Experimental evidence of randomness and nonuniqueness in unsaturated outflow experiments designed for hydraulic parameter estimation

Single transient outflow experiments are commonly conducted for inverse estimation of unsaturated hydraulic parameters. We assess the validity of this procedure through several repeated experiments on the same sample, a medium sand contained in a pressure cell. Outflow was induced by one or multiple step changes in bottom boundary suction, such that there were replicates for each of several step levels in suction. We observed that experiments with small initial step changes were poorly reproducible, even though our setup allowed reproduction of almost identical initial saturation for each run. Experiments with large step changes were well reproducible, but the out flow response was virtually the same for different step levels. Neither type of observation was predicted by a theoretical sensitivity analysis of the Richards equation, given the minimal inaccuracy in our experiments. Inverse estimation yielded incompatible apparent hydraulic parameters for different flow conditions. Our results imply experimental limitations of the inversion procedure.