Ground: isotropic, constant properties; homogeneous or horizontally layered. The effective model neglects the vertical heat flow between layers and the uneven heat transfer per layer; the segmented model resolves the second but lets heat that crosses an intermediate layer see only the two end layers (2.2 % against finite volumes at 50 years for a strong three-layer contrast). The undisturbed temperature assumes a steady, one-dimensional geothermal flux. No freezing of the ground, no temperature-dependent properties.
Groundwater: uniform horizontal flow, constant in time, without dispersion or vertical flow; aquifer layers horizontal and infinite; the pair velocity between two layers is the mean Darcy flux, checked only by bounds. Groundwater flow together with inclined boreholes is refused (the moving line source is implemented for vertical boreholes).
Geometry: tilt up to 85°; boreholes are lines (axes closer than the borehole radius are treated as apart; closer than 1 m gives a warning). The optimizer enumerates vertical, equal-length shapes; a given custom layout is sized, not optimized.
Boundary condition: UBWT assumes one wall temperature for the whole field, UHTR one heat rate, UFT one mean fluid temperature with local resistances; real fields lie in between. Networks do not yet change the g-function: the string temperatures are read-outs and the field g-function still assumes a uniform wall temperature (no axial fluid profile per string). The ground surface is isothermal at .
Borehole resistance: steady inside the borehole unless the short-time model is selected. That model represents the capacities by an equivalent single pipe (coaxial: one well-mixed fluid node), is blended out at 100 h, and in a network computes once, with the combined and the flow of the longest loop, not per string flow. Transit delay and flow schedules need a transient model along the depth. The multipole model assumes a circular borehole with pipes placed symmetrically at the nominal shank spacing.
Short times: the line source underestimates below about (a few hours) unless the CHS correction is enabled; below is linear in .
Fluid and hydraulics: properties at one temperature for the whole simulation; friction losses in the boreholes and a minor-loss coefficient per string; no header pipes.
Loads: monthly mode represents the load by monthly means plus one peak per month and direction at the month's end (the classic monthly method, conservative); the energy of the peaks is not added to the balance.
Sizing: one scale for all boreholes (equal lengths, or proportional for unequal fields); the root is the smallest feasible H found by bracketing, which may miss a second feasible window in non-monotonic cases.
Studies and Monte Carlo: the confidence of a percentile refers only to the inputs declared uncertain, with the distributions given; model error (line source, effective layers, monthly load representation) is not included. Sizing runs that cannot meet the limits within the sizing range count with the maximum length.
Thermal response tests: the confidence intervals assume independent residuals and are optimistic (the residuals of real tests have lag-1 autocorrelations of 0.93–0.99; the ΔBIC rule uses the effective number of observations). The hollow cylinder source reads λ 4–9 % below the line source over usual test durations by its physics (no heat capacity of the borehole core); the line source is the reference model and the CHS or FLS is chosen only when it fits significantly better. The time constant of the term (1 h) matters on short tests. M3 cannot tell groundwater flow from a bounded test domain and gives no flow direction; the distributed (DTS) evaluation and the dynamic-model method M4 are not available yet.
Site plans: geometry only; a shape that fits only at an unaligned position in a narrow gap can be missed by the placement of layout types (the fills still cover such sites). DXF: ASCII only; dimensions, 3D faces and multilines are skipped.
Classic-method compatibility: the compatibility options reproduce the classic monthly conventions, not the classic g-function library; results still differ from the classic method where its pre-computed g-functions differ from the finite line source (about 0.15 K on the base temperatures of the Linden worked example). Layers, groundwater, inclined or unequal boreholes, sections, UFT, explicit networks and the short-time model have no equivalent in the classic method and are flagged as such.