Soil degradation is a complex process that includes erosion, salinisation, acidification, compaction, loss of organic matter and pollution. Identifying degradation risks is crucial for sustainable land management and ecosystem productivity. Bayesian networks integrate multiple data sources and uncertainties and provide probabilistic modelling to aid decision-making. Bayesian networks are graphical models that represent relationships between variables and calculate the probability of outcomes such as soil degradation. Variables in this context include climate data (e.g. precipitation, temperature), soil properties (e.g. texture, organic matter), topography and human activities (e.g. agriculture, deforestation). Bayesian networks use Bayes’ theorem to calculate the probability of soil degradation based on the available data. The first step in creating a Bayesian network to assess soil degradation is to identify the most important risk factors. These factors range from climatic conditions such as precipitation and temperature to soil properties such as texture, organic matter and pH. Topographical features such as slope and altitude also play a role. Human activities such as agricultural practises and land development have a significant impact on degradation risks. Once these factors are identified, the network is structured by linking the variables together in such a way that their causal relationships are expressed. The nodes stand for the variables, the edges for the connections between them. For example, extreme rainfall can have a direct impact on erosion, while soil acidification could be linked to the use of fertilisers. This structure illustrates how risk factors interact and contribute to soil degradation. Next, the network is populated with conditional probabilities based on data. For example, the precipitation and erosion data indicate how often certain amounts of precipitation lead to erosion. These conditional probabilities allow the Bayesian network to calculate the overall risk of soil degradation. One of the biggest advantages of Bayesian networks is their ability to update predictions as new data becomes available. For example, if new measurements of soil organic matter are available, the network can refine its predictions about the probability of degradation. In practical applications, Bayesian networks can predict the risks of soil degradation in real time using multivariate data such as climate models, satellite images and soil sensors. These networks enable predictive modelling of processes such as erosion or changes in soil fertility due to changes in agricultural practises. Their results help decision-makers, such as land managers and farmers, to make informed decisions about sustainable land management. The risk assessments produced by the network can support decisions on the adaptation of agricultural techniques, the implementation of erosion control measures or the reduction of irrigation intensity. By integrating Bayesian networks with geographic information systems (GIS), spatial maps of land degradation risks can also be created, helping to visualise areas at risk and enable more targeted interventions. However, the application of Bayesian networks is associated with challenges, particularly in terms of data quality. The accuracy of predictions depends on the availability of reliable and complete data. 30 Inaccurate or incomplete data can lead to erroneous conclusions. To mitigate this, the combination of different data sources — such as satellite images, field measurements and historical records — is crucial. Another challenge is scaling these models as the volume of environmental data grows. Combining Bayesian networks with other AI techniques, such as deep learning, could improve their performance. In addition, the development of systems that automatically update the network with new data is a promising area for future development. To summarize, Bayesian networks are a valuable tool for land degradation risk assessment based on multivariate data. Their ability to integrate different data sources, account for uncertainties and update predictions in real time makes them particularly well suited for sustainable land management. As climate change and human activities increasingly affect soil health, Bayesian networks enable predictive analyses and informed decision-making, contributing to global efforts to preserve ecosystems and soil fertility.