The science
Chain of models. (1) Glycol left on the apron is washed to the drain: dM/dt = -kw i M with kw = 0.25 per mm of rain or melt. (2) The detention pond is a continuously stirred tank with first-order biodegradation: V dC/dt = W(t) - Q C - kp V C, throughflow from the rational method Q = c i A (c = 0.9, A = 25 ha). (3) The outfall is mixed into the stream and each water parcel travels at speed u, reacting by Streeter-Phelps: dL/dt = -kd L, dD/dt = kd L - kr D, where L is the remaining ultimate oxygen demand and D the oxygen deficit below saturation.
Closed form. D(t) = kdL0/(kr-kd) (e-kdt - e-krt) + D0 e-krt with the critical time tc = ln[(kr/kd)(1 - D0(kr-kd)/(kdL0))]/(kr-kd). The model advances parcels with the exact two-step propagator, so the profile equals this expression at every parcel.
Why it hurts. Propylene glycol is not toxic at these levels; the damage is oxygen. Complete oxidation C3H8O2 + 4 O2 gives 4×32/76.1 = 1.68 g O2 per g, so one tonne of glycol can consume 1.7 t of dissolved oxygen. Temperature acts three ways: warm water holds less oxygen (13.1 mg/L at 4 C, 9.1 at 20 C) and decays glycol faster (θ = 1.047 in the stream, 1.06 in the pond), which brings a deeper sag closer to the outfall; cold water slows the pond, so more glycol arrives in a longer pulse. In winter the real danger is often ice cover, which cuts reaeration kr (modelled here simply as a factor 0.2).
Honest simplifications. Rain, catchment and stream are constants; no dispersion (parcels do not mix with their neighbours, so peaks are sharper than in a real river); no sediment oxygen demand or algae; the pond is assumed saturated with oxygen and fully mixed; kd(20 C) = 0.35 per day and kp(20 C) = 0.3 per day are typical order-of-magnitude values, not measurements for a particular site. Airport ethylene or propylene glycol practice (Type I fluid, 50 % by volume, 300 L per aircraft here) varies widely; ice cover is represented only by cutting kr to 20 %. This is a teaching model, not a permit calculation.