Like many places around the world, the wildland–urban interface areas surrounding urban regions are subject to variable levels of fire risk, threatening the natural habitats they contact. This risk has been assessed by various authors using many different methods and numerical models. Among these approaches, machine learning models have been successfully applied to determine the weights of criteria in risk assessment and risk prediction studies. In Istanbul, data have been collected for areas that are yet to be urbanized but are foreseen to be at risk using geographic information systems (GIS) and remote sensing technologies based on fires that occurred between 2000 and 2021. Here, the land use/land cover (LULC) characteristics of the region were examined, and machine learning techniques, including random forest (RF), extreme gradient boosting (XGB), and light gradient boosting (LGB) models, were applied to classify the factors that affect fires. The RF model yielded the best results, with an accuracy of 0.70, an F1 score of 0.71, and an area under the curve (AUC) value of 0.76. In the RF model, the grouping between factors that initiate fires and factors that influence the spread of fires was distinct, and this distinction was also somewhat observable in the other two models. Risk scores were generated through the multiplication of the variable importance values of the factors and their respective layer values, culminating in a risk map for the region. The distribution of risk is in alignment with the number of fires that have previously occurred, and the risk in wildland–urban interface areas was found to be significantly higher than the risk in wildland areas alone.

The fires encountered in the buildings and facilities of industrial areas make up quite a small proportion of all fire cases. However, in terms of the heat release rate, size of the burned area, damage, and impact zone, such fires have a large impact as compared to other fire cases. All fires have risk in terms of propagation. However, since fires in industrial structures and plants have rather high levels, qualitatively and quantitatively, compared to residential fires and other types of fires, it should be regarded as necessary to research them extensively. In this study, fires that have broken out in various places around Turkey, such as in factories, cold storage plants, and manufacturing facilities, were investigated. We aimed to determine the level of risk of the occurrence of these fires in the environment. A large amount of detailed information gathered about these fires was analyzed. This information includes data about the causes of the fires, deformation data of various materials, data on technical problems, data on financial damage levels, and data on fire patterns. We found 40 of these fire cases in total, and the case data were used to calculate the risk scores with the Geographic Information System (GIS), Analytical Hierarchy Process (AHP), and Inverse Distance Weight (IDW) methods. For each fire case, places sensitive to damage and losses were assessed according to six main criteria. Buffer analysis maps were generated for the 40 fire cases, and the cases were ordered based on their overall risk scores. In this ordering, fire case number 21 was found in the riskiest region, and fire cases 32, 17, and 31 were found in the low-risk region. Fire case number 21 occurred in a factory used for manufacturing fabric. This factory works with high volumes of acrylic, polyester, and other raw materials. In addition, there are some structures in very close proximity. It was observed that fire cases could be well differentiated depending on the selected criteria in the model applied.

As in many parts of the world, rural-urban, forest-urban interface areas surrounding urban regions expose the natural areas they interact with to a threat of fire risk that can reach various sizes. This risk has been assessed for various regions of the world using many different methods and numerical models so far. Among these, it is seen that machine learning models have successful applications in risk assessment and risk prediction studies. For the fire risk prediction of Istanbul's yet unurbanized regions, but where the city is anticipated to potentially shift, data was collected using the opportunities provided by Geographic Information Systems and Remote Sensing technologies based on fires that occurred between 2000-2021, and the region was examined. Machine learning methods' Random Forest (RF), Extreme Gradient Boosting (XGB), and Light Gradient Boosting (LGB) models were applied for the classification of factors effective in fire. The best result was given by the RF model with 0.93 accuracy, 0.062 F1 score, and 0.753 Area Under Curve (AUC) value. In the classification of factors in the RF model, the grouping between fire-initiating factors and factors effective in spreading is evident, while this distinction is partially noticeable in the other two models.