Data for "Learning Limitations: Machine learning, value of information, and water distribution asset management"

ANONYMIZED PIPE CONDITION ASSESSMENT DATASET============================================ An anonymized dataset of water distribution pipes used for a condition assessment study. Each row is a single pipe segment, paired with the condition rating recorded at its inspection. All directly identifying information has been removed or relabeled (see ANONYMIZATION below), so the data can be shared without exposing the source utilities, pipe locations, or asset IDs. FILE---- condition_assessment_anonymized.csv - one row per pipe segment. COLUMNS------- region Anonymized water utility / service region (U_A...U_H). Categorical. pipe_ID Anonymized pipe identifier; the letter denotes the region (e.g. A_1). String. diameter Nominal pipe diameter. Numeric. Units: mm. material_type Pipe material class (e.g. cast iron, ductile iron, PVC, ...). Categorical. installation_date Year the pipe was installed. Integer (calendar year). inspection_date Year the pipe was inspected. Integer (calendar year). age Pipe age at inspection (inspection_date minus installation_date). Integer. Units: years. condition_rating Assessed condition rating at inspection. Numeric/ordinal. Green/amber/red mapping to good/fair/poor condition. length Pipe segment length. Numeric. Units: m. Note on the date fields: installation_date and inspection_date are calendar years (integers), not full day/month/year dates, despite the "_date" suffix. age is their difference in whole years. ANONYMIZATION------------- - Regions relabeled to U_A ... U_H (the mapping to real utilities is not published). - Pipe IDs replaced with <region-letter>_<n> (e.g. A_1, A_2, ...), numbered within each region. Original asset IDs are not recoverable. - Spatial coordinates (endpoint X/Y) removed entirely. - Criticality, internal condition scores, and unit-of-measure bookkeeping columns removed. INCLUSION CRITERIA------------------ The dataset was filtered from the source records as follows: - Only pipes with diameter > 50 mm (i.e. distribution mains, excluding small service connections). - Plastic pipes (PVC, PE, HDPE, MDPE, ABS, and related grades) recorded as installed on or before 1960 were excluded, as such early install dates are implausible for these materials and indicate data-quality issues. Non-plastic pipes were retained regardless of install year. ORIGINAL SOURCES---------------- City of Courtenay (2024). "OpenData WaterMains." [Feature layer]. Retrieved from https://data-courtenay.opendata.arcgis.com. Accessed in December 2024. City of Kitchener (2024). "Water Mains." [Feature layer]. Retrieved from https://open-kitchenergis.opendata.arcgis.com. Accessed in December 2024. City of Prince George (2024). "Water Pressurized Main." [Feature layer]. Retrieved from https://data-cityofpg.opendata.arcgis.com. Accessed in December 2024. Kapiti Coast District Council (2021). "KCDC Water Service Pipes." [Feature layer]. Retrieved from https://data-kcdc.opendata.arcgis.com. Accessed in December 2024. Matamata-Piako District Council (2024). "Waterline." [Feature layer]. Retrieved from https://data-matamatapiakodc.hub.arcgis.com. Accessed in December 2024. Region of Waterloo (2024). "Mains (Water)." [Feature layer]. Retrieved from https://open-kitchenergis.opendata.arcgis.com. Accessed in December 2024. Waimakariri District Council (2022). "Water Pipes." [Feature layer]. Retrieved from https://openmaps-waimakariri.hub.arcgis.com. Accessed in December 2024. Wellington Water (2024). "Regional Water Pipes." [Feature layer]. Retrieved from https://data-wellingtonwater.opendata.arcgis.com. Accessed in December 2024.