GSI Chart — Geological Strength Index Calculator, Free

Why this chart can be drawn exactly

The original chart is a picture. Hoek and Marinos published it in 2000 as a drawing with hand-placed contours, and there is no equation in that paper — so any program claiming to reproduce it exactly is claiming to have measured a drawing.

What makes this page different is that a later paper put numbers on both axes. Hoek, Carter and Diederichs (2013) proposed GSI = 1.5 × JCond89 + RQD/2, where JCond89 is Bieniawski's 1989 joint-condition rating (0–30) and RQD is the familiar core-recovery index. That is a plane, and a plane is something that can be drawn exactly. So the contours here are the true straight lines of that equation, equally spaced by construction, and the descriptive rows and columns are bands on those two axes at the published boundaries — surface at 45/36/27/18/9/0, structure at 40/30/20/10/0.

The paper is explicit that the original hand-drawn contours were neither parallel nor equally spaced, and that the two readings differ by up to about four GSI points at the grid intersections. Four points is real, and it is smaller than the disagreement between two loggers — which is the whole argument for drawing the quantified version and saying which one you drew.

A consequence worth knowing: the published quantified chart is four rows by five columns, not six by five. Its authors delete the top and bottom rows of the descriptive chart on purpose. This page keeps the intact-or-massive row, drawn above the published axis with a line across the chart marking where the source stops and the extension begins, and every scenario landing there is flagged. Laminated and sheared rock is refused a position outright, because its structure axis is RQD and RQD cannot go below zero.

Two ways to put your data in

The chart route is GSI as it was meant to be used: type a structure and a surface condition, and get the cell. This is what you do standing at a face or reading a description in a factual report. The result is reported as the cell's own range — about ±9.5 GSI points, which is what a cell is — with its midpoint beside it.

The quantified route takes a JCond89 rating and an RQD, which is what a core log gives you, and applies the 2013 equation directly. It is the route to use when nobody is going to visit the outcrop. If you supply both a description and the numbers, the numbers win and the table says so, because a value computed from a log and a value read off a face are different claims about the same rock.

You can also just type a GSI you have already agreed, and the page will plot it on the contour it belongs to. That is not a fallback — putting the number your ground model already carries next to the descriptions it was supposed to come from is often the most useful thing this page does.

Where GSI stops working

The index assumes the rock mass behaves homogeneously and isotropically, which means it assumes there are enough blocks. Hoek and Brown's 2018 edition puts a number on that: a 10 m tunnel in rock jointed at half a metre contains something like four hundred blocks and qualifies; the same tunnel in rock jointed at two metres or wider contains fewer than twenty-five and GSI should not be used. The failure is not conservatism, it is category error — that excavation fails by wedges falling out of the roof, and wedge analysis is what answers it.

The same rule bites in slopes, and it bites at a scale people rarely check. A 100 m slope at 3 m joint spacing qualifies comfortably. Fifteen-metre benches cut in exactly the same rock mass do not. The rock did not change; the number of blocks in the failure volume did.

Three more limits, all from the sources: GSI should not be applied where a single dominant structural orientation controls the behaviour — undisturbed slate is the named counter-example — unless that set dips into the slope and cannot control failure. Above about GSI 65 the criterion may not apply as written, and brittle spalling and structurally controlled wedges want checking. And heterogeneous or tectonically deformed rock needs its own charts: flysch, ophiolites and molassic rocks all have published versions, and this general chart is not a substitute for them.

The RMR correlation, and why it is labelled the way it is

The commonest question about GSI is how to get one from an RMR you already have. The published answer is GSI = RMR89′ − 5, from Hoek, Kaiser and Bawden (1995), and the prime matters: RMR89′ is your rating RE-SCORED with the groundwater rating set to 15 (completely dry) and the joint-orientation adjustment set to zero. Neither water nor the orientation of your tunnel relative to the joints is a property of the rock mass, and GSI is. This page does that re-scoring rather than pretending the raw total will do, and prints the primed number it used.

The relation is stated "for RMR89′ > 23", which reads like a calibration limit and is not one. Twenty-three is the arithmetic floor of the primed rating — the worst the stepwise tables can score on the four parameters that vary is 8, and priming then adds a fixed 15. So the condition says only "above the very bottom of the scale", which is exactly why the same authors give a different relation below it, in terms of Q rather than RMR.

And the caveat that matters more than either: Hoek's own current text says the RMR-to-GSI correlation "has proved to be unreliable, particularly for poor quality rock masses" and recommends estimating GSI directly from the chart instead. This site has a GSI chart. It is the primary tool; the correlation is the historical cross-check, and it is labelled that way everywhere it appears.

Who uses this

Tunnel engineers converting a face log into the GSI that a Hoek-Brown envelope, a ground reaction curve or a numerical model starts from. Engineering geologists writing the ground model that everything downstream is argued from, who need the estimate and its range in the same figure. Mining geotechnicians comparing structural domains along a drive. Students and lecturers, who need the chart visible and its provenance cited rather than buried in a licensed binary.

It is a characterisation tool, not a design suite. What it gives you is a defensible, correctly cited GSI with an honest range on it — and everything runs in this browser tab, so nothing about your project goes anywhere.

Common questions

Is this free? What is the catch?

Free, with no account, no trial timer and no gated options. Enter your own scenarios, change every setting, export the figure as SVG or PNG. Free exports carry a small watermark, as they do everywhere on this site. There is no catch — this is a published chart and an equation, and the page exists partly because it is the kind of thing a course links to.

Which edition of the chart is this?

The structure and surface classes are the Hoek & Marinos (2000) chart as reproduced unchanged in Hoek & Brown (2019), JRMGE 11(3), 445–463. The contours and the axis scales are Hoek, Carter & Diederichs (2013), ARMA 13-672, "Quantification of the Geological Strength Index chart". Worth being precise about one thing: the 2019 paper does NOT contain a quantification equation — its figure is the 2000 chart as drawn, and it endorses quantification only by citation. Both papers are named in the figure footer.

Why does my scenario plot as a bar rather than a dot?

Because a dot would be a lie about the precision. A cell of this chart spans about ±9.5 GSI points, the chart's own header asks for a range rather than a value, and the difference between two competent loggers is routinely five to ten points. The bar runs along the direction in which GSI changes fastest, so its length is exactly the range in the results table — the picture and the number are the same statement. Turn it off in the options if you need a clean figure for a report that quotes the range in text.

Why will it not accept laminated or sheared rock?

Because the quantified chart has no position for it. Its structure axis is RQD/2, running from 40 at the top down to 0, and a row below "disintegrated" would need a negative RQD. The 2013 paper deletes that row for exactly this reason, and also deletes the intact/massive row at the other end because sparsely jointed rock does not satisfy the homogeneity the index assumes. This page keeps the massive row as a marked extension, since the arithmetic is still exact there, and refuses the sheared one, since it is not. For tectonically deformed and heterogeneous rock masses, use the Marinos & Hoek flysch charts instead.

What is JCond89, and where do I get one?

It is Bieniawski's 1989 joint-condition parameter — the fourth of the five RMR ratings — scored 0 to 30. You can take it straight from the descriptive table (30 for very rough, discontinuous, unweathered surfaces; 0 for soft gouge over 5 mm), or build it from the guideline sub-table by adding five sub-ratings of 0 to 6 each for persistence, aperture, roughness, infilling and weathering. One caution from the 1989 original: where infilling is present it overshadows roughness, and you should score from the descriptive table rather than summing the sub-ratings.

Can I plot more than three scenarios?

Yes. Add rows and each gets its own symbol and colour, and the results table prints up to twelve in full. The most useful figure is usually a small number — one row per structural domain along a drive, or the same domain described by two different loggers, which is a remarkably effective way to show a client what the range on the page actually means.

Does anything leave my browser?

No. The whole calculation runs in this tab, there is no account and there is no upload. For a ground model on somebody else's licence area, that is usually the entire question.