European Journal of Sustainable Agroecosystems

Climate change is exacerbating the incidence and seriousness of interrelated risks related to agriculture such as drought, flooding, heat stress, frost, and pest and disease outbreaks, posing a danger to food security and crop productivity in areas prone to climate hazards. Traditional methods of forecasting natural hazards give a poor account of the cumulative and cascading effects of hazards. The authors create a theoretical framework of risk prediction supported by artificial intelligence that can assist in decision-making in agriculture while considering climate change adaptation. The framework brings together satellite and remote sensing information, data from ground meteorological and Internet of Things (IoT) sensor networks, information from soil and agricultural surveys, and archives of previous natural disasters within one data integration layer. Machine learning and deep learning algorithms were used to develop the risk assessment model and get layered forms of prediction in cooperation with the explanatory module explaining which features influenced the decisions. Area under the curve and receiver operating characteristic are based on statistical and validation assessments which include use of variance analysis, correlation analysis and principal components analysis. The stacked ensemble method produced the best and most reliable results across all groups of hazards. In addition, analysis of features showed that the most significant predictors were soil moisture, land surface temperature, vegetation indexes and standard precipitation index. As a result, spatial risk classification provided us with information about zones of low, moderate, high and very high risk which made it possible to implement early warning and adaptive management programs for hazards. In conclusion it should be said that this research showed how modern and innovative technological solutions based on artificial intelligence can help in making a risk and sustainability management system more robust and increase the chances to address issues related to adaptation to climate changes.