Over the last few decades, Casablanca city became the biggest industrial, commercial center in Morocco with rapid urbanization and explosive population growth, more than 4 million people. Urban expansion has reached to suburban areas due to population growth and socio economic development, not to mention the rapid increase of transportation. Result of these changes causes a change of microclimate in urban areas. The most evident phenomenon is the increase of urban surface temperature as compared with suburban areas, "heat island" is formed in the atmospheric boundary above urban area. It could make serious environmental problems for its inhabitants (e.g., urban waterlogged and thermal pollution). Thermal infrared remote sensing bands, proved its capability in monitoring temperature field. The purpose of this study is to evaluate the use of Landsat TM, ETM+, OLI and TIRS data for indicating temperature differences in urban areas, in order to achieve a spatiotemporal study, using data between 1984 and 2014, and showing the relationship between urban expansion and the heat island effect during time, producing maps that shows the distribution of urban temperature. Results can be combined with land use/ land cover maps or thermal-land cover and operated as reference for urban planning and future solutions to reduce heat island effect.

In this work, we present a hybrid classification technique combining an expert system and an object-oriented approach. The expert system allows the integration of a knowledge base built through a series of deductive rules, that will guide the classification whose primitives requires informations on the highest level and will be represented by semantic objects, not pixels. Instead of the original bands only, other derived data combining textural, spectral information and shapes, are included in the classification process. The result is then combined with an expert system whose rules use variables such as vegetation index (NDVI), shading of building objects and other indicators. In conclusion, this approach has allowed us to improve the accuracy of the feature extraction method by extracting objects like, roads, trees, grass, bare soil and shadow on a very high-resolution image of the city of Rabat.