RQD from Volumetric Joint Count

Sometimes there is no core. You are standing in a trench, at a bench face, or on an outcrop, and you need an RQD to feed into a classification that will not accept anything else. Arild Palmström's answer, published at the 1982 Lisbon symposium on engineering geology for underground construction, is to count the joints in a unit volume of rock and correlate: $RQD = 115 - 3.3\,J_v$.

The volumetric joint count $J_v$ is the number of joints crossing one cubic metre of the mass. You do not count them one by one. Take each joint SET in turn, measure its mean spacing, and $J_v$ is the sum of the reciprocals of those spacings. Three sets at 0.5 m, 0.3 m and 1.0 m give $2 + 3.3 + 1 = 6.3$ joints per cubic metre. Random joints belonging to no recognisable set are conventionally added at about a fifth each.

The relation is a straight line fitted through field data, and it has the two boundaries a straight line always has. Below about $J_v = 4.5$ it returns more than 100%, which is not a possible RQD; the rule is to take RQD as 100 and accept that the index has saturated. Above about $J_v = 35$ it returns a negative number; take RQD as zero. Both ends are the correlation telling you that RQD has run out of discrimination, not that the answer is extreme.

Treat the result as an estimate of a measurement, and carry that distinction into whatever you write down. The scatter about the fitted line is roughly ±10 RQD points, which is enough to move a mass a whole band on Deere's scale. Where a report needs to distinguish 68% from 74%, this correlation cannot do it and should not be asked to.

The counting errors that matter are systematic rather than random, and both of them make the rock look better than it is. A joint set running nearly parallel to the face you can see is under-counted, because it barely crosses the window you are looking through — a set dipping into the face at 10° might cross your two-metre exposure twice while a set at 80° crosses it fifteen times, and the two might have identical spacings. The standard correction is to divide the observed spacing by the sine of the angle between the set and your scanline, and it matters enormously at shallow angles. The second is blast damage: fractures behind a freshly shot face are not natural joints, they were made last week, and counting them turns your own excavation method into a property of the rock mass.

Run backwards, from a logged RQD to an implied joint count, this becomes a genuinely useful sanity check. If the core says 40% and the face you can walk up to has blocks half a metre across, the two disagree, and it is usually the core that is wrong — poor drilling and rough handling break rock that was intact in the ground.