In the last decades, measurement while drilling, or MWD, technology has set foot in the drill and blast tunnelling industry. Penetration rate, thrust, torque pressure, percussive pressure, rotation speed, water flow and water pressure are registered on a centimetre scale and processed to parameters more dependent on geology. These parameters could ideally be used to adapt the blast design for a more efficient blast and to predict the amount of rock support needed. In reality, MWD technology is often only used to document geology. The main reasons for this could be that the workflow is not fully adapted to the MWD technology, the knowledge about MWD is not sufficient or the ability of MWD to represent the geology has not been extensively investigated and verified for drill and blast tunnelling. MWD has great potential since it is a relatively cheap and simple method of collecting a larger amount of data, which does not interfere with the workflow in drill and blast tunnelling. The aim is to investigate the ability to detect discontinuities and their geometrical properties in MWD data and to evaluate the usability of MWD data in terms of detecting discontinuities. The first of five objectives is a literature review on the subjects of MWD technology, discontinuities in rock masses, previous research and statistical data analysis methods. Raw MWD parameters are processed and filtered by software, which can somewhat be seen as eliminating influences such as depth dependency and percussive pressure. This results in modified penetration rate, torque or water pressure which represents rock mass properties. Previous research showed the inability to compare MWD data and available geological reporting about fractures was due to a large difference in scale. For statistical analysis of MWD data in this study 4 methods are considered. Principle component analysis and k-means cluster analyses are forms of unsupervised learning and can potentially help to understand and reduce complexity of multivariate datasets. Linear and logistic regressions are forms of supervised learning that can give insight into predictability and potentially predict the presence discontinuities. The second objective is to gather data, which is appropriate and detailed enough to study relations between MWD data and discontinuities. Two tunnels under construction are visited where limestone and highly foliated phyllite are the dominant rock types. These are the Solbakktunnelen and Bjørnegårdtunnelen. While not disturbing the construction work, different methods of detailed geological mapping are used. Blasthole remains used as scanlines for mapping discontinuities and mapping discontinuities in the contour and face led to a dataset reflecting the geological situation in tunnels. Another method for mapping discontinuities, which is not influenced as much by blasting, is borehole inspection. The use of an inspection camera and an optical televiewer resulted in 11 video footages and 25 detailed recordings of 5 meter long boreholes. The third objective is to evaluate the data by a visual comparisons of geological data and MWD data. Comparing the mapped geology and 3D images of MWD data showed that fractures with a certain infill or aperture are visible in MWD data. The more detailed geological data showed that an open fracture or a fracture with soft infill and an aperture wider than 1cm often leads to a peak in penetration rate, rotation pressure, processed penetration and processed rotation pressure. The fourth objective is to apply statistical methods to come to a more in depth understanding of the relation between MWD data and discontinuities and confirm findings in objective 3. Since it is suspected that only the actual location and aperture can be predicted, a vector is made by assigning the number 1 to each MWD sampling depth between the upper and lower boundary of a discontinuity. For intact rock, a 0 is assigned to each MWD sampling depth. The principle component analyses showed that the dataset could be reduced to a manageable number of 5 to 7 components and that around 80% of the variability was retained. K-means cluster analyses is found to be an appropriate analysis for 2 out of 4 datasets. It led to the understanding that responses in MWD data due to lithological discontinuities without an aperture cannot be separated from intact rock. Training the data with logistic regression analyses confirmed this finding. Logistic regression did however differentiate MWD data of open fractures from other MWD data in half of the collected data. 2 out of 3 fractures were predicted, but the exact location and size of predicted fractures differ slightly from the real location. The regression for the fractures with soft infill was successful for a quarter of the data. 6 out of 9 fractures with clayey infill were predicted, again with a small deviation in size. Testing the regression equations on test datasets, which were not part of the input for data training, did not lead to the correct prediction of fractures. The last objective is to discuss the current and future usability of MWD data for the Norwegian Public Roads Administration. Even though the statistical analyses did not fully succeed in separating responses of fractures in MWD data, visually responses are found to be characteristic. The fact that not all discontinuities give a distinctive response, makes calculations of rock quality designation unreliable. Therefore, in terms of rock support, MWD data can only assist in decision making concerning spot bolting to secure wedges due to large fractures. This research might contribute to help understand and predict grout volumes. The use of 3D images of MWD data could give a better understanding of the in situ fracture structure. Ideally, this knowledge can be used to anticipate possible under- and overbreak due to these fractures before blasting. Considering the findings in this study, it is still presumed that MWD technology has great potential even if it might not lead to the prediction of each type of discontinuity.