An RF machine-learning model was developed to predict lithium concentrations in Smackover Formation brines throughout southern Arkansas. The model was developed by (i) assigning explanatory variables to brine samples collected at wells, (ii) tuning the RF model to make predictions at wells and assess model performance, (iii) mapping spatially continuous predictions of lithium concentrations across the Reynolds oolite unit of the Smackover Formation in southern Arkansas, and (iv) inspecting the model for explanatory variable importance and influence. Initial model tuning used the tidymodels framework (52) in R (53) to test XGBoost, K-nearest neighbors, and RF algorithms; RF models consistently had higher accuracy and lower bias, so they were used to train the final model and predict lithium.
Explanatory variables used to tune the RF model included geologic, geochemical, and temperature information for Jurassic and Cretaceous units. The geologic framework of the model domain is expected to influence brine chemistry both spatially and with depth. Explanatory variables used to train the RF model must be mapped across the model domain to create spatially continuous predictions of lithium. Thus, spatially continuous subsurface geologic information is key, although these digital resources are often difficult to acquire.
Interesting to me that RF performed better the XGBoost, would have expected at least a similar outcome if tuned correctly.
However, kriging is really quite difficult to use with non-continuous inputs. RF is a lot more forgiving there. You don't need to develop a covariance model for discrete values (or a covariance model for how the different inputs relate, either).
We had this discussion a couple of days ago: "Why do Random Forests Work? Understanding Tree Ensembles as Self-Regularizing Adaptive Smoothers".
https://arxiv.org/abs/2402.01502
I would hazard a guess that with better tuning, XGBoost would still have won. (The paper notes that the authors chose a suboptimal set of hyperparameters out of fear of overfitting - maybe the same logic justifies choosing a suboptimal model type...)
I haven't read in detail what their validation strategy is but this seems like the kind of problem where it's not so easy as you'd think -- you need to be very careful about how you stratify your train, dev, and test sets. A random 80/10/10 split would be way too optimistic: your model would just learn to interpolate between geographically proximate locations. You'd probably need to cross-validate across different geographic areas.
This also seems like an application that would benefit from "active learning". given that drilling and testing is expensive, you'd want to choose where to collect new data based on where it would best update your model's accuracty. A similar-ish ML story comes from Flint, MI [1] though the ending is not so happy
[1] https://www.theatlantic.com/technology/archive/2019/01/how-m...
It's in a caldera in a mountain that I-80 bypassed to go through Winnemuca, Nevada. Nearest town is Mill City, NV, which is listed as a ghost town, despite being next to I-80 and a main line railroad track. The mine site is about 12km from Mill City on a dirt road not tracked by Google Street View.
Google Earth shows signs of development near Mill City. Looks like a trailer park and a truck stop. The road to the mine looks freshly graded. Nothing at the mine site yet.
It's a good place for a mine. There are no neighbors for at least 10km, but within 15km, there's good road and rail access.
"to shut down the tar sands, we actually have to shut down the tar sands, not just blow up other mountains elsewhere and hope that leads to the end of the tar sands."
https://maxwilbert.substack.com/p/the-long-shadow-of-the-tar...
Fortunately, checking to make sure the entire Internet does not have a website disagreeing with the decision to start a mine, is not part of the process by which mining is started.
https://www.usgs.gov/media/images/locations-yellowstone-hots...
I specifically went out of my way on a trip a couple years ago to check out Thacker Pass to see where this planned Lithium mine was going. Unfortunately there was thick smoke followed a significant thunderstorm as a front came through and I didn't get to explore much.
We are in a crisis of climate change, biodiversity and habitat loss. Thacker Pass is critical wildlife habitat for threatened, endangered, and endemic species including the greater sage-grouse, pronghorn, Lahontan cutthroat trout, and golden eagles. Thacker Pass, known as Peehee Mu’huh in Paiute, is sacred to regional Native American tribes.
It’s too late to prevent Phase 1 of the Thacker Pass Lithium Mine, but there are opportunities to help prevent Phase 2. More broadly, we hope to protect the rest of McDermitt Caldera from Southern Oregon down to Thacker Pass from catastrophic lithium mining.Searching in Google Maps, Thacker Mine comes up as 40.58448942010599, -117.8912129833345. As you say, that is near I-80 and Mill City, and there is nothing there.
But Wikipedia says it's at 41.70850912415866, -118.05475061324945 in the McDermitt Caldera, nowhere near Mill City or I-80.
I'm thinking probably don't trust Google on this one. :)
"Lithium Americas will contract with a bus company to drive workers an hour to the site for 10-hour work shifts, he added. An additional two hours will be added for transportation time. If you go to work on our project, you will have free room and board and free transportation to the site every day. You would get three free meals a day." If you're an unemployed coal miner in West Virginia, that might look good.
[1] https://www.nevadaappeal.com/news/2024/oct/12/nevada-operati...
4-5 digits should be enough for any use outside of surveying, that's a precision of 10 meters and 1 meter respectively.
Even Wikipedia is making me suspicious by using hundredths of arc seconds, despite linking the document that came from. How do you localize a mining site down to a single foot?
It is interesting to see how much of this data could be modelled based on wastewater brines from other industries in the area, assuming we go on to mine the lithium it will say a lot if the ML predictions prove accurate.
One thing I couldn't tell, and its probably just a limitation of how much time I could spend reading the source paper, is what method would be needed to extract the bulk of the lithium expected to be there. If processing brine water is sufficient that may be easier to control externalities than if they have to strip mine and get all the overburden out of the way first.
This mining offsets mining for other things that is happening at several orders of magnitude larger scale. Oil, coal, gas, etc. mining is huge and lithium batteries plus renewables are already reducing the need for those. So, the transition to renewables and batteries might actually result in a net reduction of mining.
Of course doing lithium mining cleanly and responsibly is an important topic. Especially in places close to where people live. But considering the vast amounts of other stuff we mine already at a much larger scale than we'll ever need to mine lithium, this is a drop in the ocean.
And of course the lithium that is mined can be used and recycled over and over again. Once it is in circulation, we'll be re-using it forever. And given the improvements in battery tech, production processes, etc. the amount currently in circulation is likely to power a larger amount of battery capacity when we do recycle it eventually. Even when considering inevitable losses during recycling.
Lithium recycling processes are working fine already of course but there's very little recycling being done at scale for the simple reason that most lithium batteries in use are still very young and quite far away from needing any recycling. If anything, the improved life times of batteries is pushing the date that we need to be recycling at scale further and further away.
Extraction methods very much depend on composition of the deposits and whether they are in brine or other form and what other materials are present. There's a wide variety of brines, rock compositions, clays, etc with some lithium in them.
This point is overlooked so often in these discussions. Lithium is not a consumable in batteries, whereas oil / tar / coal etc. is. So, we do some ugly mining for a bit, and then basically stop once we have the lithium we need for use in batteries over and over again. It’s a completely different model than extract-and-burn.
It's mining brine. I.e. the "mines" are basically deep water wells.
The limestone itself doesn't have any lithium. It's the water in the pores in the limestone that is relatively concentrated in lithium.
In most of these cases, you're already producing brines from the smackover formation as a part of existing oil and gas production, but the brine is being re-injecting after oil is separated from it. The idea is that it's better to keep those and evaporate them down for lithium production.
That does require large evaporation ponds, generally speaking, but it's not strip mining.
As far as evap ponds go, are there usually chemicals or elements in the same brine water as lithium that is important when evaporating into the atmosphere?
Do you have the same trepidation about aluminum, iron, dish soap, and table salt? I ask because the amount of "ripping open" involved in lithium production is like a speck in the eye of a whale compared to all the other mining. In terms of scale all existing and proposed lithium mines are teensy tiny by the standards of mines.
Lithium supply is not an issue. Here in oz we have plenty, there is surplus in market (see current lithium prices).
Conversion however is an issue, majority of plants are in China. Build some refiners that turn it into lithium carbonate and oz will fill them.
All those minerals. All that sunshine. Terrific combo.
h/t Saul Griffith.
Sulfur, currently extracted by desulfurization of oil and gas, gets more expensive in the post fossil fuel society, but there are other sources (like pyrite).
Global reforestation is almost entirely the result of households switching from wood to coal in the 20th century.
This is ludicrously off-base for fossil fuels, even if we're only talking about local pollutants from the plants themselves, nevermind things like Exxon Valdez or the pipelines or the act of mining. Nuclear seems likely, though as the other commenter noted it's not a magic bullet either.
> Global reforestation is almost entirely the result of households switching from wood to coal in the 20th century.
This is a European phenomena mostly, and is a result of urbanization mostly.
It's high time we realize that Pax Americana is our era to lose, (re)start mining and (re)start development.
we don't need it happening upstream.
and watch as the nations destroy themselves (ecosystems)
Very good work - but typically we don't build prospectivity models this way (or rather we don't validate them this way anymore). Great to see the USGS starting to dip their toe back in this though, they and the GSC were long the leaders in this, but have dropped it on the last 5-7 years.
In the US environmental regulations, the cost of producing power, labor costs, would all drive up the price of the end product in a way that makes it totally noncompetitive. That's also why the US and some other countries are investing in other ways to find lithium (among other things) on seabeds, where it's hoped that extraction would be less expensive. Of course the threat to the seabed environment is a concern, which in turn might drive up prices by imposing regulation, etc etc etc.
In an export model, yes. However, given their negative externalities (including geo-political factors), importing countries may place tariffs on Chinese lithium in order to make use of other sources.
If the total embodied value of lithium in any particular product is small compared to the overall value of the product, the tariff might not represent a significant drag on the indigenous industry.
As proof of that there are sodium-ion batteries on the market right now, but they're not price-competetive yet[0] despite using largely the same infrastructure.
[0] The potential is there though as they have an important advantage: you can safely discharge a sodium-ion battery to 0V for storage/transportation.
So yea desert sand is essentially free, even if you pay for shipping.
To be honest, the energy problem is more or less a solved problem with the current technologies we have. We just need to accelerate our pace of adoption to hard-reverse on fossil fuels (except Germany). We already have large reserves of Uranium, of which only a small amount is needed to fuel a power plant. We already have lithium battery tech to store the power. We already have solar panels being mass produced and adopted to fill in the gaps. All we need is connecting the dots and making sure these resources play well with each other in symbiosis.
I'm skeptical. China is already mass-producing batteries, securing as much lithium as possible. Additionally, US regulations will significantly increase costs for battery manufacturers.
Does that mean the entire field has enough lithium for the requirements of 2030, 9 times? Or in other words, it can supply the lithium needs of car batteries from 2030 to 2039? That's not particularly long...
Look at steel. Most of the steel used is recycled steel, we don’t mine a lot of it any more. If you asked someone 90 years ago, they would have assumed global steel demand would continue to rise.
https://www.nytimes.com/2024/10/21/business/energy-environme...
If not that’s funny timing given that was a few weeks ago
Given the mood alerting properties of lithium, are people living here chiller than would be expected (controlling for instance for poverty / SES) ?
[1] https://www.kcl.ac.uk/news/lithium-in-drinking-water-linked-...
I think I heard that long term usage of lithium has nasty side effects like damaging kidneys, but perhaps not at these very low concentrations.
[1]: https://www.sciencedirect.com/science/article/abs/pii/S01691...
I agree it's not likely you'd get a measurable effect from the local groundwater.
[1]: I'm working on a DB of water quality, https://www.cleartap.com/water-systems/AR0000550
That said, I honestly am unsure. It also is a requisite that it must be in the water in sufficient but low amounts
Source: am bipolar and take 600mg daily.
So, the so-called therapeutic dose of lithium is merely a sub-toxic level, and must be monitored by frequent blood tests.
There are horrific side effects from using lithium in the long term, including convulsions, hair loss, diarrhea, suicidal and homicidal ideations, and extreme thirst (polydipsia).
So personally, I would rather not be tapping into lithium reserves for my health.
ML is a toolbox of methods. Not every problem needs a transformer.
They do if they want to get the intention of a Venture Capitalist!
The USGS predictive model provides the first estimate of total lithium present in Smackover Formation brines in southern Arkansas, using machine learning, which is a type of artificial intelligence.This is what it was called back in the day. https://link.springer.com/article/10.1007/BF02478259
It would be amazing for some low weight, low volume, high energy density, high discharge rate, high charge rate, cheaply manufactured from abundant materials, low thermal sensitivity, high thermal tolerance, low passive loss, non-explosive, high cycle count, low memory, shelf stable battery chemistry to appear, but thus far every one fails in several of the categories.
People in the U.S. would rather be slaves to China than be self sufficient as we once were...
Or are they literally just announcing that "Hey, we told the computer to tell us something, so it told us something"? Yes, that is how it works. The computer will always tell you something if you make it tell you something. That isn't the hard part. The hard part is getting it to tell you things that correspond to reality.
In the absence of validation, this means very little, especially in an environment where the USGS is fairly incentivized to loudly announce to the world that we've totes got plenty of lithium, my fellow countries, any effort to keep lithium away from us is just a waste of time, look at us just rolling in lithium over here.
Or, maybe they did do the validation, and it's just the reporting that doesn't consider that an important aspect of the story. Somewhere between funding and press release someone's lost the trail but I don't know who exactly.
> The study, which was published in Science Advances, can be found at https://www.science.org/doi/10.1126/sciadv.adp8149 .
This kind of article can perhaps be understood as an attempt to turn a federal organization's sails into the prevailing political winds, so to speak, at a time when funding seems insecure. I say this as someone who strongly supports most of the survey's mission. It would be ideal if national power brokers recognized the value of water science, geology, ecology, etc, on their own terms.