Predicting bare soil temperature. II. Experimental testing of multi-day models

Two agroclimatic models described earlier are tested against measurements of temperature at the soil surface, T₀, and in bulk soil, T(z,t), at an experimental site near Aberdeen during spring, summer and autumn. Soil temperature is predicted most accurately in spring and autumn, when the constant-windspeed model ENBL1 predicts the mean within 0.3 K of measurement. Thedynamic-windspeed model ENBL2 predicts within 0.1 K in all intervals, and with error<0.3 K to depth z=32 cm in spring and autumn. A sensitivity analysis of the output diurnal wave, T₀(t), shows that maximum input economy requires only time-dependent shortwave radiation, R_s(t), and air temperature, T_a(t), other inputs being constant. Two or three harmonics in ENBL1 should suffice, requiring four or six measurements per day, depending mainly on the time-structure of the primary driving function R_s.