Three-dimensional reactive transport simulations were undertaken to study the sorption and degradation dynamics of three herbicides in a shallow aerobic aquifer with spatially variable pH during a 216 days injection experiment. Sorption of two phenoxy acids [(+/-)-2-(4-chloro-2-methylphenoxy) propanoic acid] (MCPP) and [(+/-)-2-(2,4-dichlorophenoxy)propanoic acid] (dichlorprop) was found to be negligible. Degradation of the phenoxy acids was rapid after an initial lag phase. Degradation of the phenoxy acids could only be reproduced satisfactorily by growth-linked microbial degradation. The model fit to the field data was slightly improved if degradation was assumed to be influenced by the local pH that was observed to increase with depth ( approximately 4.5--5.7). In the observed pH-range the nitroaromatic herbicide [2-Methyl-4,6-dinitrophenol] (DNOC) was partly dissociated (pK(a)=4.31) and present in both the neutral and ionized form. The model simulations demonstrated that most of the observed spatial variation in sorption of DNOC could be explained by assuming that only the neutral form of DNOC was subject to sorption. A varying flow field was observed during the injection experiment and the model simulations documented that this most likely resulted in different migration paths for DNOC and the non-sorbing solutes. The model simulations indicated that degradation of DNOC was an important process. The degradation rate of DNOC remained constant over time and was simulated adequately by first-order kinetics. Again, the model fit to field observation was slightly improved if local pH was assumed to influence the degradation rate. Only the maximum utilization rate was estimated from the field data, while the remaining degradation parameters where successfully transferred from the laboratory study.