Jean-François Mas

Comparison of pixel‐based and object‐oriented image classification approaches—a case study in a coal fire area, Wuda, Inner Mongolia, China

Pixel‐based and object‐oriented classifications were tested for land‐cover mapping in a coal fire area. In pixel‐based classification a supervised Maximum Likelihood Classification (MLC) algorithm was utilized; in object‐oriented classification, a region‐growing multi‐resolution segmentation and a soft nearest neighbour classifier were used. The classification data was an ASTER image and the typical area extent of most land‐cover classes was greater than the image pixels (15 m). Classification results were compared in order to evaluate the suitability of the two classification techniques. The comparison was undertaken in a statistically rigorous way to provide an objective basis for comment and interpretation. Considering consistency, the same set of ground data was used for both classification results for accuracy assessment. Using the object‐oriented classification, the overall accuracy was higher than the accuracy obtained using the pixel‐based classification by 36.77%, and the user’s and producer’s accuracy of almost all the classes were also improved. In particular, the accuracy of (potential) surface coal fire areas mapping showed a marked increase. The potential surface coal fire areas were defined as areas covered by coal piles and coal wastes (dust), which are prone to be on fire, and in this context, indicated by the two land‐cover types ‘coal’ and ‘coal dust’. Taking into account the same test sites utilized, McNemar’s test was used to evaluate the statistical significance of the difference between the two methods. The differences in accuracy expressed in terms of proportions of correctly allocated pixels were statistically significant at the 0.1% level, which means that the thematic mapping result using object‐oriented image analysis approach gave a much higher accuracy than that obtained using the pixel‐based approach..

The application of artificial neural networks to the analysis of remotely sensed data

Artificial neural networks (ANNs) have become a popular tool in the analysis of remotely sensed data. Although significant progress has been made in image classification based upon neural networks, a number of issues remain to be resolved. This paper reviews remotely sensed data analysis with neural networks. First, we present an overview of the main concepts underlying ANNs, including the main architectures and learning algorithms. Then, the main tasks that involve ANNs in remote sensing are described. The limitations and crucial issues relating to the application of the neural network approach are discussed. A brief review of the implementation of ANNs in some of the most popular image processing software packages is presented. Finally, we discuss the application perspectives of neural networks in remote sensing image analysis.

Modelling deforestation using GIS and artificial neural networks

Abstract This study aims to predict the spatial distribution of tropical deforestation. Landsat images dated 1974, 1986 and 1991 were classified in order to generate digital deforestation maps which locate deforestation and forest persistence areas. The deforestation maps were overlaid with various spatial variables such as the proximity to roads and to settlements, forest fragmentation, elevation, slope and soil type to determine the relationship between deforestation and these explanatory variables. A multi-layer perceptron was trained in order to estimate the propensity to deforestation as a function of the explanatory variables and was used to develop deforestation risk assessment maps. The comparison of risk assessment map and actual deforestation indicates that the model was able to classify correctly 69% of the grid cells, for two categories: forest persistence versus deforestation. Artificial neural networks approach was found to have a great potential to predict land cover changes because it permits to develop complex, non-linear models.

Tropical Deforestation, Community Forests, and Protected Areas in the Maya Forest

Community forests and protected areas have each been proposed as strategies to stop deforestation. These management strategies should be regarded as hypotheses to be evaluated for their effectiveness in particular places. We evaluated the community-forestry hypothesis and the protected-area hypothesis in community forests with commercial timber production and strict protected areas in the Maya Forest of Guatemala and Mexico. From land-use and land cover change (LUCC) maps derived from satellite images, we compared deforestation in 19 community forests and 11 protected areas in both countries in varying periods from 1988 to 2005. Deforestation rates were higher in protected areas than in community forests, but the differences were not significant. An analysis of human presence showed similar deforestation rates in inhabited protected areas and recently inhabited community forests, but the differences were not significant. There was also no significant difference in deforestation between uninhabited protected areas, uninhabited community forests, and long-inhabited community forests. A logistic regression analysis indicated that the factors correlated with deforestation varied by country. Distance to human settlements, seasonal wetlands, and degree and length of human residence were significant in Guatemala, and distance to previous deforestation and tropical semideciduous forest were significant in Mexico. Varying contexts and especially colonization histories are highlighted as likely factors that influence different outcomes. Poorly governed protected areas perform no better as a conservation strategy than poorly governed community forests with recent colonists in active colonization fronts. Long-inhabited extractive communities perform as well as uninhabited strict protected areas under low colonization pressure. A review of costs and benefits suggests that community forests may generate more local income with lower costs. Small sample sizes may have limited the statistical power of our comparisons, but descriptive statistics on deforestation rates, logistic regression analyses, LUCC maps, data available on local economic impacts, and long-term ethnographic and action-research constitute a web of evidence supporting our conclusions. Long-inhabited community forest management for timber can be as effective as uninhabited parks at delivering long-term forest protection under certain circumstances and more effective at delivering local benefits.

Land use-cover change processes in highly biodiverse areas: the case of Oaxaca, Mexico

Abstract Land use-cover changes (LUCC) such as deforestation, have resulted as global warming and a reduction of environmental services, with large negative consequences for mankind. Effects based on statistics alone have not been sufficient enough to detect, stop and eventually revert negative LUCC processes that are strongly related to biodiversity loss. It is, therefore, of prime concern to assess and depict cartographically, major LUCC processes simultaneously. Mexico harbors a large pool of biodiversity, mostly restricted to a few locations among which, The State of Oaxaca plays a major role. In this state, nevertheless, drastic negative LUCC processes are taking place. Land cover types, mapped in previous surveys, overlaid on recent Landsat imagery and 300 ground truth sites, were used to detect current LUCC. Rates of conversion of the most important LUCC processes were computed and mapped simultaneously. Oaxaca has lost over half a million hectares of forested areas during the last 20 years. The core results may contribute to the understanding of how LUCC and GIS methods can provide better and more targeted information that may help to improve conservation policies and land use planning strategies.

Comparing two approaches to land use/cover change modeling and their implications for the assessment of biodiversity loss in a deciduous tropical forest

Land use/cover change (LUCC) modeling is an important approach to evaluating global biodiversity loss and is the topic of a wide range of research in ecology, geography and environmental social science. This paper reports on development and assessment of maps of change potential produced by two spatially explicit models and applied to a Tropical Deciduous Forest in western Mexico. The first model, DINAMICA EGO, uses the weights of evidence method which generates a map of change potential based on a set of explanatory variables and past trends involving some degree of expert knowledge. The second model, Land Change Modeler (LCM), is based upon neural networks. Both models were assessed through Relative Operating Characteristic and Difference in Potential. At the per transition level, we obtained better results using DINAMICA. However, when the per transition susceptibilities are combined to compose an overall change potential map, the map generated using LCM is more accurate because neural networks outputs are able to express the simultaneous change potential to various land cover types more adequately than individual probabilities obtained through the weights of evidence method. An analysis of the change potential obtained from both models, compared with observed deforestation and selected biodiversity indices (species richness, rarity, and biological value) showed that the prospective LUCC maps tended to identify locations with higher biodiversity levels as the most threatened areas as opposed to areas that had actually undergone deforestation. Overall however, the approximate assessment of biodiversity given by both models was more accurate than a random model.

Optimal region growing segmentation and its effect on classification accuracy

Image segmentation is a preliminary and critical step in object-based image classification. Its proper evaluation ensures that the best segmentation is used in image classification. In this article, image segmentations with nine different parameter settings were carried out with a multi-spectral Landsat imagery and the segmentation results were evaluated with an objective function that aims at maximizing homogeneity within segments and separability between neighbouring segments. The segmented images were classified into eight land-cover classes and the classifications were evaluated with independent ground data comprising 600 randomly distributed points. The accuracy assessment results presented similar distribution as that of the objective function values, that is segmentations with the highest objective function values also resulted in the highest classification accuracies. This result shows that image segmentation has a direct effect on the classification accuracy; the objective function not only worked on a single band image as proved by (Espindola, G.M., Camara, G., Reis, I.A., Bins, L.S. and Monteiro, A.M., 2006, Parameter selection for region-growing image segmentation algorithms using spatial autocorrelation. International Journal of Remote Sensing, 27, pp. 3035–3040.) but also on multi-spectral imagery as tested in this, and is indeed an effective way to determine the optimal segmentation parameters. McNemars test (z 2  =  10.27) shows that with the optimal segmentation, object-based classification achieved accuracy significantly higher than that of the pixel-based classification, with 99% significance level.

Evaluating drivers of land-use change and transition potential models in a complex landscape in Southern Mexico

Understanding and analysis of drivers of land-use and -cover change (LUCC) is a requisite to reduce and manage impacts and consequences of LUCC. The aim of the present study is to analyze drivers of LUCC in Southern Mexico and to see how these are used by different conceptual and methodological approaches for generating transition potential maps and how this influences the effectiveness to produce reliable LUCC models. Spatial factors were tested for their relation to main LUCC processes, and their importance as drivers for the periods 1993–2002 and 2002–2007 was evaluated by hierarchical partitioning analysis and logistic regression models. Tested variables included environmental and biophysical variables, location measures of infrastructure and of existing land use, fragmentation, and demographic and social variables. The most important factors show a marked persistence over time: deforestation is mainly driven by the distance of existing land uses; degradation and regeneration by the distance of existing disturbed forests. Nevertheless, the overall number of important factors decreases slightly for the second period. These drivers were used to produce transition potential maps calibrated with the 1993–2002 data by two different approaches: (1) weights of evidence (WoE) to represent the probabilities of dominant change processes, namely deforestation, forest degradation, and forest regeneration for temperate and tropical forests and (2) logistic RM that show the suitability regarding the different land-use and -cover (LUC) classes. Validation of the transition potential maps with the 2002–2007 data indicates a low precision with large differences between LUCC processes and methods. Areas of change evaluated by difference in potential showed that WoE produce transition potential maps that are more accurate for predicting LUCC than those produced with RM. Relative operating characteristic (ROC) statistics show that transition potential models based on RM do usually better predict areas of no change, but the difference is rather small. The poor performance of maps based on RM could be attributed to their too general representation of suitability for certain LUC classes when the goal is modeling complex LUCC and the LUC classes participate in several transitions. The application of a multimodel approach enables to better understand the relations of drivers to LUCC and the evaluation of model calibration based on spatial explanatory factors. This improved understanding of the capacity of LUCC models to produce accurate predictions is important for making better informed policy assessments and management recommendations to reduce deforestation.

Comparison of simulation models in terms of quantity and allocation of land change

Our article illustrates how to compare the outputs from models that simulate transitions among categories through time. We illustrate the concepts by comparing two land change models: Land Change Modeler and Cellular Automata Markov. We show how the modeling options influence the quantity and allocation of simulated transitions, and how to compare output maps from pairs of model runs with respect to a reference map of transitions during the validation interval. We recommend that the first step is to assess the quantity of each transition and to determine the cause of the variation in quantity among model runs. The second step is to assess the allocation of transitions and to determine the cause of the variation in allocation among model runs. The separation of quantity and allocation of the transitions is a helpful approach to communicate how models work and to describe pattern validation. We compare three runs of models that simulate transitions among land categories.Pattern validation compares a reference map of transition to maps from pairs of runs.Quantity and allocation are helpful concepts to describe models and to compare maps.Quantity refers to the size of each transition from one category to another category.Allocation refers to the spatial distribution of the transitions.

Change Estimates by Map Comparison: A Method to Reduce Erroneous Changes Due to Positional Error

A common method to assess land use/cover change (LUCC) is the comparison of digital maps of an area within a geographic information system (GIS). However, positional errors of the maps involved in the comparison affect this assessment and much of the change shown by means of this comparison may be an artifact due to these errors. This note presents a simple method to improve change estimates by detecting and correcting erroneous changes resulting from positional errors. It allows an important reduction of error in change area estimates and is likely to be useful in LUCC assessment studies.