We need more water than rain can provide: refilling rivers with desalination

Why?

We believe that water should be unconditionally abundant. In the face of extended droughts, aspiring for greater usage efficiency is not, by itself, a sufficiently robust solution. The Colorado River, which supports $1.4t/year of US GDP, has seen annual flows steadily decline lower than water extraction rights, with no end in sight. The Colorado River is not an outlier, it is a harbinger. The Mississippi, among dozens of other economically vital rivers worldwide, is also facing record low water. Climate change is changing rainfall patterns and melting glaciers, it is not going away, and it will continue to threaten water security.

Recently announced federal policy, including the Inflation Reduction Act, cuts water allocations to the beneficiaries of the lower Colorado and even offers up to $400/acre-foot for usage reductions, but this will not be enough. $400/acre-foot is not enough money to compensate for lost economic activity and income, and yet $400/acre-foot would be ruinously expensive if enough demand reduction was actually obtained to match the ever-diminishing supply. Our economy rests on an assumption of abundant water and we can’t simply pay people, year after year, to not grow food and to not eat. Water scarcity creates hunger.

Water demand curtailment is only half the story. We must also address supply. For too long, water policy in the American West has been a zero sum game, reallocating intrinsically scarce surface water. This approach has required ever more ambitious dams, canals for an ever-diminishing supply of excess water brought from further and further away, and has long since reached its practical limit. The key is to break the tyranny of meager and inconsistent rainfall by harnessing solar energy to create more water.

We advocate for an aggressive and proactive stance against incipient catastrophic water scarcity. It is not enough to cut usage now and hope for rain. Hope is not a strategy. We have the tools to take control of our destiny and it is long past time that we approach this problem with the seriousness it demands. As our forebears pioneered reservoirs and canals that brought water abundance to the Western US for more than a century, we must safeguard that hard-earned desert fertility for our descendants. 

What?

What if we could solve water scarcity not by choosing lots and killing entire industries, but by simply generating more water? Water is not scarce on Earth, only fresh water upstream of our farms and cities. 

We propose refilling parched rivers by desalinating sea water and piping water upstream. 

We have the technology to convert sea water into fresh, via reverse osmosis (SWRO). Once an extremely expensive, niche technology applied mostly in the Middle East for drinking water alone, SWRO is now relatively mature and ready to be scaled. Much of the cost is energy, and solar electricity is now irresistibly cheap. 

The remaining cost is capital costs associated with building SWRO plants, but like solar this too is falling fast as deployment scales.

To take just one example, Arizona imports 2.8m acre-feet of water from the Colorado annually. SWRO could generate this with just 1 GW of electricity. Even including pumping and 25% solar utilization, just 20,000 acres of solar panels (a square 6 miles on a side) would be adequate to power 100% of Arizona’s imported water. This may sound like a lot but it’s just 2% of the land that Arizona already irrigates.

We could substitute the entire lower Colorado River’s annual flow of 9m acre-feet/year with about 13 GW of solar power, or roughly 3 weeks of global PV manufacturing output in 2021.

It seems absurd that such a trivial investment of land and energy is all that stands between our future and eternal unconditional water abundance.

How?

We advocate for fresh water supply augmentation, to minimize impacts on downstream utilization and infrastructure. Our approach tops up economically important rivers in strategic locations. 

Tech review

The precise technical details of how SWRO works are not critical to this analysis, but a summary is presented here.

Sea water is pre-treated to remove fish, grit, and other macroscopic contaminants. It is then fed into the RO system (1) into a high pressure pump (A). High pressure is necessary to force fresh water through the osmotic membrane (C) which separates fresh water from a remaining  salty brine. Modern systems typically have multiple stages of RO filters. The desalinated fresh water leaves at low pressure (2), while the pass through brine leaves the cell still at substantial pressure (3). It passes through a pressure exchanger (D) where ~96% of its pressure energy is transferred to part of the incoming sea water stream. Low pressure brine leaves the system (5) and can be disposed of either by diluting it in a large stream of ordinary sea water, or by evaporating it in basins potentially to extract dissolved minerals. 

The fresh water, in a drinking water system, is typically subjected to post treatment to stabilize pH, add some salt back in to improve taste, and add chlorine and fluoride.

Modern SWRO systems are very efficient, producing a cubic meter of fresh water for just 2.5 kWh of electricity.

Learning curves

While SWRO is now a relatively mature technology, it has yet to be deployed at anything like the scale needed to refill rivers or irrigate millions of acres of land. This next generation of construction and development will both increase scale and reduce production cost, according to a phenomenological parameterization called the “learning curve”.

The learning curve describes the rate at which a process decosts, or becomes cheaper, per doubling of production rate. For solar power, the learning rate is between 30% and 40%, which is one of the reasons that solar power is now so cheap. For SWRO, the historical learning rate is around 15%.

Large scale present day plants cost around $6 per gallon per day, or $2m/acre-foot/day. This is relatively high compared to the cost of the solar panels, transmission lines, and feeder pipes or canals. 

Current SWRO global water production is 10m acre-feet per year, of which 22% is in Saudi Arabia. 10m acre-feet/year is roughly equivalent to the entire lower Colorado allocation, so we should not expect gratuitous reductions in construction cost from learning rate alone in the near future. That said, total US river flows are about 1.2b acre-feet/year, which represents nearly 7 doublings over current global SWRO water production, and would push SWRO prices down to 32% of their current costs.

In the short term, economies are available through relaxing the purity standard of the output water, most of which will be used for agriculture. 

Model System

To build up our intuition, let’s examine a model system that consumes just 1 GW of solar photovoltaic electricity. 1 GW of solar occupies roughly 20 km^2 of land, or 5000 acres, and costs about $1b to develop. We’ll assume 25% utilization, which is equivalent to 6 hours per day. It turns out that SWRO capital costs are high enough to justify charging and using batteries to keep the plant running overnight, so we’ll assume the plant consumes 250 MW and we have 4.5 GWh of batteries installed locally to operate the plant for the 18 hours of the day that the sun is mostly down. At $300/kWh, the batteries cost $1.35b. 

8760 hours per year at 250 MW is 2.19 GWh, or 876 million cubic meters of water, nearly a cubic km, or 710,000 acre-feet per year. This works out to 1945 acre-feet per day. A SWRO plant this size with present day technology costs $3.9b to build.

Value Proposition

Total capital costs are approximately $1b for the solar array, $1.35bm for the battery, and $3.9b for the SWRO plant, not including canals. 

In total, approximately $6b up front for 710,000 acre-feet/year. At $400/acre-foot, that’s $284m of revenue per year, or 21 years of operation to cover the capital cost. With solar and battery costs declining more than 10%/year, the payback period will continue to diminish.

The Central Arizona Project carries 2.8m acre-feet/year from Lake Havasu to Phoenix and Tucson. Just 4 GW of solar+SWRO are enough to completely replace the river as a source of water, providing a basis for unlimited abundance of water in Arizona, forever.

Logistics

SWRO plants are typically operated close to the ocean. Fresh water must still be delivered to its end users. We have a fairly comprehensive network of existing canals in the American South West, but occasionally new canals will have to be built. Existing dams on rivers also simplify pumping water upstream, within the footprint of the relevant lake. 

Central Arizona Project canal emptied for maintenance. Credit: Tom Tingle/The Republic  

As an alternative, water tunneling technology is a mature, quick, and relatively inexpensive way to build a whole invisible underground parallel river system without disturbing anyone on the surface. Tunnels are less prone to evaporation than canals, but like them may suffer seepage unless the water course is lined with an impermeable material. 

While we can easily cover tunnels, evaporation and additional humidity is less of a concern if we’re generating post-scarcity water. The whole point of transporting SWRO fresh water inland is to increase humidity and extend the growing season. 

Scarcity and production ramp

While we develop the Colorado case study further below, there are dozens of other major rivers worldwide whose flows are threatened by shifting climate and whose inhabitants could apply GW-scale SWRO to ensure adequate agricultural output without further straining parched ecosystems. 

For the following rivers, we will assume a 10% shortfall of flow and compute what scale solar+SWRO plant would be required to make up the shortfall. 

River	Region	Annual flow (acre-feet/year)	Size of plant to replace 10% of flow	2021 deployment cost
Colorado	US south west	17.5m	2.5 GW	$15b
Columbia	US north west	198m	27.9 GW	$179b
Mississippi	US central	556m	78 GW	$468b
Sacramento	US west	22m	3.1 GW	$18.6b
Mekong	South-east Asia	385m	54 GW	$324b
Rhine	Europe	72m	10.1 GW	$60.6b
Danube	Eastern Europe	167m	23.5 GW	$141b
Indus	Pakistan	142m	20 GW	$120b
Teesta	South Asia	203m	28.6 GW	$172b
Ganges+Brahmaputra+Meghna	India	1014m	142.8 GW	$857b
Red	US south	41m	5.8 GW	$34.8b
Murray	Australia	19.6m	2.8 GW	$16.8b
Amu Darya	Central Asia	60m	8.5 GW	$51b
Yellow River	China	45m	6.3 GW	$37.8b

While none of these projects are cheap, it is easy to underestimate the scale they address. Cheap food requires abundant water, and water scarcity begets starvation. Projects of this nature will, under climate change adaptation, stand between a future of abundance and civilizational collapse. To further put costs in perspective, the aggregate cost of every project listed above comes to about $2.3t, a number that will diminish further with improved technology, and which pales in comparison to even US expenditure on COVID-related stimulus. In other words, we have entered the age of SWRO-enabled agricultural production.

Case Study – The Colorado River

Average flows on the Colorado have declined by more than 20% since 1900, a trend consistent with increased air temperatures in the Upper Colorado basin, which in turn increases evaporation and reduces precipitation. This trend is extremely unlikely to go away, leaving the US southwest with an insatiable thirst which cannot be quenched, except by the generation of additional fresh water.

Chart credit.

Chart credit. A graphical summary of water diversions from the Colorado River c. 2009, aptly illustrating just how little margin there is for supply reductions. Almost half the water is diverted through a small handful of canals into California, and more than half of the remainder into Arizona. 

Chart credit. Diagram showing the geographical location of the major Colorado-fed water diversions. 

Stage 1 – All American Canal and Colorado River aqueduct

The first step to relieving pressure on the Colorado River is to augment the supply of water diverted into the All American Canal, which totals nearly 3m acre-feet per year, almost a third of the total lower Colorado allocation. Additionally, the Coachella canal could in principle connect to the Colorado River Aqueduct, augmenting up to 1m additional acre-feet per year otherwise drawn from Lake Havasu to water Los Angeles. 

4m acre-feet need 5.6 GW ($33.4b capital cost) of solar powered sea water reverse osmosis, most likely located in the southern Imperial Valley east of the agricultural area. A connection to the gulf of California is required, making this an international project. As Mexico also relies on water from the Colorado, a deal to relieve pressure on natural flows would enjoy support on both sides of the border. Mexico’s current allocation of 1.5 m acre-feet per year could be freely increased to any desired level, ideally restoring the health of the Colorado delta and Mexico’s agricultural activity there.

Also in the Imperial Valley is the Salton Sea, a festering brackish lake in dire need of large-scale intervention. Local desal capability would extract economically valuable mineral salts from the lake, and thus regulate its level and salinity at a point more conducive to life and health for surrounding inhabitants.

Stage 2 – Central Arizona Project 

The Imperial Valley desalination system may augment the Colorado’s flow enough to solve the water scarcity problem in the US south west forever, but it is also possible that natural flows will continue to diminish as increased consumption of water upstream of the Imperial Dam once again encounters scarcity. In this case, the next largest consumer of water is Arizona, with a historical allocation of 2.8m acre-feet per year feeding both Phoenix and Tucson via the Central Arizona Project (CAP), completed in 1993. 

There are two potential ways to augment this flow. The first way is to expand the Imperial Valley SWRO project and pipe water north through the Palo Verde valley into Lake Havasu, where it can feed the CAP. The second is to build additional desal further south, in the Rio Sonoyta catchment, pump it north over the divide, and then reverse the flow direction of the CAP (this time in gravity’s favor!) to feed Tucson, Phoenix and the Salt River agricultural area, and perhaps even Lake Havasu on the Colorado itself. 

Stage 3 – Nevada and Lake Meade

As nearly the entirety of the lower Colorado is dammed, pumping water upstream from Lake Havasu to Lake Powell via Lake Mohave is relatively straightforward. A vertical lift at each dam using existing hydroelectric infrastructure, with relatively modest pipes or canals where the dams do not quite meet. Via this expedient the entire flow of the Lower Colorado could be assured even if the Grand Canyon ran dry indefinitely.

Stage 4 – Lake Powell

While it seems unlikely that, in the event of total augmentation of the lower Colorado, Lake Powell would continue to suffer water scarcity, it is still physically possible to pump water further up the river, feeding Lake Powell from Lake Meade.

The geographic obstacle to this is the Grand Canyon itself, with terrain that is decidedly unfriendly to canals! Several options present themselves. 

Conventional canals could be built to the north of the canyon, bypassing the Virgin River Gorge, St George, Hurricane, then traversing east parallel to the Utah/Arizona border.

Water tunnels can be built faster and cheaper than transport tunnels, and the Grand Canyon stratigraphy presents numerous layers of competent, horizontally bedded impermeable rock through which conventional boring techniques could drive a large tunnel in as little as a year, depending on haste and technology used.

Stage 5 – Restoration of natural rivers

Artificial lakes serve several purposes, including hydroelectric generation, flood control, and water storage. They also have benefits including climate moderation, transport, and recreation. On the other hand, they seriously disrupt ecosystems and flood enormous areas, so much so that dam removal is now in vogue.

What does the Colorado look like with solar desal supported flow and fewer (or no) dams? First, a somewhat more extensive water distribution system is needed to transport coastal desalinated fresh water inland and to plumb it into existing canals and/or tributaries. Second, the ability to rapidly control extreme flows regardless of precipitation and thereby to avoid catastrophic floods. 

In the limit, western rivers could return to their wild, economically unexploited former state while human uses of water could be fed directly via a parallel river system fed by SWRO. 

Scalability

How much solar is needed to do this? Is this an extensible solution to an eternal problem?

Generation of fresh water from salt, using electricity, is relatively inexpensive. By contrast, direct electrical synthesis of gasoline uses 15,000 times as much energy per unit volume. While a 1 MW solar plant may be able to generate the equivalent of one barrel of oil, in hydrocarbons, that same solar plant running SWRO can generate 2385 cubic meters of water, which is nearly two acre-feet. Fortunately, per capita consumption of oil is about 300x less than water!

If all 8 billion people on Earth consumed as much fresh water as the US average (~1.1 acre-feet/cap/year), and 100% of that use was created with SWRO, we would need about 12.4 TW of solar power. That’s just 248,000 sqkm, or just 31 m^2 of solar array per person. This is an extreme scenario but aptly illustrates that, in 2022, water scarcity or abundance is no longer restricted by the vagaries of weather and climate; it is a choice. We should choose abundance.

Impact

Building anything carries economic and environmental impacts, but it is important to center the fact that choosing not to build is a choice that also carries a cost. In the case of water provision to the thirsty American West, decades of underinvestment have led to the collapse of natural river flows, the depletion of fossil water aquifers, and an otherwise undesirable brake on economic growth. 

Yes, gigascale SWRO has an environmental impact, but what is that impact compared to the default option of “business as usual”? Using a small portion of our practically infinite arid land to generate solar power seems a small price to pay for future-proof flows of enough fresh water to return much of the West to its Pleistocene climate of lakes, raging rivers, lush meadows and endless forests, should we so choose to return this land to its Eden-like state during the ice age just 10,000 years ago.

In the context of the proposed projects in California, Nevada, and Arizona, there is the potential for negative environmental impact, stemming from three major sources: solar power, brine disposal, and landscape modification through canal construction.

The stage 1 Colorado project described above, intended to substitute 4 million acre-feet per year, which is essentially all westward-bound extractions from the Colorado River, requires 5.6 GW of solar energy. For context, this is roughly 10 days of solar panel production in 2021, though the industry continues to grow explosively. 5.6 GW consumes roughly 112 square km of land, and if it was a single farm it would be the largest yet built, exceeding the Indian Bhadla Solar Park by a factor of two. This seems like a lot but it could easily be located east of Brawley far enough from any major road that most people would remain unaware of its existence.

This screenshot shows how 5.6 GW of solar would fit neatly to the east of the Imperial Valley agricultural areas, close enough to the All American and Coachella canals to feed them, while requiring new water transport only to access the ocean.

At greater remove, the city of Los Angeles is visible. The solar array is smaller than Palm Springs, smaller than Pasadena, smaller than the Salton sea, which it would help to rehabilitate.

Finally, at regional scale we can see the area of the land in California, Arizona, and Nevada that could finally be watered with a relatively modest development of existing technology for significantly less capital than Elon Musk used to buy Twitter.

Solar arrays have both positive and negative impacts on land. They increase shading and moisture retention, and involve substantially less modification to the surface than any other kind of human development, least of all agriculture. 

Brine disposal is a substantial concern with large scale desalination. Depleted brine returned from SWRO is roughly twice as salty and denser than sea water. Dumped into the ocean, it sinks to the ocean floor, and there can kill sea life rather than rapidly mixing with the rest of the ocean. 

SWRO brine is enriched in salt, and also in all the other minerals dissolved in the ocean, many of which are the lighter alkali metals, which are increasingly critical to the energy transition. In other words, we may see lower impact lithium, magnesium, and calcium extraction tacked onto SWRO, transforming brine disposal from a cost center to a profit center. Even boring old sodium chloride, or table salt, can be used as an additive in cement in enormous volumes. 

The key to reducing the biotic impact of SWRO brine is to ensure it is thoroughly mixed with an adequate supply of sea water that its salinity is not raised to an intolerable level. Different oceans have differing baseline salinity, and the quantities of fresh water extracted by SWRO are so vanishingly small compared to the ocean that the net impact on ocean salinity is the same as evaporation and natural rainfall, which is to say, zero. 

In the Colorado Stage 1 project, the SWRO plant can be sited anywhere in the Colorado delta, on either side of the US-Mexico border, since it is close enough to sea level to make little difference. The key is that the intakes flush a relatively large volume of sea water past the filtration system to prevent ingestion of sea life, while also creating a large enough current flow to dilute the returning brine stream. 

Canals also have impacts on the landscape, inhibiting the movement of animals if sufficient bridges are not built. Not every canal is open, however. The Colorado River Aqueduct is almost entirely underground, either tunneled or covered over to prevent both evaporation and sabotage. The landscape impact of canals is comparable to a similar sized road and is significantly less damaging than, say, urban construction. 

There are also positive impacts well worth enumerating, such as alleviation of salinity, improved agricultural productivity in both the US and Mexico, and the loosening of the yoke we’ve placed on the Colorado and other rivers. 

View from the future

With solar powered SWRO we can break the default balance between natural flows of fresh water and ever-present scarcity, permanently in favor of abundance. We can choose to argue and fight over relatively insignificant side issues for another decade or three, we can choose to kick the can down the road and leave yet another increasingly hopeless mess to our children’s generation. Or, we can choose to recognize that the times have found us here, and now, where our technology has broken down previously unbreachable barriers, and all we have to do is build our future as though we are eager to see it in our own lifetimes.