A better physical understanding but also prediction of convective drying processes of fruit is essential for further process optimization. This study uses validated conjugate modeling to gain insight in how fruit drying kinetics are related to the convective heat and mass exchange with the surrounding turbulent airflow via the fruit surface. Conjugate modeling implies that the heat and mass transport in both air and fruit domains are solved simultaneously. We explore the impact of several model assumptions and different convective drying conditions. The conjugate model is inherently more accurate than the use of constant convective transfer coefficients (CTCs), so the non-conjugate approach. However the gain in accuracy was found to be limited in terms of overall fruit drying kinetics, such as total mass loss. Nevertheless, conjugate modeling allowed to identify spatial and temporal variability in CTCs, which locally affected drying rates and internal moisture content distribution. Thereby, we identified the occurrence of negative convective transfer coefficients, which led to rehydration at specific locations on the fruit surface, due to the surrounding high-humidity microclimate. The ability to identify the direct relation between non-uniformities in the airflow to those in the tissue is a unique trait of the conjugate approach. Furthermore, it was shown that isothermal modeling should not be used, even for near-isothermal conditions such as low-temperature drying, and that including thermal radiation exchange with the environment clearly affected the drying rates. Regarding the drying conditions, the impact of the air speed and approach flow temperature was found to be smaller compared to altering the approach flow humidity. When direct solar radiation was present, the presence of airflow provided significant cooling of the fruit, which is beneficial for preserving heat-sensitive nutritional compounds in the fruit, and also enhanced the drying rate. This study will aid drying technologists to define the required complexity of their model.