area. The results show that for the study area and the classes considered here, all accuracy indices decrease when the class proportions are considered in the reference database, achieving a maximum difference of 16% between the two approaches. This demonstrates that if the UA is considered as representing reality, then the true accuracy of CLC is lower than the value obtained when using the reference data set that assigns only one class to each pixel. Arguments for and against using class proportions in reference data sets are then provided and discussed.

An effective accuracy assessment indicator for credible land use change modelling: Insights from hypothetical and real landscape analyses

Land use change (LUC) modelling has been widely used to inform landscape planning and adaptive management practices. The validation of LUC modelling results is critical for justifying the usability of LUC models. Along this line, global-level accuracy assessments with the kappa and error matrix approach are accepted as the common method for the validation of LUC models. However, high global-level accuracy does not always guarantee good model performance and high accuracy in characterizing the local LUC. It is necessary to develop indicators to exactly assess the accuracy of land use modelling in characterizing the detailed land use change. In this study, both hypothetical and real landscapes are used to analyze the differences between global-level and local accuracy assessments, and all possible simulation scenarios are considered in the hypothetical landscape by exhaustive methods. The results derived from the hypothetical landscape show that the local accuracy tends to increase with the increase in the proportion of the regional area showing land use transitions throughout the landscape. A real landscape simulation by the Dyna-CLUE model in the middle reaches of the Heihe River Basin (M-HRB) in China also showed a similar trend, where the land use transition from 2000 to 2015 accounted for approximately 10% of the total area. The simulation results showed high global-level accuracies of 97.17% (2010) and 85.01% (2015). The local accuracies for regions with little land use transition were 98.45% (2010) and 96.56% (2015), but the local accuracies for regions with significant land use transitions were only 0.99% and 6.08% in 2010 and 2015, respectively. According to the global-level accuracy, LUC simulations are reliable, but they do not correctly reflect the local changes in regions with significant land use transitions. Therefore, both global-level accuracy and local accuracy should be used to avoid possible misleading LUC modelling results. An effective simulation and accuracy assessment procedure was proposed in this study to increase the credibility of LUC modelling.

EXTENDING ACCURACY ASSESSMENT PROCEDURES OF GLOBAL COVERAGE LAND COVER MAPS THROUGH SPATIAL ASSOCIATION ANALYSIS

Abstract. High-resolution land cover maps are in high demand for many environmental applications. Yet, the information they provide is uncertain unless the accuracy of these maps is known. Therefore, accuracy assessment should be an integral part of land cover map production as a way of ensuring reliable products. The traditional accuracy metrics like Overall Accuracy and Producer’s and User’s accuracies – based on the confusion matrix – are useful to understand global accuracy of the map, but they do not provide insight into the possible nature or source of the errors. The idea behind this work is to complement traditional accuracy metrics with the analysis of error spatial patterns. The aim is to discover errors underlying features which can be later employed to improve the traditional accuracy assessment. The designed procedure is applied to the accuracy assessment of the GlobeLand30 global land cover map for the Lombardy Region (Northern Italy) by means of comparison with the DUSAF regional land cover map. Traditional accuracy assessment quantified the classification accuracies of the map. Indeed, critical errors were pointed out and further analyses on their spatial patterns were performed by means of the Moran’s I indicator. Additionally, visual exploration of the spatial patterns was performed. This allowed describing possible sources of errors. Both software and analysis strategies were described in detail to facilitate future improvement and replication of the procedure. The results of the exploratory experiments are critically discussed in relation to the benefits that they potentially introduce into the traditional accuracy assessment procedure.

Effects of climate and land cover changes on habitat for herbivores at Mole National Park, Ghana.

The aim of this study was to determine climate and land cover changes in the MNP and the effect these could have on the large herbivores. Monthly temperature and rainfall data and Landsat satellite imagery from 1980 to 2013 for the Mole National Park (MNP) were acquired. Land cover classes of the imageries were validated through ground truth accuracy assessment procedure. Autoregressive integrated moving average (ARIMA) model was used to model the data using time series statistics with R. The results indicated a rise in temperature that became significant after two decades and a reduction in the amount of rainfall every decade. The decade mean temperature for the period 2000–2010 (2000s) reached the highest of 27.2 ± 0.15 0C while the 1980s recorded the least of 26.7 ± 0.25 0C. The decade mean rainfall for the 1980s reached the highest of 129 ± 9.91 mm while the least of 93.9 ± 7.41 mm was recorded in the 2000s. There was a strong and significant negative correlation between the mean rainfall time series and the mean temperature time series for the period under review. The shrub-land occupied the largest part of the Park (about 26%) followed by the open savanna (23%). Rainfall influenced the open savanna, grass and shrub classes immediately, but the closed savanna lagged behind rainfall by five years. As forecast suggested a relatively high temperature and low rainfall, the shrub-land and grassland are likely to shrink while open and closed savanna woodland areas expand by 2020. Accordingly, there may be shrinking forage resources for the grazers and abundant forage for the browsers during the period. It is recommended to reduce the extent of burning at MNP during the forecast period to make forage sufficiently available for grazing. Regular follow up to this study could provide a guideline to securing habitats, and forage, for MNP’s herbivores.

Scale Issues Related to the Accuracy Assessment of Land Use/Land Cover Maps Produced Using Multi-Resolution Data: Comments on “The Improvement of Land Cover Classification by Thermal Remote Sensing”. Remote Sens. 2015, 7(7), 8368–8390

Much remote sensing (RS) research focuses on fusing, i.e., combining, multi-resolution/multi-sensor imagery for land use/land cover (LULC) classification. In relation to this topic, Sun and Schulz [1] recently found that a combination of visible-to-near infrared (VNIR; 30 m spatial resolution) and thermal infrared (TIR; 100–120 m spatial resolution) Landsat data led to more accurate LULC classification. They also found that using multi-temporal TIR data alone for classification resulted in comparable (and in some cases higher) classification accuracies to the use of multi-temporal VNIR data, which contrasts with the findings of other recent research [2]. This discrepancy, and the generally very high LULC accuracies achieved by Sun and Schulz (up to 99.2% overall accuracy for a combined VNIR/TIR classification result), can likely be explained by their use of an accuracy assessment procedure which does not take into account the multi-resolution nature of the data. Sun and Schulz used 10-fold cross-validation for accuracy assessment, which is not necessarily inappropriate for RS accuracy assessment in general. However, here it is shown that the typical pixel-based cross-validation approach results in non-independent training and validation data sets when the lower spatial resolution TIR images are used for classification, which causes classification accuracy to be overestimated.

Geographic stacking: Decision fusion to increase global land cover map accuracy

Techniques to combine multiple classifier outputs is an established sub-discipline in data mining, referred to as “stacking,” “ensemble classification,” or “meta-learning.” Here we describe how stacking of geographically allocated classifications can create a map composite of higher accuracy than any of the individual classifiers. We used both voting algorithms and trainable classifiers with a set of validation data to combine individual land cover maps. We describe the generality of this setup in terms of existing algorithms and accuracy assessment procedures. This method has the advantage of not requiring posterior probabilities or level of support for predicted class labels. We demonstrate the technique using Landsat based, 30-meter land cover maps, the highest resolution, globally available product of this kind. We used globally distributed validation samples to composite the maps and compute accuracy. We show that geographic stacking can improve individual map accuracy by up to 6.6%. The voting methods can also achieve higher accuracy than the best of the input classifications. Accuracies from different classifiers, input data, and output type are compared. The results are illustrated on a Landsat scene in California, USA. The compositing technique described here has broad applicability in remote sensing based map production and geographic classification.

CLOSE-RANGE PHOTOGRAMMETRY ENABLES DOCUMENTATION OF ENVIRONMENT-INDUCED DEFORMATION OF ARCHITECTURAL HERITAGE

Deformation, damage and permanent loss of heritage assets due to various physical and environmental factors has always been a major problem. As the availability of funds for conservation and restoration is limited, the digital documentation of heritage objects and monitoring of environment-induced deformations are increasingly important for cultural heritage preservation. Our study elucidates developments in the digital image capturing and processing for recording architectural heritage objects focusing on the digital camera calibration, close-range imaging, and photogrammetric modelling of complex structures using image matching techniques. A particular consideration in this paper is given to the ortho-photographic image compiling and accuracy assessment procedure. The practicality of the methodology is demonstrated by applying photogrammetric system PhotoMod for documentation of decorative elements in Uzutrakis manor, a national heritage site in Trakai, Lithuania.

The relationship among counselor experience level, empathic accuracy, and counseling outcome in the early phase of counseling

The current study examined a mediating effect of empathic accuracy on the relationship between counselor experience level and counseling outcome. Data were collected from 48 counselor–client dyads in real-life counseling settings. Empathic accuracy and counseling outcome were assessed by using client perceptions of the first three audiotaped counseling sessions. Ickes’ standard empathic accuracy assessment procedure was used with modifications to assess empathic accuracy. The procedure to assess empathic evaluation included three steps: (a) clients recognize and write down their thoughts and feelings while listening to the audiotapes, (b) counselors infer their client’s thoughts and feelings, (c) The third raters rate the statements of counselors’ empathic accuracy. Correlational analysis revealed that counselor experience level, empathic accuracy, and counseling outcome were positively related to one another. A path analysis was used to test the model that counselor experience level affects counseling outcome through the effect of empathic accuracy. Both paths from counselor experience level to empathic accuracy and from empathic accuracy to counseling outcome were significant, along with a significant mediating effect of empathic accuracy. These results show enhancing empathic accuracy is critical for counseling outcome. Limitations of the current study and suggestions for future research were discussed.

PRELIMINARY STUDY FOR THE IMPLEMENTATION OFAN IMAGE ANALYSIS ALGORITHM TO DETECT DAIRY COW PRESENCE AT THE FEED BARRIER

The objective of this study was to investigate the applicability of the Viola-Jones algorithm for continuous detection of the feeding behaviour of dairy cows housed in an open free-stall barn. A methodology was proposed in order to train, test and validate the classifier. A lower number of positive and negative images than those used by Viola and Jones were required during the training. The testing produced the following results: hit rate of about 97.85%, missed rate of about 2.15%, and false positive rate of about 0.67%. The validation was carried out by an accuracy assessment procedure which required the time-consuming work of an operator who labelled the true position of the cows within the barn and their behaviours. The accuracy assessment revealed that among the 715 frames about 90.63% contained only true positives, whereas about 9.37% were affected by underestimation, i.e., contained also one or two false negatives. False positives occurred only in 2.93% of the analyzed frames. Though a moderate mismatch between the testing and the validation performances was registered, the results obtained revealed the adequacy of the Viola-Jones algorithm for detecting the feeding behaviour of dairy cows housed in open free-stall barns. This, in turn, opens up opportunities for an automatic analysis of cow behaviour.

PRELIMINARY STUDY FOR THE IMPLEMENTATION OFAN IMAGE ANALYSIS ALGORITHM TO DETECT DAIRY COW PRESENCE AT THE FEED BARRIER

The objective of this study was to investigate the applicability of the Viola-Jones algorithm for continuous detection of the feeding behaviour of dairy cows housed in an open free-stall barn. A methodology was proposed in order to train, test and validate the classifier. A lower number of positive and negative images than those used by Viola and Jones were required during the training. The testing produced the following results: hit rate of about 97.85%, missed rate of about 2.15%, and false positive rate of about 0.67%. The validation was carried out by an accuracy assessment procedure which required the time-consuming work of an operator who labelled the true position of the cows within the barn and their behaviours. The accuracy assessment revealed that among the 715 frames about 90.63% contained only true positives, whereas about 9.37% were affected by underestimation, i.e., contained also one or two false negatives. False positives occurred only in 2.93% of the analyzed frames. Though a moderate mismatch between the testing and the validation performances was registered, the results obtained revealed the adequacy of the Viola-Jones algorithm for detecting the feeding behaviour of dairy cows housed in open free-stall barns. This, in turn, opens up opportunities for an automatic analysis of cow behaviour.

Mapping land cover changes in Mexico, 1976-2000 and applications for guiding environmental management policy

Land use conversion typically implicates deforestation and fragmentation of primary land cover types, which invariably translates into impoverishment of both natural and cultural capital. Understanding where conversion is taking place crucially underpins sound environmental policy instruments to prevent these enormous social and economic costs. This paper examines 30 years of semi-detailed (1:250 000) land cover mapping in Mexico. Pre-existing analogue databases describing land cover patterns in the 1970s and 1990s were reviewed, corrected, reorganized and transformed into a digital format. Current land cover patterns were depicted by conducting updated reinterpretation on Landsat ETM+ imagery. Digital cartographic overlaying was performed and the results were used to construct a spatially explicit land use/land cover change (LULCC) database with an additional accuracy assessment procedure. The value of the results of this analysis is also seen in the light of their direct applications for identifying critical watershed trends, for guiding the allocation of financial funds for sound land use planning and for assessing the effectiveness of established protected areas. This effort highlights the importance of new and more effective geographical approaches to depict, understand and contribute to informed measures to mitigate ongoing negative trends in land cover and climatic changes.

Benthic Mapping Using Sonar, Video Transects, and an Innovative Approach to Accuracy Assessment: A Characterization of Bottom Features in the Georgia Bight

Benthic maps provide the spatial framework for many science and management activities in coastal areas such as identification and protection of fish distributions and associated habitat as well as for monitoring changes in benthos and fish communities. To meet this need at Gray's Reef National Marine Sanctuary off the Georgia, U.S.A., coast, we created fine-scale benthic maps by visual interpretation of sonar imagery within a geographic information system. The major bottom types in the sanctuary—flat sand, rippled sand, hard bottom that is sparsely colonized with sessile invertebrates, and densely colonized hard bottom—were delineated through combined analysis of backscatter from side-scan sonar, bathymetry from multibeam sonar, scuba surveys, and video transects. Maps showed that unconsolidated sediments cover 75% of the bottom of this region; 8% occurs as flat sand plains with obvious burrowing and reworking of surface material by mobile benthic invertebrates, whereas 67% occurs as rippled sand without such fauna. The rest of the sanctuary consists of limestone bottom in two types of formations; either flat, sparsely colonized regions (25% of the sanctuary's total area) or as vertical ledges that are densely colonized with a diverse fauna of sessile invertebrates (<1%). Despite their limited area, these 0.5–2-m-tall ledge features harbor the majority of the sanctuary's biodiversity and biomass of both sessile invertebrates as well as ichthyofauna. A modified accuracy assessment procedure was used to account for spatial autocorrelation in the validation data and to separate thematic from positional accuracy. Overall thematic accuracy of maps is 95% for those areas of the map in which thematic accuracy and positional accuracy could be separated (87% of the mapped area). This fine-scale characterization provides a benthic inventory for a marine sanctuary and novel methods for mapping using sonar and accuracy assessment using transects.

Multitemporal land-cover classification of Mexico using Landsat MSS imagery

A complete land-cover classification of Mexico was performed using Landsat Multi-Spectral Scanner (MSS) imagery corresponding to years 1974, 1986 and 1990 ( - 1 y). The categorization of the approximately 2 M km 2 geographical region included the classification of approximately 300 equivalent scene images. Vegetation experts throughout the country provided an initial 250-class inventory of major vegetation associations by applying an unsupervised classification approach. A final regrouping was performed to produce a generalized thematic product containing 12 classes to provide a consistent national scale product. Classification accuracies were evaluated for each scene by means of cartographic comparison using two independently developed reference datasets corresponding to the 1970s and 1990s. An automated evaluation procedure was developed that incorporated decision rules to duplicate the results obtained using a manual accuracy assessment procedure. Overlaying both the image and the digital cartographic information allowed for the comparison of randomly selected pixels within each image scene. An overall accuracy for the three epochs of 62% was obtained for the 300 image scenes. Study results have provided a historical baseline documenting vegetation extent and distribution across Mexico over the two-decade period. This study serves as a possible model for subsequent North American land-cover characterization efforts.

Land cover mapping in a tropical hillsides environment: A case study in the Cauca region of Colombia

Results are presented from a study of land cover mapping undertaken in a tropical hillsides environment. The study area is located in the foothills of the Cordillera Central of Colombia, where a conventional maximum likelihood classification was performed upon Landsat TM imagery. A comprehensive accuracy assessment procedure performed on the resultant land cover map suggested that relatively low rates of classification accuracy were achieved. However, reported accuracy levels were found to vary substantially, depending on the specific methodology used to generate them. This suggests that caution is needed when making comparisons between classification accuracy figures reported by different workers, unless their methodologies are also clearly identified. It is further argued that a low accuracy land cover map still makes a valuable contribution to our knowledge of this hitherto little studied environment, provided that its limitations are understood and respected.

Land cover inventory in the Netherlands using remote sensing; application in a soil and groundwater vulnerability assessment system

Abstract Information on actual land cover is necessary for various applications, such as soil and groundwater protection studies and hydrological studies. Therefore, a decision to produce a national land cover data base of the Netherlands, using satellite images, was made. The first version of the data base is now available for the whole of the Netherlands. Prior to the supervised classification the area was stratified in more or less homogeneous areas. Because cost, time and logistics required for a random sampling of the entire country were prohibitive, a mixed quantitative/qualitative classification accuracy assessment procedure was proposed. Classification performances were quantitatively assessed by comparing the classification results with digitized ground reference maps using a Geographical Information System (GIS). This offers a flexible method for locating the incorrectly labelled pixels and determining the possible reasons thereof. Classification accuracy of the land cover types which do not cha...