Since the framework for fully-integrated surface water and groundwater hydrologic models was first laid out over 50 years ago, considerable progress has been made towards realizing that vision. Today, fully integrated hydrologic models (like HydroGeoSphere) are routinely applied to a wide variety of problems and scales.
Compared to traditional groundwater or surface water only models a fully integrated approach to simulating the hydrologic cycle presents a different set of opportunities (and challenges!). Making the transition to fully integrated modelling often requires a bit of a shift in perspective to fully realize the benefits of this powerful approach to water resources engineering.
Join Dr. Steve Berg to learn how you can maximize the power of HydroGeoSphere in your own modelling projects.
In this talk, we will:
1️⃣Briefly discuss the seminal paper for this field on numerical modelling (Freeze and Harlan,1969)
2️⃣Review the key process hydrologic processes handled by HGS and discuss what makes tightly-coupled integrated modelling so powerful
3️⃣Review key applications for which this modelling approach excels
4️⃣Review of ‘advanced’ HGS functionality
5️⃣How to design an HGS model for your problem to best leverage it’s strengths
6️⃣Discuss the future of integrated hydrologic modelling (where are we heading)
Want to read up on the sample applications listed in the webinar? Check the Aquanty blog: https://www.aquanty.com/blog
References:
Berg, S.J. and Sudicky, E.A., 2019. Toward large‐scale integrated surface and subsurface modeling. Groundwater, 57(1), pp.1-2.
Chen, J., Sudicky, E.A., Davison, J.H., Frey, S.K., Park, Y.J., Hwang, H.T., Erler, A.R., Berg, S.J., Callaghan, M.V., Miller, K. and Ross, M., 2020. Towards a climate-driven simulation of coupled surface-subsurface hydrology at the continental scale: a Canadian example. Canadian Water Resources Journal/Revue canadienne des ressources hydriques, 45(1), pp.11-27.
Erler, A.R., Frey, S.K., Khader, O., d’Orgeville, M., Park, Y.J., Hwang, H.T., Lapen, D.R., Peltier, W.R. and Sudicky, E.A., 2019. Evaluating climate change impacts on soil moisture and groundwater resources within a lake‐affected region. Water Resources Research, 55(10), pp.8142-8163.
Fatichi, S., Vivoni, E.R., Ogden, F.L., Ivanov, V.Y., Mirus, B., Gochis, D., Downer, C.W., Camporese, M., Davison, J.H., Ebel, B. and Jones, N., 2016. An overview of current applications, challenges, and future trends in distributed process-based models in hydrology. Journal of Hydrology, 537, pp.45-60.
Frey, S.K., Miller, K., Khader, O., Taylor, A., Morrison, D., Xu, X., Berg, S.J., Hwang, H.T., Sudicky, E.A. and Lapen, D.R., 2021. Evaluating landscape influences on hydrologic behavior with a fully-integrated groundwater–surface water model. Journal of Hydrology, 602, p.126758.
Kollet, S., Sulis, M., Maxwell, R.M., Paniconi, C., Putti, M., Bertoldi, G., Coon, E.T., Cordano, E., Endrizzi, S., Kikinzon, E. and Mouche, E., 2017. The integrated hydrologic model intercomparison project, IH‐MIP2: A second set of benchmark results to diagnose integrated hydrology and feedbacks. Water Resources Research, 53(1), pp.867-890.
LeRoux, N.K., Frey, S.K., Lapen, D.R., Guimond, J.A. and Kurylyk, B.L., 2023. Mega‐Tidal and Surface Flooding Controls on Coastal Groundwater and Saltwater Intrusion Within Agricultural Dikelands. Water Resources Research, 59(11), p.e2023WR035054.
Maxwell, R.M., Putti, M., Meyerhoff, S., Delfs, J.O., Ferguson, I.M., Ivanov, V., Kim, J., Kolditz, O., Kollet, S.J., Kumar, M. and Lopez, S., 2014. Surface‐subsurface model intercomparison: A first set of benchmark results to diagnose integrated hydrology and feedbacks. Water resources research, 50(2), pp.1531-1549.
Schilling, O.S., Cook, P.G. and Brunner, P., 2019. Beyond classical observations in hydrogeology: The advantages of including exchange flux, temperature, tracer concentration, residence time, and soil moisture observations in groundwater model calibration. Reviews of Geophysics, 57(1), pp.146-182.
Wright, L. and Davidson, S., 2020. How to tell the difference between a model and a digital twin. Advanced Modeling and Simulation in Engineering Sciences, 7(1), pp.1-13.
Xu, S., Frey, S.K., Erler, A.R., Khader, O., Berg, S.J., Hwang, H.T., Callaghan, M.V., Davison, J.H. and Sudicky, E.A., 2021. Investigating groundwater-lake interactions in the Laurentian Great Lakes with a fully-integrated surface water-groundwater model. Journal of Hydrology, 594, p.125911.
Yang, J., Wang, Q., Heidbüchel, I., Lu, C., Xie, Y., Musolff, A. and Fleckenstein, J.H., 2022. Effect of topographic slope on the export of nitrate in humid catchments: a 3D model study. Hydrology and Earth System Sciences, 26(19), pp.5051-5068.
Visit our websites for more information on Aquanty’s projects and products!
Aquanty: https://www.aquanty.com/
Canada1Water: https://www.canada1water.ca/
HydroGeoSphere: https://www.aquanty.com/hydrogeospheresw
HGSRT: https://hgsrt.com/
Hi everyone thank you so much for joining us in today’s webinar my name is Bren mcneel I’m the technical sales and marketing lead here at aanti I think I’ve been in contact with almost everyone here but if uh if I if you haven’t or if you have questions about
Hydro geosphere aanti any of our technology or our projects feel free to reach out to me anytime my email address is in there uh in the chat um we’re really hoping that the webinar today will spur thoughts and and lead to further Partnerships with water water resources experts across Canada and the
World um if there if there’s anything in this project or in this presentation that interests you if you have any ideas or projects of your own that we might be able to help with definitely encourage you to reach out to us um Steve you can go to the next slide before we truly
Begin I’d like to make a quick land acknowledgement as is the case with almost every nation with a colonial history Canada has uh a very difficult history fraught with historical injustices is perpetrated against the indigenous people of this land as a country we’re attempting to embark on a journey of Truth and Reconciliation
Acknowledging the truth of the historical crimes committed against the original inhabitants of this land and making sincere efforts for reconciliation uh and to restore Canada’s relationship with the indigenous people of this land so this land acknowledgement is just one small way that we at aanti can contribute to
That effort a quanti is based in waterl Ontario it’s a City situated right in the middle of what’s known as the Haldeman tract which you can see in green on that image to the right this tract of land is located on the traditional territory of the neutral adaw wanderon Anisha Bay and hoden
Shon’s people the land was quote unquote given to the six nations of the Grand River as compensation for their role in the American war for independence during which these indigenous Nations fought for the British as you can see today less than 5% of this treaty land is
Still under the administra ation of the six nations of the Grand River and aanti would just like to acknowledge our privilege and Express gratitude for being able to live and work on this territory now I’ll just do a quick introduction for today’s speaker Steve Berg is the president and CEO at aanti
And is also an adjunct assistant research professor in the department of Earth and environmental Sciences at the University of waterl Steve received his PhD in Earth Sciences from the University of watero focusing on hydro geology and numerical modeling and also has an MBA from Wilford Lauer University Steve joined aanti in 202
2013 as employee number three with a focus on integrated modeling for mining applications in 2015 Steve took over over the role of president and CEO and has been straddling both the technical and business sides of aanti ever since thanks in advance Steve for taking the time to deliver today’s presentation
We’re all really eager to hear your perspective on the history of integration ated hydrologic modeling with hgs and what the future holds for hydro geosphere take it away that’s great thanks so much bradden so before we get started into the meat of the the presentation I wanted to start with a a
Few definitions just to make sure we’re all on the same page um and really the the main reason for this is the term digital twin has been gaining in terms of usage recently particularly in hydrology but it’s often used in many different ways and um sometimes s
Incorrectly as well so I thought it’d be good just to kind of ground us and have a common definition for what we’re talking about today so uh in terms of a model this is fairly straightforward I think we’re all familiar with this concept so I just went straight to the
Oxford English Dictionary for this definition it’s a simplified or idealized description or conception of a particular system uh often in terms of mathematical expression uh it could be put forward as a basis for theoretical or empirical understanding and for calculations and predictions so this is the camp that something like a
Hydrogeosphere mod flow Etc when you build up models with them would fall into in contrast a digital twin has a model as a component of what a digital twin is uh in addition to the model it’s an evolving set of data related to the object so that could be observations
From the field that are ingested on a regular basis and then a means of dynamically updating or adjusting the model in accordance with the data so this is some sort of simulation process and then you might use the digital twin on the fly to make predictions uh and I
Wanted to draw this contrast because um where we’re going with our models is to turn them into digital twins in terms of real- time hydrologic forecasting and so for this talk we’re going to uh kind of go give an overview of the history of integrated modeling what are the key
Strengths of integrated modeling look at some applications and then look at how do we turn them into digital Twins and then where is that heading in the future uh so I kind of just covered this outline we’ll talk about you know the history of integrated modeling what are
Models what are digital Twins and then some thoughts on on the future so um all integrated models integrated hydrologic models can draw their history back to this um Landmark paper by Freez and Harland that came out in 1969 it was titled uh blueprint for a physically based digitally simulated
Hydrologic response model and really they laid out that the framework for um how to build these models and how to conceptualize them uh this is figure one from that paper um the red highlights are just my highlights to emphasize the Bolding that they had in the figure because it didn’t
Stand out very well uh and what we see here is what their recommended path within the world of different approaches for uh modeling were to develop a physics-based digitally simulated hydrologic response model you can see that they considered lumped versus distributed even physical models analog models uh and this is kind of the
Recommended path forward that they headed back in 1969 and this is the path that models like modflow hgs Etc have followed in in this paper they identified um three major obstacles at the time we have to remember this is over 50 years ago so the science was
Still in its infancy uh there were three major obstacles that needed to be overcome before this blueprint could be fully realized the first one was do we understand the key water cycle processes well enough to represent them within a numerical model uh and I’d say over the
Past 50 years you know this has definitely been accomplished both for uh Overland flow as well as subsurface flow um you know both hydrologists and hydrogeologists have been working on these separately for decades and um that science is quite mature with an integrated model we’re bringing that together the knowledge from both camps
Uh similarly uh for the question of does the data exist to build such models um one of the questions that we often get from New modelers who have only used a surface water only or groundwater mod only model is well where do I find the data for the other domain and it’s not
That it’s not out there it’s just a lack of familiarity both disciplines have been collecting data or using expert judgment when there’s missing data for for decades now and recently um there’s been a big push for open data uh Canada is an excellent example of this with the
Canada one water initiative where uh national scale homogenous data sets are being made available that can be used to drive the construction of both surface water and groundwater models as well as fully integrated models so I think the second uh concern that freeze and Harland head has also been sufficiently
Addressed and then the last one they had is uh do we the computational resources to run the simulations in an economical Manner and you can imagine you know 50 years ago the Computing resources were nowhere near what they are today um and not just you know advances due to M’s
Law but there’s also been um you know advances in in numerical methods uh parallelization HBC Etc that has really brought us to the point where we can start to build these U large scale complicated nonlinear fully integrated hydrological models um and and deploy them in meaningful manners because we
Have the uh computational resources available so these are the three major objectives or um obstacles that frze and Harland identified and I’d say that you know we’ve addressed them um sufficiently over the past 50 years this graph just shows kind of the cumulative citation count um for the
Freez and Harland paper going back to 1969 and you can see for the first couple of decades had very little movement and then in the late 90s early 2000s uh really started to take off as integrated modeling particularly became more popular and I’m just going to
Overlay kind of the history of of hydro geosphere on on this growth path so back around 1995 maybe was ’94 uh fra 33 DVS was first published which which was Renee teran’s PhD which I think was finished perhaps in ‘ 92 um and that was really the basis of the subsurface
Component of hydrogeosphere then in the early 2000s uh the Overland domain and abouta transporation was added uh and that’s when it became hydrosphere then Hydro geosphere was paralyzed in 2012 by Dr hunai Wang as part of his PhD um Ed sidiki and the founders founded a quanti in
2012 and then um you know we’ve been working away with Hydro geosphere for several years and started to develop the forecasting platform hgs RT uh in the late 201 early 2020s and most recently uh well it’s not released yet but I believe January 2024 will have a gooey
For HDs released as well um and so it’s just interesting to see kind of the evolution of integrated modeling from our perspective as it overlays on the citation count in history of Freez and Harland so this slide kind of shows the bottom left here the conceptual model
For for hydrog geosphere it doesn’t have all of the features but it has a lot of them uh hydrog juser has a 3D variably saturated subsurface uh it has 1D domain 1D Channel flow as well as 2D Overland flow in the Overland flow domain uh it’s got a vapid transpiration so that’s uh
Root water uptake and transpiration as well as soil and uh surface evaporation winter processes so snow accumulation snow melt and so freeze th thaw as well as um you know many different types of subsurface domains as well such as fractures tunnels uh you know represent quite complex models um so once you have
The conceptual model you move it to a numerical model uh you parameterize it and then you get to your operational numerical model uh we’re going to kind of talk through some of these key steps in a little more detail um with particular emphasis on the conceptual model and how to think about um
Designing your problem so that it’s appropriate for a fully integrated hgs simulation before I get into the conceptual model in a little more detail um I wanted to talk a bit about the the domain coupling and this is really where um a tightly coupled model like Hydro
Geosphere shines and what makes it so powerful in comparison to a Loosely coupled model which might be um say like a surface water model bolted onto something like mod flow where it wasn’t two models that were designed to work with each other whereas hydrogeosphere from the ground up has been designed to
Be a a fully coupled Model A tightly coupled model fully integrated and the way this is accomplished is by embedding exchange flux terms within the flow equations for the different domains so here we Show an example for the porest medium domain and the surface water flow domains and within those governing
Equations there are exchange flux terms that allow for the free movement of water between domains when the equations are solved so we assemble all the equations into one Global Matrix and solve it simultaneously so that way we don’t have to iterate between different model types um it improves runtime mass
Balance accuracy Etc and so this type of approach is taken uh within hgs when we’re coupling any domains so there it could be fractures or 1D Channel flow uh dual procity dual permeability domains uh but as really kind of the core uh Gem of what makes hgs so powerful in the way
It solves these equations so what what does this mean in in practice if you’ve got um say a surface and subsurface domain that’s tightly coupled um you know what are the advantages of that where what it makes it unique and so one of the first things
That really jumps out is that it’s um much simpler to set up these models so the the surface boundary condition is is simplified considerably compared to say a groundwater model where you have to think about what is the recharge and estimate it and use that as a boundary condition with
Hgs we like to say you just rain on the model and it solves everything else and you know it is pretty close to that simple you bring in your precipitation uh rasters or time series your potential vapid transporation and let the model solve the movement of water between surface and subsurface domains so
There’s a free exchange of water there and recharge is calculated internally it doesn’t have to be prescribed to the model uh which is a real huge benefit in terms of uh setting up the models and and minimizing sources on cty um it’s also really good this Approach at uh reflecting complex near
Surface Dynamics such as perched water tables um you know if you had say a surface water model and a groundwater model that weren’t built to work to with each other and the groundwater model was fully saturated you wouldn’t be able to represent H water table conditions um this approach also excels in steep
Topography like mine environments or engineered Landscapes um and really good at looking at aapo transpiration and and your surface um Dynamics as I mentioned in the previous slide multiple domain domain interactions possible so it doesn’t just have to be surface and subsurface you can put in fractures you can put in 1D
Channels um you know we don’t recommend turning everything all on at once but you can definitely uh layer in quite a bit of complexity and the animation on the the right here is just an illustration of a simple cross-sectional model that is a Surface domain a subsurface domain and then discrete
Fractures within it and you see here that there’s some precipitation on the land surface the depression fills up it spills over into the next depression starts to infiltrate you get these perched conditions on a clay aquatard there’s fractures through that aquatard they start to activate and then you get
The water flowing through uh to the deeper water table um and so this just a nice simple illustration of how all these domains can kind of be packaged up together within a single simulation and really complex Dynamics can be represented within hydrog geosphere so then if we take a a step
Back um to thinking about how we design these models the you know conceptual model is is really the first place that everyone should start when when thinking about their models um we’ve got you know just some things to consider um this is by no means a comprehensive list and
Bradden’s got some uh webinars and sessions that dig a little bit more deeper into model setup but you know you want to start thinking about what are the questions that you’re trying to answer with your model um so that the model you know that should carry through and the model is designed appropriately
To address those questions um do you need all the domains what are the governing equations how do they get parameterized you want to make the model as complex as necessary but not overly complex if you can avoid it um think about what are the water sources um and
Sinks within your model uh if it’s a water shed scale model it could be as simple as you’re just raining on it and letting water come out with a vapa transpiration or flow to the domain uh for engineering type applications it could be much more complicated where you’ve got water routing pumping all
Sorts of you know engineered activities um and then also where are the areas of interest and what parts of the model should be refined in more detail so that you can get the resolution necessary to answer answer your questions in that area uh and the last comment here is
That you know you should strive for scale appropriate physical realism uh we’ll see a little later on what that means but the idea is that if you’re building say a a let’s say a continental scale model model you’re probably not going to be able to represent uh you
Know each and every little pumping well within that model and there might be ways you need to upscale that so you always want to think about the scale and um whether what you’re including in it is appropriate for the scale of the model some of that comes from experience
And some of that um you we probably have some examples that would help guide users uh on how to think about those problems as well uh some other considerations when when designing your conceptual model and these are just kind of ideas to get you thinking about some of the points on the
Last slide so you know if you’re working on a say a deep geological repository do you need to consider um unsaturated flow or Surface conditions with hgs you know we can do just a single domain so you can simplify the model considerably depending on your use case uh for
Problems that let’s say uh involve quantifying ET then obviously you’re going to want to carry a lot more refinement near surface so that you can resolve the the root depth profile and get a more accurate quantification of the evapo transpiration uh both spatially and temporarily uh maybe you’re looking at a contaminant
Transport problem you’ll want to think about you know mes mesh resolution uh maybe particle tracking can be of use before you go to full contaminant transport simulation to help guide that mesh resolution and refinement um and then we’ve got some other tricks within hgs where you can take an existing flow
Field and use that to drive a transport simulation so that you’re not solving both flow and transport at the same time it’s a way to really gain efficiencies and leverage existing flow simulations uh the other one we’re looking at here is winter processes uh this can also go for pet as well where
Sometimes there’s multiple methods for calculating different types of processes and so you really want to think through what is the appropriate method and do you want to do an internal calculation in hgs or if you want more control you can always do an offline calculation and
Feed that in an example of that might be be say using a full energy balance solution for uh snow accumulation snow belt that’s not really an hgs but you can compute it offline and then add it as a boundary condition into hgs um and we always recommend start
Simple and then add complexity as you get more comfortable with the model the main benefit of this is it helps to find any errors and setup or bugs conceptual issues um I put the asterisk there because if you’ve built the same type of model many times you can definitely jump
Further ahead as you get that experience but for newer users it definitely makes sense to start simple then layer in additional boundary conditions um you know domains Etc and check the model at each stage to make sure that it’s behaving the way you would expect it
To uh a few other considerations for the conceptual model um you know this we’re spending a lot of time on this because it is such a critical stage and a good conceptual model does guide the model construction process so you want to get that correct up front um likewise if you
Have a conceptual model that’s going to carry through your model build and application process and may come back to bite you later on um I like to think of the conceptual model as kind of a living guide I usually put it into like a PowerPoint presentation as like a
Dumping Ground of all my data sets conceptualization Etc and as new data comes in from a site you can update that and kind of keep it as a living conceptual model um one thing we recommend to and this is mostly targeting newer users who may come from
Different domains um is to forget about the processes that are internally handled within HDs and not forget about them completely but don’t try to control them is is the message here so um sometimes we see groundwater only modelers who are new to integrated modeling who try to prescribe recharge
Within a fully integrated model framework and it doesn’t work and it leads to all sorts of problems so understanding kind of how to build the conceptual model and then apply it to hgs as a critical step to make sure that you’re letting the strengths of hgs really Excel
Uh and not getting into some of these common troubles and pitfalls and Braden does cover that in his new user training course so if you’re uh new to hgs I would definitely suggest checking those out so that you can get some more in-depth insight into some of these
Considerations so that was kind of the conceptual model things to think about things to be aware of as we uh build up our models or prepare to build up our models um the kind of next area that talk briefly about is data sets um that
Was one of the question that came out of the free and Harland paper and you know definitely they would not have been able to imagine that these National homogenized uh data sets would have been available for model construction back in 1969 we’ve come a long way with open
Data uh what I’m showing here is an example of the data sets coming out of Canada one water uh many of these are directly applicable to a fully integrated model like hgs uh so once you have your conceptual model in hand you need to think about your uh model domain
Extents you know is it going to follow a shed boundary or is it following a site plan Etc so what are the implications for boundary conditions based on that choice um you know your surface hydrologic features both River Network and Lake bimet these are useful for constraining um your mesh development to
Make sure you’ve got refinement in areas where you need it uh demm is obviously critical as well and then with that in hand you can start to build your finite element mesh um and also bring in additional refinement in areas where where it’s needed then we get into hydrography and soils so subsurface
Parameterization and conceptualization uh land cover is Al also critical so that you can have appropriate vegetation and and um or land flow friction representation across the model and then you get into your forcing data to drive you know precipitation snow accumulation snow melt soil freeze Tha
Etc so once we’ve built the model I’m going to skip over the running the model stage we’ll assume that’s done uh and just look at some of the typical model outputs that come out of a fully integrated model um one that stands out as you know quite interesting is the
Recharge and discharge distribution over time uh so here we’re looking at an example from the Great Lakes region uh soil saturation is also one that comes out of hgs surface water flow rates surface water levels as well uh water table position groundwater levels Etc uh and then actual vapo transporation
And there’s a lot of other ones these are kind of some of the key ones that stand out um but if you’re doing solute transport obviously you can look at you transport distrib or distribution within the model as well uh and all these can also be used as calibration targets
Which is kind of interesting with hgs so um because it’s such a rich uh modeling framework there’s a lot of potential endpoints that could be used to help constrain a calibration uh and there’s a paper recently that came out by Shilling it all that looked at the benefit of what
They call unconventional observation data and and I’ll just read what they say here but it kind of advocates for the benefits of the additional outputs from hgs that can be used to constrain a calibration and they say in general the reviewed studies confirm that including at least one unconventional observation
Type of a relevant process for the modeled system and for the desired type of prediction alongside classical observations strongly reduces the ill posedness of the inverse problem uh that is it improves parameter identifiability and reduces uncertainty of predictions made with the flow model uh and uh one of the calibration parameters that have
Been really powerful for us recently is actual VAP transpiration uh and so the data that we would be uh calibrating to would come from flux Towers uh they’re becoming a lot cheaper and easier to deploy and in some regions that actual VAP of transpiration sink is a major component
Of the water water budget and being able to constrain that uh for calibration purposes can significantly improve model calibration and performance uh so I just highlighted that one because that’s one we’ve been working with recently but a lot of these uh outputs are very useful for for model calibration and when
You’re designing your model calibration you should think about what are the possible outputs from the model that would be useful for the type of questions and problem that you’re you’re looking at so the next thing we’re going to talk a bit about is where does hgs Excel so
Or fully integrated models in general but today we’re talking you know quite a bit about hgs um and really it comes down to uh you know if we’re talking about a fully integrated system issues where the surface water groundwater interaction is important um you know HDs can do a lot of groundwater only
Problems a lot of um deep subsurface problems where that surface water groundwater interaction isn’t important but if you’re really taking advantage of the uniqueness of hgs is going to be uh in the surface water groundwater interaction realm so things like climate change impacts uh land use change impacts agricultural studies are are one
That we do a lot of work with uh mining applications especially in in Canada it’s hard to uh operate a mind and not deal with surface water groundwater issues um design topography so this could be stream restoration Restorations s design and Analysis mining closure uh planning and assessment for mine closure
Ure uh it’s also powerful in um regions where maybe you have less data because it is a fully distributed model and we have a sense what the physically meaningful ranges are for a lot of the parameters that go into hgs plus the support from remote sensing and these
National scale data sets you can start to build models for ungaged basins and get you know meaningful insights from them it’s not going to be perfect but it is useful um and even in models where maybe it’s not necessarily for an Engaged Basin but you want in inputs at
Locations outside of an calibration point you can still get meaningful insights even if the model wasn’t directly calibrated to that location which is something you can’t do with like an empirical model um the other kind of key benefit of of hgs is that it can be applied across a very wide range
Of scales so everything from the small column sandbox model all the way up to the Continental scale uh and everything in between so there’s many many use cases is out there uh we’ve got a Blog on the website that kind of talks through a lot of the recent Publications
If you’re also someone who has a publication that’s come out with hgs contact Braden he’s happy to post that on there um what I’m going to do now though is walk through just a number of examples that cover different scales and really highlight um Power of hgs across a number of different use
Cases and I believe we’re taking uh a look across nine orders of magnitude and scale so really shows how versatile um this type of modeling framework is for addressing problems of different sizes uh so we’ll start from the small scale and work our way up so at the the very
Small scale um here we’re looking at less than a meter squared typically for these types of problems you’re really digging into detailed processes uh trying to gain some understanding about the fundamental uh physics of the problem you’re looking at in this case we’re looking at a study that came out
From Beth Parker’s group at the University of galf where they looked at the backward diffusion problem where you have a solute that diffuses into low permeability material and then even after it’s been flushed you get this back diffusion out of the clay uh and that’s what we’re showing in the from
The paper here in the pictures uh the top is just showing um the visual representation of the Ploom the bottom is the numerical comparison and then on the right we see the simulated versus observed concentrations coming out of the sandbox as well so showing that you
Know for very small scales this is a groundwater problem with transport um hgs is is very amenable to and applicable and then as we get a little bit larger here we’re looking at uh detailed soil moisture forecasts uh for agriculture specifically and in this case we’re using uh 1D column models
That are 5 meters long uh and this is a a fully integrated model so it’s got surface water ground water uh vable transpiration built in uh and allows for resolving in high detail the soil moisture profile within that soil column um and in this case it pulls from those
Open data sets I was talking about soil properties as well as um like crop type uh these can be overridden by the user and then you can generate high resolution soil saturation profiles and forecasts uh for a point of interest um so that’s at the order of uh say several meters in
Size uh if we step up to go a little bit larger we’re looking at hundreds of meters in size um and in this case is an example with density dependent flow and transport looking at uh saltwater intrusion uh along the coast and particularly they were trying to understand the impacts of climate change
And if you had engineering structures like a dyke how that would influence um water quality and that saltwater intrusion into the system over time um and you can just see here we’re looking at the extent of the salt water in the middle and then the calibration between simulated and observed head so
This is at the order of several hundred meters in in scale again it’s going to be fairly high resolution um and at a scale we’re doing solute transport with density dependent uh effects is is Meaningful and and can be applied as as we go a little bit larger
We get to the order of uh several kilm squared uh and this is an interesting paper where they built a model of a real site and use it as a virtual laboratory to look at um feedback between different features of the Watershed specifically the slope of a watershed and the amount
Of nitrate um that was loading to the streams and running off and what they found is that there was a positive correlation between uh catchment slope and nitrate loading to the streams because as it becomes steeper there’s less opportunity for the nitrate to infiltrate and degrade um and so it runs
Off in there’s more loading to the stream in these steeper W sheds so this is a neat example of you know you can take a real world problem but then modify it as a Virtual Lab to start to look at um some of these relationships between processes within the
System uh and again this was a solute transport model on the order of you know several square kilometers moving a little bit larger this is a a mining application of 8 square km uh looking at Reclamation of an oil sands mine um and this was one where they used the the thermal
Transport capability with an hgs to look at the impact of soil freezing uh on runoff behavior and uh solute loading if you will um with and without freezing conditions uh the graph at the bottom right shows the impact of that it’s a little bit pixelated my apologies uh so
The solid line that’s dark as observed the lighter solid line is with Frozen and then the dash line is without Frozen soil uh and you can see that there’s a huge difference in terms of the ability to represent the observed uh runoff characteristics really showing the importance of considering soil freeze
Thaw in in Northern regions and what’s interesting is not just is there a mismatch where the fresh at there in the spring but also that mismatch carries through for the rest of the spring where we get higher runoff Peaks uh for the non-frozen condition because there’s
Still a lot more water in the in the system and what they found is that by considering um the Frozen soil conditions you actually get less chloride being released to the environment over an eight-year period than if you hadn’t so this can have significant impacts for assessing um
Engineer design plans and so making sure that appropriate processes are included in your numerical model is absolutely critical when thinking about your conceptual model and and model setup uh moving a bit larger this study is at the 190 Square kilm scale uh this was a groundwater only study but used uh
Backward transport capability within hgs to look at uh domestic well vulnerability from fracking operations within a water shed and from this assessment they’re able to produce a vulnerability map for the Watershed of Interest so that they can start to focus on you know key Wells for more detailed
Investigations to make sure that the Water Supplies are protected uh going a bit larger we’re now up to 2,000 square km so about 10 times jump um and now we’re getting to the scale where we need to think about um Sub sub grid features so you may have uh
Processes that are operating at a scale that are smaller than your mesh resolution in this case it was specifically looking at real storage uh for which is a subgrid dep depression so you might have little Pawn or potholes that are smaller than your element size and this is parameterized
In hgs through a process called real storage uh and this can be represented uh spatially distributed in a spatially distributed Manner and that was a key feature that was developed through this paper by Steve frad all and they use that to look at what are the impacts on you know real storage
Distribution and how that affects runoff so the figure on the right shows um the spring freshet and the impact of having more real storage in green reduces those Peak flows and so this has implications for Land Management and Wetland management to make sure that or shows the benefit
Of keeping these depressions in place um to mitigate you know Peak flows and and hopefully offset flood risks as well and so these types of studies can be informative for decision makers when they’re thinking about how to manage their Watershed maybe incentivize stakeholders to take certain actions to mitigate
Downstream runoff and and flooding potential uh so then now if we go a little bit larger to the Grand River Watershed which is where um we’re sitting today with in aanti uh this is a 7,000 square kilometer model and this was one of the first ones um to really
Lay out the framework for how to look at climate change impacts with a fully integrated uh surface subsurface model and that was Andre erer LED that study um and really nice methodology in there for anyone who’s looking at um you know applying climate change impact analysis
To a fully integrated Model H this slide just shows some of the sample outputs from that but looking at you know changes in soil saturation under future conditions or average groundwater recharge average depth to water table you can look at stream flows it’s really anything that can come out of the fully
Integrated model you can do a you know future condition versus current condition and look at what the potential change ches and impacts are to help uh guide planning and decision making uh for the future um I think we’ve got just a couple more left for these sample
Applications but I just really wanted to highlight you know the the different scales across which hgs can be applied to and how the types of questions that are being uh looked at change as we go to the larger scales uh this one is uh looking at the athabaska River Basin
Which is about 160,000 square kilometers uh and specifically uh anti- weing it all we’re looking at the groundwater contribution across the water shed uh that’s what the plot in the bottom right shows so the red is surface flow and the oh sorry the blue is surface flow and
The the red is the groundwater seepage uh this study found that at the downstream end the mean annual annual contribution of groundwater to flow conditions was about 45% which was in good agreement with uh local isotope studies um and for some of these large scale systems really the
Only way to start to investigate um some of these groundwater contribution estimates is through uh numerical modeling and we’ll see that in the next uh study here which is looking at the the Great Lakes um and this is a model that uh Bruce Shu and team built up to
Look at um groundwater contribution and seasonality of groundwater contribution into the great lakes and this was a very large model 166,000 Square km so obviously you know um there’d be some upscaling not uh as much detail as you’d carry in a smaller model but can still be meaningful to address some of these
Big picture questions and um you know it’d be very difficult to go out and measure groundwater discharge along all the Great Lakes but with a numerical model that’s calibrated you can start to gain insights on you know what is the contribution to groundwater and what are the seasonal dynamics of it and what
They found is that there was quite strong season ity with uh groundwater contribution being highest in the winter months and and lowest in the the summer months okay so the last one we’re going to look at here is uh just very quickly the Continental scale applications um and this is a very large
Model 10 million square kilometers Jeremy Chen did this as part of his PhD study and you know I think it’s fair to say this is kind of a demonstration proof of concept model at this scale uh but a lot of the lessons learned are supporting the Canada one water
Initiative particularly around uh data management uh data collection the value of homogenized data sets and now for the Canada one water initiative they’re actually taking in Breaking this model domain up into still large pieces but into smaller pieces so that higher resolution can be carried within the
Model so so that was just a quick tour of kind of the uh some of the use cases of hgs to get people thinking about you know different types of applications different scales um I think importantly you need to kind of match the scale of the model with the questions that you’re
Asking and this table or this this figure is just a kind of a summary is by no mean set in stone but you know at the small scale you’re typically looking at especially the very small scale looking at models that help to support process understanding maybe running experiments
To support upscaling probably going to be looking at you know solute transport soil moisture Maybe fracture Dynamics very small scale detailed uh process models and then as we get larger the questions start to change a bit because you obviously can’t carry the same level of detail so you need to start to
Upscale um some of those small scale processes um you know as you get to the site scale you can start to really look at you know engineering questions for maybe it’s a detailed engine engineering design mine application um stream restoration as you get bigger um you know you might be
Getting to the Watershed scale you can still carry some local detail including say mining applications or industrial activities um but the questions may start to change or you may start to get into some upscaling representation especially as you start to get above several thousand square kilometers and
Then when you get to the large Basin scale um you’re starting to look at big picture questions you know climate change impact land use management um any smaller features are going to have to be upscaled in terms of their representation in in some way as well
Um just to kind of cap off the the model phase of this talk just wanted to show you know integrated models and here we’re just looking at Hydro geosphere and their mentions on Google Scholar over time so this curve kind of matches what we were seeing for that freeze and
Harland where definitely we’re seeing uh increasing uptake within the community uh as hgs and integrated models become more more widespread and more widely used I know we’re getting close on time braen so I’ll try to get through this last section uh quickly here um so just coming back to the the definitions that
We had initially with model and digital twin so I I put digital twin in the camp of you know moving beyond frez and Harland so Freez and Harland really focused on the model aspect but the digital twin piece has taken the Freez and Highland framework and extending it into an operational decision support
Type framework um and that’s where we have a a real-time forecasting PL platform called hdsr and it takes an HDs model and turns it into a digital twin this figure kind of shows a schematic of how that how that’s done so at the heart of the platform we
Have an hgs model of the area of Interest usually a watershed uh we ingest data from the field uh so it can be remote sensing Stream flow soil Mo or ground water level data that is used um to update the model to have it reflect current conditions and then uh we use
Weather forecast to drive the model forward to provide uh forecasts from that hydrologic model and so with this we’re you know pretty close to that definition of a digital twin words it’s living is bringing in data from the field and then it’s providing insights back to the stakeholders and decision
Makers and we’ll just look at a few kind of sample outputs that come from this uh so here’s a you know surface water depth distribution map um so any of the outputs that come out of hgs can be displayed uh we’re looking at you know Stream flow through time as well the
Solid black line is the observed and then we do a probabilistic forecast so you get the median response shown in red and then the 10th and 90th percentiles in green uh these types of forecasts can also be done for uh groundwater levels uh soil moisture so really anything
That’s within the fully integrated model framework can be pulled out in a forecasting uh application and displayed or shared with the user uh recharge is another one as well as as groundwater discharge can be displayed and and shared um and then recently we’ve added in uh water quality and this is a
Machine learning water quality forecast where it’s been trained on historic water quality data as well as historic hgs simulation model output so this is a really interesting pairing of machine learning with a physics-based numerical model that can yield um you know pretty interesting insights and provide water quality forecast that otherwise would
Have been intractable with trying to solve the inve dispersion equation at a at a watershed scale so primarily we’ been focused uh on building this up in Canada you know it’s you always want to get started in your backyard and test things out um we currently have a number of different
Deployments across Canada our first one was with the sou Nation water shed up near Ottawa that’s about 4,000 square kilometers uh We’ve also got the southern Ontario platform which is about 76,000 Square kilm the South Saskatchewan River Basin is about 120 uh cineo River Basin is about 150,000
Square kilm and then the pem valley is uh just over is about 13,000 square kilometers and what you’ll notice with um a lot of these larger platforms is we’ve taken and broken the models up into smaller pieces typically around 10,000 square kilometers so that we can better answer local questions with the
Forecasting models instead of having these very large uh River Basin Scale Models So currently we’ve got uh 30% coverage of Canada’s habital landscape and about 38% of Canada’s population covered with these forecasting platforms um so you know the question is why would we turn a fully integrated
Hydrologic model into a digital twin um you know one one of the early motivations was to get out of the routine of building a model writing a report and then having it sit on a shelf and die we wanted to you know turn these into living living tools that can uh you
Know really allow the the client or stakeholder to get the most out of their investment um you know some of the other interesting things that come out of it is you may get into new stress scenarios where you can get more insights on your model ability and so find new mismatches
Between simulated observed and improved calibrations um and then also can support you know um site operations and management and really you know the focus is on decision ERS who maybe aren’t technical in their own right but uh or technical enough to dig into say fully integrated model but can Val gain value
From the outputs that come out of the model and so bridging that gap between you know really detailed technical output from HDs to a simple to understand simple user interface was a big motivation in the early stage of building up this platform as well um so some use cases uh you know
Flood and drought forecasting are some obvious ones especially as seasonal weather forecasts become more common um field sampling and monitoring planning and guidance so using these models and to help guide decision-making in the field um lots of potential agricultural operation or applications as well um mining applications anywhere where
Realtime uh insights from a model could be useful we can also layer on things like automated alerts so if a forecast is showing that uh a variable of interest is going outside of certain bounds you know you can get an email or text to be alerted and then also these
Forecasts can be pushed into other dashboards and Water Management tools it doesn’t just have to be through the hdsr website you know lots of integration with other platforms um so that was kind of the the digital twin side of things I know I went through that a bit fast just
Cognizant of time here so I just got a few slides to wrap up kind of looking at where we’re heading with the science uh kind of what the future is going to bring so if we take a look at the model side of it which is hydrog geosphere um you
Know we’re continuously developing new features and processes some interesting ones that are coming up are uh Dynamic meshing which allows us to modify node elevations during and change material properties during a simulation to represent say mining or land subsidence without having to do a snapshot simulation approach um lots of stuff
Around Water Management uh for routing water within the model or irrigating props on demand those sorts of water management applications uh we’re also looking at coupling with other models uh whether that be you know climate or weather models and land surface models uh we’ve done some of that in the past but the
Coupling should be a little more generic so it can work with more models more easily uh also uh standard hydrology models for for flood inundation like HEC Rass um you know hgs outputs can be used to drive those models and getting that kind of pipeline and process automated
Is something we’re working on other models of Interest would be sediment transport reactive chemistry and since there’s a lot of interest in coupling with hgs really we want to step back and look at that kind of coupling framework in general um other things coming up would be the hgs guey which I mentioned
Earlier so that’s in partnership with aquaveo hgs will be plugged into the GMS guei and that’s going to be pushed out in early 2024 um and then the last one worth mentioning is just the can and water uh initiative is is wrapping up in early
2024 as well and a lot of those data sets if they’re not already publicly available will be pushed out to the public as well which can really support uh rapid deployment of hgs modelsa across Canada uh when it comes to to real time um some of the things we’re working on
Or the digital Twin Side some of the things we’re working on are 1D soil moisture models and forecasting so we talked about that a little bit uh earlier but having the ability for the user to to trigger the models on demand customerize the parameterization and then get the results back we’re also
Looking at extending water quality forecasting capabilities that we talked about um to other parameters but also you know there could be use cases for forecasting for contaminated sites with plumes where there can be real-time feedback between field sensors and uh maybe operations around pump and treat or or plume
Management um coupling hgs with machine learning is a an active area of res search for us um we see a lot of opportunity for using hgs outputs to inform machine learning training as well as supplementing some of the forecasts with using machine learning for short-term forecasts but then rely on
The you know physics-based numerical model for longer term predictions where those tend to to perform better uh and then also allowing just more user interaction with the web platform for entering say material Properties or water management scenarios uh s Dam operations those sorts of things where
There’s more of a input from the user to help guide the the forecasts that are going to be run as well and I think this is where all wrap up um so just a few thoughts on kind of what what we need in the field um first bullet training training training uh I
Think there’s a lot of opportunity in University programs to get um students thinking about surface water and groundwater as a holistic uh water cycle which it is we all know it is but often times it’s still siloed in universities uh and really we need to you know find a
Way to to kind of break down those silos and getting you know students who know both domains being graduated um Additionally you know there’s unfortunately a trend towards decreasing the quantitative requirements um in some of these programs and I think you know we really should maintain them or
Increase them if possible as they have slipped a bit over the the past decade or two um and also introducing integrative modeling into to courses so the release of the guey should really help accelerate that um but we do offer licenses to anyone who wants to teach
With hgs you can get licenses for free and we’ve got some great examples of that locally with uh Andrea Brookfield at the University of watero and Janna levenson at the University of galf teaching with hgs in their curriculums um and then also just you
Know there seems to be a bit of a gap or disconnect with um the decision makers and stakeholders uh in in government Regulatory Agencies so figuring out how to bridge that Gap through education and Outreach and increased stakeholder engagement uh just so that everyone’s aware of what is possible with modern
Tools and what the state of the science is um so with that um there’s a bunch of references for the uh example applications and other kind of interesting big picture papers that were shown um if anyone’s interested uh reach out to bradden he can make this
Available to you and with that uh I will stop and pass it back to bradden for any questions thanks Berg Steve that was that was really good um I haven’t seen oh there’s just boom a bunch of questions just landed um so maybe we can go through
These one by one Steve if you have access to the chat I’ll just let you kind of read through them and answer them as you like how does that sound uh yep one second let me just see if I can find the chat here oh yeah there’s a lot okay I got to
Get to the top uh oh okay yes uh slide deck um will be the the presentation will be posted I don’t know if we’re going to uh share the slide deck uh explicitly Braden I don’t know if you can answer that one or not if we typically share the SL um it’s
Just the last post has a long series of questions some of which were already answered in the in the present presentation but I’ll just maybe make a note about the slides um it usually takes me a day or two to sort of format the video and get it uploaded onto
YouTube at which point I will send a follow-up email to everyone who registered uh just providing you with the link to the recording so you can review at a later date since there is a lot of interest in having access to the slides I’d be happy to share a PDF of
Those uh with you at that time as well so just keep an eye on your inbox and uh I’ll definitely be following up with those things um I can since I’m already talking I will just make a couple of other quick notes I just wanted to point
Out Steve did mention this already but I I want to point out again that almost all of the sample application slides that you saw presented here uh they are they can be found these papers and and short sort of um research highlights can be found on our blog um several of those
PR of those research papers also have webinars that that we’ve recorded so for example the saltwater intrusion sample application that Steve shared we actually just hosted a webinar maybe a month and a half or two months ago with the lead researcher um so check our YouTube channel there’s plenty of past
Webinars and you know presentations with focuses on on different types of sample applications and with that in mind I also just want to point out that we have one final webinar for the year um scheduled for next Wednesday from 11: to 12 p.m. or 11: am to 12:00 pm Eastern
Standard Time um that webinar is going to be delivered by some collaborators of ours at Ducks Unlimited Canada we built uh with them a watershed scale model of the the dog Lake Watershed I’m not exactly sure where that would fit in the scale the size scale range um but the
The focus of that research project was really on evaluating the impact of wetlands on maintaining base flow to streams so that’ll be another interesting webinar for you before the end of the year um those are all my notes other than a thank you of course for everyone
For attending I’ll pass it back to you Steve now that you’ve had a chance to look at some of those questions yeah thanks Bron I just chewed through here it looks like so most of them look like um were from one person but I’ll I’ll
Answer a few of them but I think tahara probably be good for you to connect with Braden as well to dig into detail on some of these um but I think one kind of interesting question is is about Computer Resources and like what the computer specification would be for
Running some of these models uh so typically we run our models on high-end gaming PCs so you know good i7 chip uh maybe 16 to 32 gig of RAM uh SSD hard drive nothing fancy um we do start to do our calibrations on larger servers and stuff now but for a single individual
Run uh a good gaming PC uh is more than sufficient um for that I see another question here from Ying about visualization um so Tech plot is you know the tool we use in house primarily but we do support pairview as well which is uh open source and it’s kind of like
The open source version of tech plot and then also through the aquaveo GMS uh guey integration they will have some visualization tools available there as well um see another one here about hydrography so it it really depends on the um location you’re working in if it’s um you know’s say large scale
Watershed we’ll probably start with um open open data sets so like in Canada there’s the Canada one water data sets now that can get you base hydr Strat and then you could refine with local information um if you’re working at a say a industrial site typically they’ll have um Hydro statgraphic information
That you can use as as well to to build your models uh we tend in house tend to work through um GIS and Leap Frog but there’s lots of other um different tools available out there as well that can then be imported into hgs for building up your geological
Model um okay just looking at the next question here from delonga uh oh sorry it’s moving up the screen uh for water quality is it possible to model gluten concentration Based on data of the groundwater near the body of water yeah so we you can use the inion
Dispersion equation to simulate um the movement of solutes within the system obviously it’s going to depend on your scale the amount of information you have available to Define your Source zone is it you know point source or non-point Source what are the Dynamics of your
Solute of interest but there is a lot of literature um that talks about hgs being used for water quality modeling uh and if you have any questions um in addition to that please reach out and we can help with that um any other ones you want me to jump on
Braden or just um yeah I’ve been trying to answer a few here as we go I think given the long list of questions from tahara I’ll probably maybe just have a follow-up call with her to go over some of those uh considering many of them were already answered as well um but I
Think we’ve answered most of them oh there’s a couple of new ones coming in here so is hgs compatible with Lea frog or can it import a Leap Frog model directly yep so I can take that one so at the moment um it’s compatible in that you build up your hydrographic model
Within LeapFrog and then you can export the surfaces and then use those as layer control within hgs um another workflow we’ve had is to build a uh fee flow model uh within like build a f flow grid within Leap Frog export it and then map the uh Zone distribution within HDs so
There’s a couple different ways that it can be done um next one is talking about surface water groundwater coupling parameters um right so for that there’s a term that uh a key term called coupling length and in that case um it there’s there’s actually a study out I
Think it was around 2010 um I can’t remember if it was Jessica Le or perhaps someone else but um they looked at the sensitivity of models to coupling length and um in general they found that is with if you’re within a fairly like within a reasonable range it’s fairly
Insensitive over several orders of magnitude I think we typically use on the order of you know 1 cm to maybe 10 cenm it is a little bit scale dependent so larger models you might go a bit larger smaller scale models you’ll go a bit smaller but there is some literature
Out there on um setting up those values one so okay yeah I haven’t seen any other questions come in I just wanted to make one um extra little comment just about uh visualization capabilities in addition to PIR review we have also recently been working on some commands that would allow us to
Export hgs model outputs as a point Cloud uh which would allow users to just visualize hgs outputs directly in GIS that’s not quite available yet but uh should be coming down the line pretty soon um we just need to do a little bit more testing on that
And yeah there were no other questions so I guess I’ll just wrap things up here we are after the hour thank you everyone um for attending the training course or sorry for attending the webinar today if you are interested in learning more about hgs we have a webinar next week we
Have a free training course tomorrow afternoon and if there are any other questions that come to mind after the web you know after the meeting ends here we are available to answer questions anytime just reach out through our our emails again my email is listed in the
Chat you can also email us at info aan.com so uh one final thank you very much and have a great day thanks for hosting this Bren and thanks everyone for attending