Impact dimensions are not independent
Energy, carbon, water and waste heat all descend from the same capacity figure, so they correlate almost perfectly by construction. The analytically interesting relationships are the weak ones — where a large footprint does not translate into local exposure.
Strong correlations between energy, carbon, water and waste heat are an artefact of the modelling chain, not an empirical discovery: all four are derived from capacity through fixed coefficients. Only relationships involving independently sourced variables — population, water stress, land use, confidence — carry information.
Annual water against IT capacity
Bubble size is IT capacity. Click a point to open the facility record.
Spearman rank correlations across all facilities
Computed in the source workbook over the full 1,098-record dataset, independent of the filters above.
Questions worth asking of this data
Location-based Scope 2 only, using national average grid factors. Scope 1 and Scope 3 are absent entirely; sub-national grid variation is not represented.
Every energy value is MODELLED (Capacity × 8,760 × Load Factor × PUE). No facility in this dataset is metered.
Waste heat is an energy-balance proxy (an annual average rejection RATE), never a measured local temperature rise. Cooling technology is inferred from the water ratio.
Water = Energy × WUE and covers ON-SITE cooling only; it excludes upstream water embodied in electricity generation. Water stress is an area-level band, not a coordinate-level lookup.
THE WEAKEST PILLAR IN THIS WORKBOOK. Noise is a capacity-based engineering proxy and setback is a land-use class band. Neither has been measured at any facility — use for screening only.
Zero documented complaints does not mean zero complaints occurred — it means no complaint record was located in the sources consulted.