A water utility needs a managed graph for sensor bronze and silver tables with expectations attached. What should they create?

Creating a Lakeflow Spark Declarative Pipeline gives a managed graph for municipal IoT refine paths: declare sensor bronze and silver tables, wire dependencies, and attach expectations in one designed pipeline. Water-utility telemetry that needs governed bronze landing plus silver cleanses with quality rules fits that model far better than ad-hoc statements or unmanaged folders. The pipeline definition is the operational artifact operators schedule, monitor, and evolve. A Databricks Job with a notebook that defines the sensor bronze/silver graph and attaches expectations misattributes pipeline capability — a Job orchestrates tasks, it is not a declared table graph with attached expectations, so quality rules would remain hand-coded in the notebook rather than enforced and reported as a managed unit. A Databricks SQL warehouse with a view that defines the sensor bronze/silver graph and attaches expectations misattributes pipeline capability — a warehouse provides serving compute and a view is only saved query logic, not the declarative pipeline construct that defines bronze/silver dependencies and attaches expectations. A Lakeflow Connect ingestion pipeline that defines the sensor bronze/silver graph and attaches expectations misattributes pipeline capability — Connect is a connector for getting source data in, not the managed graph-with-expectations construct itself, so it cannot define bronze-to-silver dependencies or attach quality expectations. Exam caveat: declarative pipelines excel at table graphs and expectations; irregular one-off sensor backfills may still use jobs/notebooks alongside the always-on pipeline. Operational check: deploy a declarative pipeline with bronze/silver sensor tables and at least one expectation, trigger a run, confirm tables update in Unity Catalog, and verify expectation results appear in the pipeline quality UI.