Using two GEOBIA (Geographical Object Based Image Analysis) algorithms on a set of segmented images compared to grid partitioning at different scales, we show that statistical metrics related to both objects and sets of pixels are (more or less) subject to the Modifiable Areal Unit Problem. Subsequently, even in a same spatial partition, there may be a bias in statistics describing the objects due to some size effect of the pixel samples. For instance, pixels homogeneity based on Grey Level Cooccurrence Matrices (GLCM), Landscape Shape Index, entropy, object compacity, perimeter/area ratio are studied according to scale. The approach consists in studying the behavior of a given statistical metrics through scales and to compare the results on several image segmentations, according to different partitioning processes, from GEOBIA (Baatz & Schäpe algorithm and Self Organizing Maps) or using reference grids. We finally discuss about the relationship between GEOBIA metrics and scale. By analysing object shape and pixels composition from different metrics points of views, we show that GEOBIA does not always mitigate the Modifiable Areal Unit Problem.

More than 72,000 hectares of western Tasmania were burnt in 2016 due to bushfires. Bushfires in Tasmania has high social, economical, and environmental impacts. The remote delineation of these bushfires has paramount importance for decision-making authorities to help people in emergencies and planning. Considering the fact that delineation uncertainty from Earth Observation [EO] data is inevitable, this study uses MODIS, Landsat and Sentinel-2 imageries covering the 2016 burnt areas from Tasmania. We test the hypothesis that the difference in Normalised Difference Vegetation Index (NDVI) before and after the fire event can detect the accurate delineation of burnt areas and hence the changes. MODIS, Landsat and Sentinel-2 products before and after fire are used independently in delineating and mapping bushfire boundaries. We map in three thematic classes burnt, damaged and both. Delineated boundaries are examined for uncertainty and error maps are produced. The uncertainty examination and validation are performed using ground truth data obtained from local fire authorities. Developed error metrics are used to obtain statistical measures like sensitivity, specificity, positive predictive value, negative predictive value, kappa coefficient and overall accuracy. Our results show that there is minimal difference in overall accuracy from both the sensors MODIS: [0.94 vs 0.92] and Sentinel [0.94 vs 0.93] for the classes burnt & damaged vs only burnt. Furthermore, we propose a conceptual framework for bushfire mapping uncertainty in a multiple-scale environment incorporating sensitive thematic parameters that could affect initiation of fire and blaze direction.

Using the R language as a GIS applied on forest fire data in South of France, the goal of the research is to emphasize how spatial statistics may depend on the areal units chosen. First, we propose to map the forest fire data at different scale levels based on administrative boundaries. Second, we measure the MAUP by showing scale sensitivity in descriptive statistics and in regression analyses. Finally, although many tools can be used for vector or raster data aggregation and mapping, we discuss why we choose R as a primary analysis tool and R added-value.

No abstract