Reed v Department of Natural Resources and Mines & Ors No. 2 [2014] QLC 6

APPEARANCES: D Yarrow instructed by p&e Law for the appellant

Reed v Department of Natural Resources and Mines & Ors No. 2 [2014] QLC 6 | Case summary | Zoe

[105]

[19] At the hearing of the appeal an expert report, exhibit 1, and a reply report, exhibit 2, were tendered on behalf of the appellant. These reports were prepared by Dr Scott Rayburg, who also gave oral evidence. Dr Rayburg is a senior lecturer in water resources engineering at Swinburne University of Technology in Victoria. His expertise is in fluvial geomorphology and hydrology and since 2004 he has published more than 35 peer-reviewed scientific publications in the fields of geomorphology, hydrology and hydraulics and been the co or lead investigator on research projects and consultancies.

[20] Dr Rayburg expressed the opinion that the hydraulic modelling employed by Brisbane Stormwater Management is invalid and therefore the results of that modelling are invalid.[6] While also critical of the hydraulic model used, "Sobek", his evidence was that the particular model used was a good deal less important than the fact that the modelling had been carried out and the results relied upon when there had not been accurate input and calibration data used. The required data, in the most ideal case, would include measures of flow level and discharge, velocity, shear stress and stream power for at least three flow levels, low, medium and high. However, some validity could be given to the model by data on water levels and velocity from two or three low to intermediate flow events.[7] This stream is ephemeral in nature so will only flow during and soon after rain events which could occur at any time of the day or night. When not flowing it will be dry with perhaps a series of ponds. In recognition of the practical difficulties and no doubt the danger of attempting to measure a stream on short notice in view of the obstacles presented potentially by darkness and swift flow, Dr Rayburg was of the view that some useful calibrations could be achieved with data from the observations made in at least two flow events in the range of up to what could be expected to occur in an event with an ARI of one in two years or less. The thrust of his evidence was that the precise details of what should be collected was not a useful focus of attention. His point was that no flow data at all had been collected so no calibration of the model had been done. In the circumstances, his view was that it would have been better not to model at all rather than to potentially have a modelling outcome which was conveying an unrealistic impression of accuracy.[8] Dr Rayburg has visited the site and did not collect any stream flow data at the time as he maintained that was not his role. He does not contend for a different result from modelling, but rather that the result relied upon is unreliable. The focus of Dr Rayburg's evidence became the use of a measure called "Manning's n", a measure of friction or resistance to flow. The degree of meandering, the type and size of obstructions and the form of the streambed are elements of resistance included in Manning's n.[9] Without having collected data to arrive at Manning's n, the applied value used in this case cannot be relied upon, invalidating the result of the modelling. Using the modelling without a disclosure of the extent of possible error and a sensitivity analysis showing what effect varying one element can have on the output of the model has resulted, in his opinion, in unsubstantiated claims being made.[10] Concern expressed by Dr Rayburg in his initial report concerning the reliability of topographical data were not pursued at the hearing. His view was that sensitivity analysis, validation, calibration and error assessment are fundamentals and none were conducted, invalidating the output of the modelling.[11] The model is highly sensitive to Manning's n and small errors in the accurate determination of this value will result in significant underestimations of flow velocity, shear stress and stream power which have the potential to threaten the stability of the diversion and the creek above and below it.[12] I note that Table 1 in the guideline sets out upper limits to parameters to which regard must be had when designing a diversion. No criticism was made in the appeal of the actual design. The stream power and shear stress relate to the effect on the bed and banks of the flow in the stream.

[21] Dr Rayburg, as a scientist, could not identify a single paper in the published scientific literature over the last 30 years where a similar hydraulic model had been used without first calibrating it.[13] He describes as a useful comparison the work of the International Panel on Climate Change which has conducted global climate modelling where the most sophisticated models in the world are calibrated and validated against data from thousands of sites, many of which have data covering 100 years or more.[14] The accuracy and uncertainty of the model is disclosed.[15] Such error reporting is standard practice.[16] Dr Rayburg is of the opinion that the modelling performed in this case does not meet the minimum criteria for best practice for hydraulic modelling.[17] This is not a case where the relevant data was existing but was not used, it is a case where the data was not collected.

[106]

[22] Manning's n is the only freely varying parameter in the equations, the Saint-Venant equations, that underlie the model[18] of flow in an open channel. The accuracy of the applied value of Manning's n will be almost exclusively determinative of the accuracy of the output of the model.[19] The most reliable way to arrive at the correct value is to measure flow and to work back to the value using the Manning equation. More measurements make for more accurate outcomes. The next method, second best, is to use published tables based on sediment size. This method tends to produce an overestimation. The least accurate method is to use handbooks with photographs of sample streams with Manning's n values. An example of this is exhibit 10, "Roughness Characteristics of New Zealand Rivers" by DM Hicks and PD Mason, 1991 Water Resources Survey. While a useful example of such a handbook, in Dr Rayburg's view it is of no practical assistance presently as the New Zealand streams depicted in it are not usefully comparable to Coral Creek. The directions entitled "How to use this book" on pages 11 and 12 require selecting a similar stream and there was no suggestion that this reference contained anything which could be fairly so described. An earlier (1967) publication in the United States was similarly of no assistance for present purposes.[20]

[23] Dr Rayburg also pointed out that a single estimation of Manning's n ignores real-world channel variability and the effects of woody debris and vegetated islands.[21] The modelling of the existing stream and of the proposed new channel both use a Manning's n of 0.045 when it is not possible that the new channel, in the absence of obstructions, would have friction approaching that of the existing stream.[22] This makes it certain that the estimate for the existing creek is incorrect and is another reason why the modelling is invalid.[23]

[24] Referring to the "Bowen Basin River Diversions Design and Rehabilitation Criteria" (2002),[24] Dr Rayburg notes that none of the hydraulic modelling results for the more than 30 streams considered was the result of a model which had been calibrated. The model results must be considered therefore to be invalid.[25] It provided the context within which the modelling results for Coral Creek were evaluated.[26] The earlier report "Maintenance of Geomorphic Processes in Bowen Basin River Diversions" (2000) illustrated how every single previous diversion in the Bowen Basin had resulted in increased sedimentation and erosion[27] which the 2002 design criteria was an attempt to address for future diversions. The guideline, in Table 1, states the criteria drawn from this document. Dr Rayburg states that the Coral Creek diversion is likely to fail by exhibiting accelerated erosion and sedimentation that spreads to the natural stream above and below it.[28] He gives it as his considered opinion that all of the Brisbane Stormwater Management modelling results are invalid. Should the diversion proceed on the basis of this modelling "alone"[29] (emphasis added), it is his view that the diversion and the natural stream above and below it "will experience accelerated erosion and deposition rates that are likely to result in instability in the stream that lasts for decades or longer and may result in a permanent change in the river's form and process operation".[30]

[25] In his reply report, exhibit 2, Dr Rayburg responds to reports obtained by the second respondents and adheres to his views. He provides, at Figure 1, eight photographs of Coral Creek showing "elements of friction" such as vegetation, fallen trees and accumulations of sand and rocks. In reply to the report of Dr Connor, who was called by the second respondents, he states that from Dr Connor's statement that when on the limited occasions when calibration and verification of the model have been available, the performance of the model has been adequate, it is implied that Dr Connor has very limited experience with proper modelling protocol.[31] He states that the lack of existing data is not a reason to proceed without it and that a basic data set for calibration could have been collected by the efforts of no more than two people for half a day's work on two separate occasions about 2 - 4 weeks apart.[32] He disagrees with Dr Connor that error assessment is rarely done in industry.[33] Dr Rayburg refers to Dr Connor's statement that where data is absent or minimal practically everyone concerned knows that the results are limited by this as "categorically untrue".[34] In his view, to not acknowledge that the model results "must be construed as highly suspect in the absence of adequate input data is both inappropriate and misleading".[35] Dr Rayburg refers to Dr Connor's emphasis on the reduced slope of the diversion as a factor favouring its stability and states that such a discussion is irrelevant in the absence of an accurate determination of Manning's n.[36]

[115]

[29] Dr Rayburg's opinion that long term instability in the stream is likely to result[44] is contradicted by Dr Connor, the registered professional engineer who notes that the 20% increase in stream length of the diversion with the consequent reduction in stream slope, together with the extensive vegetation and construction works, provide confidence in the long term stability. This is a matter directly within Dr Connor's expertise as a registered professional engineer, and is not a matter within the expertise of Dr Rayburg so the Court must accept Dr Connor's opinion on the point and reject that of Dr Rayburg.

[30] Dr Connor also provided a site visit report, which became exhibit 6. In this he describes his visit, on 25 November 2013, to observe the full length of the section of Coral Creek to be diverted. He took the photographs which are attached to his report. At the time, there was no water in the creek.[45] He states that when a major flood occurs in the small catchment the creek is likely to reach peak height in a matter of hours, fall quickly and tail off steadily over a day or two.[46] For most of the time, the creek will be dry.[47] It will fill quickly with no real warning[48] and forecasts for heavy rain or storms in the area cannot pinpoint small areas such as the creek catchment.[49] This is relevant to the practicality of collecting the data referred to by Dr Rayburg.

[31] Based on his observations made during the site visit, Dr Connor considers a Manning's n of 0.045, as used in the modelling "a very reasonable assessment for roughness of the whole creek system in substantial events".[50] He points out that it cannot be known when flow in the creek will occur and measurement would need to be made quickly. Mobilising a measurement team to the site in time would be unlikely unless they were stationed nearby.[51] The need to measure two events compounds the difficulties.[52] Dr Connor considered that automatic level indicators would give "some idea" of peak water levels and levels at different stages of a flood. Although actual flows would not be known, testing with a variety of flood discharge time histories and varying values of Manning's n "would attempt to match the profile".[53] This would give some indication of Manning's n but the likelihood of an event with an ARI of 20 years in the period of a study would be small. Dr Connor was less sanguine than Dr Rayburg about the usefulness of projections from events in the ½ to 2 year ARI to what might be expected in a 50 year ARI event, for instance, not agreeing that it could be done.

[32] Dr Connor did not believe that the New Zealand book, exhibit 10, would be useful as New Zealand streams regularly experience low flows while Coral Creek is more likely to be largely dry until the occasional flood. The handbook did not show any streams similar to Coral Creek.[54] He did note that for the New Zealand streams experiencing significant flows, greater than 100 cumecs (m³/sec), although high Manning's n values may be measured at low flows, 86% of the streams exhibited Manning's n values equal to or lower than 0.047 for flows greater than 100 cumecs. He was of the opinion that this gave "some support" to the view that 0.045 is "not an unreasonable value for roughness in Coral Creek when modelling substantial flows".[55] (emphasis added) Dr Connor provides his working on pages 3-5 and 3-6 of his report, exhibit 6. He was not challenged on this in cross-examination and there is no evidence from Dr Rayburg on this particular aspect. This is a matter within Dr Connor's expertise and the Court must accept his evidence on the point. Dr Rayburg did not make any estimate of Manning's n by any means.

[33] Dr Connor was present during the evidence given by Dr Rayburg. When cross-examined, Dr Connor was shown exhibit 11, an extract from the "Good Modelling Practice Handbook" published by the Dutch Department of Public Works. It sets out the modelling process in a series of steps. Dr Connor was of the view that it should not be a criticism just because you can't see all those steps in a work; here referring to the modelling that was done. The handbook itself states that the steps in it are not relevant at all times.[56] He was also shown an article titled "Calibration of process-oriented models" which was published in Ecological Modelling 83 (1995) 55-66. This became exhibit 12 and was used in support of the idea that model calibration is a critical phase in the modelling process. Dr Connor was of the view that it is critical "if you have measurements".[57]

[116]

[34] The second respondents also called Mr Stephen Clark, a director of Water Technology Pty Ltd. Mr Clark is a registered professional engineer in Queensland, specialising in water engineering. He was called as a witness of fact. His affidavit became exhibit 4. He has not been to the Coral Creek site.[58] He is familiar with the steps taken to conduct the hydraulic modelling. He took steps to satisfy himself of what the value of Manning's n should be. This included reviewing approximately 34 photographs of the creek, referring to a 1967 United States Geological Service paper and text, the text "Open Channel Hydraulics", Chow, (1959) and a 2009 Australian Handbook of stream roughness coefficients.[59] He states that there is no calibration data available for Coral Creek. If there was such data he would have used it.[60] In his experience, in the absence of calibration data, it is common for an engineer to identify the Manning's n value having regard to reference texts and previous experience.[61] Consideration was given to the possible consequences of not establishing the revegetation at the time of a flood so a Manning's n of 0.035 was used to model the results in such a circumstance.[62] Some parts of Mr Clark's affidavit were excluded after objections taken on behalf of the appellant. These related to expressions of opinion. Mr Clark states that throughout the development of the existing case and design case models and adoption of the Manning's n, he and the two other engineers who were under his supervision undertook sensitivity testing of the model to ensure that it was producing reasonable and appropriate results.[63] At several stages, the model was run using Manning's n values of 0.030 and 0.080.[64] A Manning's n of 0.030, in his experience, is a very low value not often representative of flows in natural channels like Coral Creek. A Manning's n of 0.080, on the other hand, is a relatively high value for a channel where the vegetation is sparse, such as in this part of Coral Creek.[65] In addition to using the "Sobek" software, which would take up to 12 hours to run on the computer, Mr Clark and his staff used a spreadsheet to estimate the likely impact on the hydraulic modelling of particular changes to the model.[66] He provides an example of these calculations in exhibit SQC-2 of his affidavit, exhibit 4. Mr Clark points out that because the diversion channel was an engineered construction, its regularity in comparison to the creek made it more likely that the spreadsheet would accurately indicate its hydraulics.[67]

[35] Mr Clark points out that the guideline values for velocity, shear stress and stream power are whole channel values for any particular cross-section[68] whereas the modelling results reported detailed distributions of those parameters for all locations along Coral Creek on the basis of a 5 m grid. The results reported show the maximum values.[69] Mr Clark has undertaken an analysis of the "Sobek" hydraulic modelling results for the existing creek and for the designed diversion to identify the whole channel values for a number of cross-sections. It will be recalled that Table 1 in the guideline requires consideration of three flow situations:

An ARI of 2 years when there is no vegetation.

An ARI of 2 years when there is vegetation.

An ARI of 50 years.

[118]

[36] Mr Clark has chosen five cross-sections of the existing Coral Creek and analysed the "Sobek" hydraulic modelling results for the parameters shear stress and stream power in a flow with an ARI of 50 years. The selected cross-sections are at the beginning and end of the relevant section of the creek and roughly evenly within the section of creek. Channel shear stress and stream power in the 1 in 50 year flow event are generally lower in the creek when a Manning's n of 0.045 is used than in the diversion when a Manning's n of 0.035 is used for the diversion. The values for the shear stress and stream power in the creek are below the guideline values but above them in one example in the diversion channel. When the Manning's n for the diversion channel is 0.045 the shear stress and stream powers are close to or below the design guidelines for the 1 in 50 year ARI flow event.[70]

[37] In his evidence, Dr Rayburg said that "from a general perspective and from an engineering perspective"[71] calibration protocols are well known. Since he is not an engineer he is unable to give evidence of what may or may not be well known to engineers so his statement must be taken as referable to a general perspective, the extent of which was not expressed to be limited in any way. He explained that gathering the data for model calibration would involve the people doing the measuring walking "across the river when there is flow in it".[72] He was of the opinion that two or three events less than Q2, a flow event expected to recur with a two year interval, would recur "on average more frequently than once every two years".[73] Measuring velocity at one point, he said, "takes about 60 seconds".[74] In view of the evidence concerning the rapid rise and fall to be expected in this creek, the uncertainty of when it might take place, day or night and the obvious dangers to the people expected to "walk across the river", in this case the creek, in conditions which are to be expected to be rapidly changing, opportunities for measurement might be expected to be somewhat limited. Droughts and floods are characteristic of Australia and even a Q2 event is a "bankfull" event[75] which stands apart from Dr Rayburg's intention that the events to be measured should be those that are "safe".[76] Dr Rayburg said that data for less than a Q2 event was required.[77] This is understandable when it is recalled that Dr Connor's evidence was that an ARI or (Q) 2, event will result in peak flow generally from 2 to 4 m deep in this creek.[78] The difficulty with Dr Rayburg's concept of a "safe" event for measurement is that there was no evidence at all of what the ARI of a "safe" event is and how it may relate to an ARI of ½ or 1 or 2 years. If it is defined as a certain depth, say half a metre, that is a completely different unit of measurement to say, an ARI 1 flow and the evidence does not contain the equivalent of a Rosetta stone that would allow a "safe" and therefore measurable event to be located anywhere on the ARI (or to use the other term Q) 2 scale. I also note that the flow data collection, in view of the practicalities to which I have referred, could not be described as able to be collected "with very little effort",[79] which is Dr Rayburg's view.

[38] Dr Rayburg states that measurements could have been made on the day he visited the site.[80] He did not make measurements himself[81] but even if he had, and done a subsequent measurement cycle on another day, it is not possible to say where what was measured would be located on the ARI scale. If it cannot be said that what was measured was an ARI ½ or 1 to 1.5 event, for instance, it would be impossible to use it as a basis for extrapolation to any other ARI level such as Q2 or Q50. On the evidence presented, it is not possible to be satisfied that measurements made in the way explained by Dr Rayburg would yield any useful data for modelling of a flow event on the ARI scale such as is required by the guideline. His evidence on this point is unhelpful and it does not actually point to an easily remediable defect in the approach which was actually employed, although he says that it does.

[126]

[41] Dr Connor's evidence was that at a "bankfull" flow, "which we're saying is a Q2 flow" "two to four metres deep", Manning's n is going to be "vastly different" to what is measured with people wading out across a stream. That activity should only be done when the flow is "up to the knee to be safe".[83] Using instruments to make it unnecessary to wade in the creek will address the physical risk involved in entering the stream but not the logistical challenges of attending at the right times to do the measuring. Dr Connor's view was that it was "fanciful" to think the measurements could be achieved simply.[84]

[42] Dr Connor was of the view that data collected in low flows "may give a very inaccurate outcome for a Q2 or a higher flow".[85] Dr Rayburg's view that it might be better not to model at all rather than to do so in a manner which he believes to be unsatisfactory is not an option when the guideline requires that hydraulic modelling is to be completed. The method to determine Manning's n which Dr Rayburg describes as the least accurate "although perhaps the most commonly used approach"[86] of recourse to a handbook with pictures was employed by Dr Connor in exhibit 6. Dr Connor pointed out that none of the streams appear similar to Coral Creek but he was able to derive "some support" for the view that 0.045 "is not an unreasonable value for roughness in Coral Creek when modelling substantial flows".[87] (emphasis added)

[43] Dr Rayburg does not contend for another value of Manning's n, just that there could not be confidence in the value applied. After considering:

The accessibility of data which would be needed to proceed as Dr Rayburg states is necessary;

The lack of utility of what would be collected by the methods he describes;

The lack of information on where on the ARI scale a measured flow is located;

The evidence of the difficulty of extrapolation of values for Manning's n found at low flows to that applicable at higher flows up to, for instance, the required 50 year ARI flow level;

Some support from exhibit 10 for a Manning's n of 0.045 being not an unreasonable value;

At a glance

Court

Decision date

Catchwords

Source

Catchwords

Judgment (133 paragraphs)

Parties

Legislation Cited (1)