Insights from a geoscience communication activity,
verified using preliminary investigations with an artificial neural network,
illustrate that observation of humans' abilities can help design an
effective artificial intelligence or “AI”. Even given only one set of
“training” examples, survey participants could visually recognize which flow
conditions created bedforms (e.g. sand dunes and riverbed ripples) from their
shapes, but an interpreter's geoscience expertise does not help. Together,
these observations were interpreted as indicating that a machine learning
algorithm might be trained successfully from limited data, particularly if
it is “helped” by pre-processing bedforms into a simple shape familiar from childhood play.
Introduction
Environmental flows shape the surface they flow over. The variety of
features produced (e.g. sand ripples on a beach), known as bedforms, reflect
and preserve characteristics (e.g. speed and depth) of the flowing ice, water
or air (Venditti, 2012; Bullard et al., 2011;
Storrar and Stokes, 2007). The relationships between bedform morphology and
flow are contested where observation is extremely difficult, such as under
ice sheets (e.g.
Hillier et al., 2018; King et al., 2009), and best understood for
unidirectional water flow over sand in a laboratory setting, mimicking a
river. Even in this idealized fluvial setting, it is difficult to construct
a one-to-one link between bedform type (e.g. ripples or dunes) and specific flow
conditions (Venditti, 2012; Froehlich, 2020).
Illustratively, ripples have a higher aspect ratio (height / length) than
dunes (e.g. Allen, 1968), yet the observational ranges
overlap (Venditti, 2012), creating uncertainty when attempting
to link morphology with hydraulic conditions. Many variables related to
hydraulics and/or the physics of sediment movement have been proposed to
remove the overlap in bedform stability diagrams such as Fig. 1a. Only
recently has a distinct and non-overlapping zonation of bedform type and
flow-sediment condition been developed using a quantity called shear
velocity (Duran Vinet et al., 2019). Inverting this result
may help realize the aspiration of developing a means to reliably infer flow
conditions from bedform morphology, which is often the only option for
inferring past environmental conditions on Earth (Leary and
Ganti, 2020) or Mars (Ohata et al., 2017; Edgett and
Lancaster, 1993).
Machine learning or “AI” (artificial intelligence) algorithms, such as artificial neural networks
(ANNs), have great potential in geomorphology (Sofia et al.,
2016; Froehlich, 2020; Valentine and Kalnins, 2016; Shumack et al., 2020)
and offer an opportunity to examine this problem, as they do not assume
simple (e.g. linear or one-to-one) relationships between inputs and predicted
variables (Wang et al., 2009; Faruk, 2010).
Unexploited morphological subtleties may exist by which to categorize
bedforms or even to accurately position them on stability diagrams. This
work examines the scope for using ANNs to distinguish the flow conditions in
which bedforms originated by asking if the ability exists in non-artificial
(human) intelligence for two particulars:
Is it possible to identify the environment (e.g. river and
desert) of a bedform's genesis from its shape?
In the fluvial environment, is it possible to distinguish flow
conditions?
(a) Illustrative bedform stability “phase diagram” for
unidirectional fluvial (i.e. river) bedforms, synthesized from multiple
sources (Ohata et al., 2017; Lewis and
McConchie, 1994; Southard and Boguchwal, 1990). Main types considered here
(i.e. ripples and dunes) are highlighted. Experiments (Expts.) 1–4 are positioned
indicatively. (b) Distance–height profiles (strictly speaking time series)
like those given unannotated to participants, i.e. one each from Expts. 1–3, all scaled to the same dimensions. Horizontal axis is time because in
the flume tank a stationary sensor recorded height as bedforms passed
beneath it. (c) Example of how height (H) and width (W) are determined. Measured heights (thick
black line) are processed using the SWT algorithm to identify bedforms,
drawing a line beneath them (thin black line), which is then approximated as
flat-topped cones (grey lines). SWT parameters as in Hillier (2008). (d) Height–width relationships for
the four experiments, with colours as in (a): lines are sliding means with W
(Gaussian weights, width 60 s); shaded areas are full ranges for Expts. 1 and 4; and dots are the means (±2σ) of the upper quartile of
the data when the small bedforms (i.e.H< 0.5 cm) are excluded. (e) Comparison of the actual experiment number with out of sample prediction of the experiment number by the
ANN using H and W: individual bedforms (light grey) and subsets of five bedforms with
(grey) and excluding (dark grey) small bedforms.
Method, data and ethics
An online survey was conducted, initially at the “Non-Equilibrium Flow and Landform Coupling Workshop” (19 May 2021) and then expanded to participants without geomorphological expertise using
authors' close contacts (friends, colleagues and family). For Q1,
participants attributed distance–height profiles across 34 individual
bedforms and 13 bedform sequences (≥3 bedforms) to one of four
environments (fluvial (river), glacial, marine and aeolian (desert)). For Q2, participants ranked three
profiles according to flow strength (shear velocity), thrice for individual
forms and thrice for bedform sequences. Examples were provided to isolate
visual shape analysis from prior knowledge (Fig. 1b); black-and-white
profiles were used to exclude contextual clues (dataset characteristics and
other features in the landscape), and the order of options (e.g. B, A and C)
was shuffled for each participant to prevent bias. Scale (e.g. metres)
readily distinguishes the environment without using bedform shape, so it was not
given.
Ethical approval was given by the Ethics Review Sub-Committee at
Loughborough University.
Aeolian data are from ASTER (v2; Advanced Spaceborne Thermal Emission and Reflection Radiometer) across linear and transverse dune types from the
Namib Desert (Bullard et al., 2011); glacial data are from near
Lough Gara in Ireland (Hillier and Smith, 2008); fluvial data are from
four laboratory experiments (Expts. 1–4) of non-linearly increasing shear
velocity (Unsworth, 2015); and marine data are from the Irish Sea.
Distance–height profiles of these data were created, although for the
fluvial measurements “time” was used as a proxy for “distance” (see Fig. 1b). For the survey, representative examples of individual bedforms and
sequences were manually selected from these datasets. Pre-processing to
estimate bedforms' height (H) and width (W) – see Fig. 1c and d – used the spatial
wavelet transform (SWT) algorithm (Hillier,
2008) and fitting of flat-topped cones (Hillier, 2006).
ANN analysis to follow up the survey used a multi-layered perceptron (MLP)
with four hidden layers with 28, 56, 56 and 28 nodes, each with a ReLU (rectified linear unit)
activation function. In a baseline analysis, input to predict the fluvial
flow regime (coded by experiment number) was non-overlapping profile
segments 160 s long. After this, to “help” the ANN bedform, shapes
(H and W) were input once each per analysis, either (i) individually or (ii) as
pseudo-sequences – groups of five bedforms in increasing size order,
selected at random without replacement. Weights and biases were updated
using the Adam optimizer of PyTorch using a loss function that calculates
the mean squared error, all within a feed-forward back-propagation algorithm.
Results
Of the 42 survey participants 25 self-identified as geoscientists, and 16 did not. For Q1, participants correctly identified the one of four
environments (e.g. fluvial and aeolian) in which individual features originated
32 % of the time, slightly if significantly (two-tailed t test, p≪ 0.01)
better than the 25 % expected of guesswork. This rises to 51 % for
bedform sequences. For Q2, participants ranked entirely correctly three flow
strengths (Expts. 1–3) for 46 % of individual features and 60 % of
sequences, much better than the 16 % expected of guesswork (p≪ 0.01).
In none of the questions did geoscientists perform better than
non-geoscientists, with mean percentages of correct answers being
indistinguishable (two-tailed t test, p> 0.05). The overall
sentiment is encapsulated by one comment:
I felt this was a geometrical exercise of recognizing same patterns at different scales. I did not feel that my experience as an “expert” in bedforms really made any difference from, say, my son taking the test.
Several participants commented that their ability to distinguish
environments might be to do with characteristics of the data (e.g.
smoothness due to data resolution), not bedform shape. This is a potential
pitfall of training an ANN, avoided here by only analysing the fluvial data.
In the baseline ANN analysis, flow regime was predicted poorly (r2=0.03). Fitting a simplified geometry (H and W) to bedforms improves results
dramatically, particularly if pseudo-sequences of bedforms are used (Fig. 1e); individual forms are weakly predictive (light grey, r2=0.11),
but sub-sets of five bedforms are more strongly so (grey, r2=0.56),
particularly if very small bedforms present in all experiments (H< 0.5 cm) are excluded (dark grey, r2=0.80). This is consistent with
a visual assessment (Fig. 1d) where individual morphologies overlap between
experiments, but their trends, as well as averages over a number of bedforms, are
distinctly different.
Discussion
Morphologies from differing environments (e.g. glacial and fluvial) can be
viewed as similar, indicators of analogous processes at work (e.g. Shaw, 1983) and modelled with identical equations (e.g. Fowler, 2002; Duran Vinet et al., 2019) or
statistics (e.g. Hillier et al., 2016;
Einstein, 1937). Despite similarities in appearance, the survey results
clearly demonstrate a level of ability to distinguish flow conditions from
distance–height data of the bed and, unsurprisingly, imply that an ANN
should perform better if utilizing sequences of bedforms rather than
evaluating individual forms in isolation. Interestingly, geoscientists' a priori and
contextual knowledge added little, indicating that all required visual cues
lie within the distance–height profiles. Furthermore, one training dataset
sufficed for the survey's participants, a stark contrast to the thousands of
datasets required to train ANNs performing pure pattern recognition (e.g. Bishop, 1996), suggesting that participants drew on
significant previous learning (e.g. identification of basic idealized
shapes). Together, these observations prompt the testable idea that an
effective ANN might be efficiently trained by helping it via
pre-processing profiles into simple shape parameters that would have been
readily understood by all participants (H and W).
Preliminary analysis with an ANN supports our speculations. It demonstrates
that an AI with predictive efficacy can be built using limited data,
improved by using bedform sequences (Fig. 1e). The increase in predictive
skill to 0.80 with pre-processing help demonstrates, in principle, the
utility of this approach when building an effective AI for geomorphology
that avoids the crippling need for thousands of datasets when examples in nature
are often limited in number. Speculatively, it follows that machine learning
techniques might work well and be trained efficiently wherever non-experts
make good decisions based on images. This study was on equilibrium
conditions but illustrates that ANNs may be key to linking forms and flow
for transitional, non-equilibrium conditions (e.g.
Myrow et al., 2018).
Code and data availability
For completeness and transparency, the survey form and anonymized answers
are provided in the Supplement. For the ANN, input data with an associated
README file and pseudo-code are provided to make the work reproducible.
The supplement related to this article is available online at: https://doi.org/10.5194/gc-5-11-2022-supplement.
Author contributions
JKH conceived and led the work. JKH and CU designed and ran data collection at
the workshop. LDC, JKH and CU undertook analyses and data preparation. All
authors contributed to the analytical design, writing of the manuscript, and
development of the understanding and ideas presented.
Competing interests
At least one of the (co-)authors is a member of the editorial board of Geoscience Communication. The peer-review process was guided by an independent editor, and the authors also have no other competing interests to declare.
Disclaimer
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Acknowledgements
We thank Dylan Ward and an anonymous reviewer for their thoughtful and
challenging comments which significantly improved the paper. The “Non-Equilibrium Flow and Landform Coupling Workshop” was organized by Tim
Marjoribanks, Chris Keylock, Christopher A. Unsworth, Daniel R. Parsons and
Jonny Higham, with support from the British Society for Geomorphology. We
thank Matt Baddock for preparing the aeolian data and the 42 anonymous
participants for completing the survey.
Review statement
This paper was edited by Louise Arnal and reviewed by Dylan Ward and one anonymous referee.
ReferencesAllen, J. R. L.: Their Relation to Patterns of Water and Sediment Motion, North Holland Publishing Company, Amsterdam, 433 pp., 10.1017/S001675680005946X, 1968.Bishop, C. M.: Neural Networks for Pattern Recognition, Oxford University
Press, 1st edition, Clarendon Press, Oxford, 502 pp., ISBN 9780-198538646, 1996.Bullard, J., White, K., and Livingstone, I.: Morphometric analysis of
aeolian bedforms in the Namib Sand Sea using ASTER data, Earth Surf. Proc. Land., 36, 1534–1549,
2011.Duran Vinet, O. D., Adreiotti, B., Claudin, P., and Winter, C.: A unified
model of ripples and dunes in water and planetary environments, Nat. Geosci., 12,
345–350, 10.1038/s41561-019-0336-4, 2019.Edgett, K. S. and Lancaster, N.: Volcaniclastic aeolian dunes: terrestrial
examples and application to martian sands, J. Arid Environ., 25, 271–297,
10.1006/jare.1993.1061, 1993.Einstein, H.: Der Geschiebetrieb als Wahrschein-lichkeitproblem, in: Sedimentation (Symposium to honor H. A.
Einstein), edited by: Shen, H. W. [English translation: Bedload transport as a probability problem, edited
by: W. W. Sayre] Fort Collins, Colorado, 1972,
C1–C105, Zurich: ETHZ, 1937.Faruk, D. Ö.: A hybrid neural network and ARIMA model for water quality
time series prediction, Eng. Appl. Artif. Intel., 23, 586–594,
10.1016/j.engappai.2009.09.015, 2010.Fowler, A. C.: Evolution equations for dunes and drumlins, Racsam. Rev. R. Acad. A., 96, 377–387, 2002.Froehlich, D. C.: Neural Network Prediction of Alluvial Stream Bedforms, J. Hydrol. Eng., 146, 04020084, 10.1061/(ASCE)HY.1943-7900.0001831, 2020.Hillier, J. K.: Pacific seamount volcanism in space and time, Geophys. J. Inter., 168, 877–889, 10.1111/j.1365-246X.2006.03250.x, 2006.Hillier, J. K.: Seamount detection and isolation with a modified wavelet transform, Basin Res., 20, 555–573, 10.1111/j.1365-2117.2008.00382.x, 2008.Hillier, J. K. and Smith, M.: Residual relief separation: digital elevation
model enhancement for geomorphological mapping, Earth Surf. Proc. Land., 33, 2266–2276, 10.1002/esp.1659, 2008.Hillier, J. K., Kougioumtzoglou, I. A., Stokes, C. R., Smith, M. J., and
Clark, C. D.: Exploring explanations of subglacial bedform sizes using
statistical models, Plos ONE, 11, e0159489,
10.1371/journal.pone.0159489, 2016.Hillier, J. K., Benediktsson, Í. Ö., Dowling, T. P. F., and
Schomacker, A.: Production and preservation of the smallest drumlins, GFF, 140, 91–92, 10.1080/11035897.2018.1457714, 2018.King, E. C., Hindmarsh, R. C. A., and Stokes, C. R.: Formation of mega-scale
glacial lineations observed beneath a west Antarctic ice stream, Nat. Geosci., 2, 585–596, 2009.Leary, K. C. P. and Ganti, V.: Preserved Fluvial Cross Strata Record Bedform
Disequilibrium Dynamics, Geophys. Res. Lett., 47, 1–11, 2020.Lewis, D. W. and McConchie, D. M.: Practical sedimentology, 2nd ed., Springer, US, 212 pp., ISBN 9780-442012175, 1994.Myrow, P. M., Jerolmack, D. J., and Perron, J. T.: Bedform Disequilibrium, J. Sediment. Res., 88, 1096–1113, 10.2110/jsr.2018.55, 2018.Ohata, K., Naruse, H., Yokokawa, M., and Viparelli, E.: New Bedform Phase
Diagrams and Discriminant Functions for Formative Conditions of Bedforms in
Open-Channel Flows, J. Geophys. Res.-Earth, 122, 2139–2158, 10.1002/2017JF004290,
2017.Shaw, J.: Drumlin formation related to inverted melt-water erosional marks, J. Glaciol., 29, 461–479, 1983.Shumack, S., Hesse, P., and Farebrother, W.: Deep learning for dune pattern
mapping with the AW3D30 global surface model, Earth Surf. Proc. Land., 45, 2417–2431,
10.1002/esp.4888, 2020.Sofia, G., Hillier, J. K., and Conway, S. J.: Frontiers in Geomorphometry and Earth Surface Dynamics: possibilities, limitations and perspectives, Earth Surf. Dynam., 4, 721–725, 10.5194/esurf-4-721-2016, 2016.Southard, J. B. and Boguchwal, L. A.: Bed configurations in steady
unidirectional water flow part 2, Synthesis of flume data, J. Sediment. Petrol., 60, 658–679,
1990.Storrar, R. and Stokes, C. R.: A glacial geomorphological map of Victoria
Island, Canadian Arctic, J. Maps, 3, 191–201,
10.1080/jom.2007.9710838, 2007.Unsworth, C.: River Dunes in Unsteady Conditions, PhD, University of Hull, GB, available at: https://hydra.hull.ac.uk/assets/hull:13096a/content, 2015.Valentine, A. and Kalnins, L.: An introduction to learning algorithms and potential applications in geomorphometry and Earth surface dynamics, Earth Surf. Dynam., 4, 445–460, 10.5194/esurf-4-445-2016, 2016.
Venditti, J. G.: Bedforms in Sand-Bedded Rivers, in: Treatise on
Geomorphology, edited by: Shroder, J. and Wohl, E., Elsevier, 137–162, ISBN 9780-123747396, 2012.Wang, W., Chau, K., Cheng, C., and Qui, L.: A comparison of performance of
several artificial intelligence methods for forecasting monthly discharge
time series, J. Hydrol., 374, 294–306, 10.1016/j.jhydrol.2009.06.019,
2009.