Consider a scenario where you are managing a data pipeline that involves processing large datasets in Azure Synapse Analytics. During execution, you notice that some operations are causing data spills to disk, significantly slowing down the process. Describe the steps you would take to identify the causes of these data spills and how you would optimize the pipeline to minimize or eliminate them.