It is evident that temperature inside culverts is not equal to, nor a function of outside temperature. A clear variance in temperatures exists in culverts. Two temperature distribution scenarios occur in culverts depending on whether the ends of the culverts are closed (by snow) or open. Additionally, differences in top and bottom temperatures inside culverts have been recorded. Temperature distribution inside the culverts will lead to uneven frost depth under the culvert. 

Air flow inside the culvert can significantly affect temperature distribution inside the culvert. The effect of air velocity on temperature distribution can be seen in graphs starting from velocities of 1 m/s. However, nonlinear regression models suggest that the effect can become noticeable at 0.5 m/s. Air velocity inside culverts is dependent on angle of the culvert in relation to the predominant wind direction of the area and topography rather than size of the culvert. It is considered likely that one-directional air flow in culverts will result in a deeper frost penetration at the inlet than the outlet.

In the case of the instrumented culverts, the temperature distribution is more affected by external factors such as air temperature, wind, snow cover and vegetation rather than the diameter of the culvert. However, temperature distribution inside the culverts is affected by the cover depth. 

It is possible for the temperatures inside the culvert to be almost equal to outside temperature under specific conditions, as recorded in the 3.4 m culvert. In this culvert the air flow is bi-directional, it is made of steel, its entrances are never obstructed by snow and it has low cover depth. It is not clear if temperatures inside the culvert would remain equal to the outside temperature if one of the previously mentioned conditions were altered. 

Data collection from the three instrumented culverts will continue until 2023. To have a better overview of the temperature distribution inside the whole culvert, the culverts will be re-instrumented in the autumn of 2021. Thermocouples will be added to the previously uninstrumented end of the culvert, while thermocouples at CS2 will be removed. Anemometers will remain it their original position. 

Harsh climate conditions of northern Sweden cannot be applied to all regions with seasonally frozen ground. For this research to be more widely applicable, generalization that can be applied to areas with lower freezing temperatures and snow cover must be made. Collected data from the initial freezing period could potentially be used to model full length winters at a more southern location. 

Computational fluid dynamics (CFD) studies would help to gain a better comprehension of how air flow develops in culverts. A 3D CFD study allows for a detailed investigation of how combined influence of embankment geometry, wind velocity and direction as well as angle and diameter of the culvert result in air flow inside the culvert. The goal of the CFD investigation would be to predict air velocity in any circular culvert if wind velocity and direction is known. 

Finally, convective heat transfer boundary should be numerically implemented inside the culvert to fully assess the effect of air velocity on frost depth. Initially, results collected from the physical experiment should be replicated after which instrumented culverts in the field can be modelled for verification. Additionally, insulating effects of soil cover and snow must be quantified and added to the model.