Some land sinks and never comes back up

The water came back and the ground kept falling

Near Cantua Creek, on the west side of California's San Joaquin Valley, an observation well tracked water in a deep aquifer, the underground layers of sand, gravel and clay that hold the water wells pump up. Between 1968 and 1974, as canal water replaced pumping, the water level in that well rose more than 200 feet. Over the same six years the land around it sank another 2 feet. [S02]

That was the tail of decades of heavy pumping for farms. By 1970 more than 5,200 square miles of irrigable land, about half the valley, had sunk more than a foot, and the ground near Mendota had dropped more than 28 feet. [S02] A well-known USGS photograph shows the hydrologist Joseph Poland beside a pole near Mendota, with signs marking where the land surface stood in 1925, 1955 and 1977. [S02]

The cause is clay. In the valley, beds of silt and clay lie between the water-bearing layers of sand and gravel. Pumping lowers the water pressure in the sand, the clay beds slowly drain into it, and the weight of the ground above presses the clay grains closer together. Each clay bed has a heaviest load it has ever carried. Push past it and the grains rearrange for good, and the clay becomes 10 to 1,000 times easier to squeeze. [S01, S07] Water returning to the sand later cannot push those grains apart again. [S01]

Clay drains slowly, so it keeps compacting for years after the pumps ease. That lag is why the well near Cantua Creek and the ground beneath it moved in opposite directions. [S02]

Over the same six years the land around it sank another 2 feet.

The layers written off as dry held most of the storage

Hydrologists long treated clay beds as barriers between the water-bearing sands. USGS now describes these layers, in many such systems, as "once dismissed as nonwater yielding, comprise the bulk of the ground-water storage capacity" of the aquifer. Clay is more porous and more compressible than sand, and there is often more of it. [S01]

The first time a basin is pumped hard, that clay gives up water as it compacts. USGS calls it the "water of compaction" and says it "cannot be reinstated by allowing water levels to recover to their predevelopment status." [S01] Where a basin is pumped far faster than it refills, this water typically makes up 10 to 30 percent of everything pumped, and the volume of water released roughly equals the volume the land sinks. [S01]

The San Joaquin Valley showed what that spending looks like later. In the droughts of 1976–77 and 1987–91, imported water deliveries to the west side were cut and farmers turned their pumps back on. Water levels fell fast after a relatively small amount of pumping, which USGS attributed to "reduced ground-water storage capacity" from the compacted clay. [S02]

once dismissed as nonwater yielding, comprise the bulk of the ground-water storage capacity

The loss estimates differ for traceable reasons

Studies of the 2007–10 drought give numbers that look like they disagree. A valley-wide satellite study by Ojha and colleagues put the permanent loss at about 0.16 to 1.3 percent of the Central Valley's aquifer storage. [S05] Smith and colleagues, working in the southern San Joaquin Valley, put it at roughly 5 to 9 percent of the groundwater used. [S04] USGS's general figure for basins pumped far faster than they refill is 10 to 30 percent of the water pumped. [S01]

Most of the spread comes from what each number is divided by. Ojha's team compared the lost space with the volume of the whole aquifer system. Smith's team and USGS compared it with the water pumped out. One drought draws down only part of what an aquifer holds, so the same lost volume is a small share of the first and a larger share of the second. [S04, S05]

The rest comes from quantities nobody can measure directly across a valley. Ojha's team did not know how thick the compacting clay was, so it ran its numbers for 50, 100, 200 and 400 meters. Those assumptions span a factor of eight, and so does its estimate. [S05] Smith's team could not pin down each well's old lowest level, because the records are patchy, so it built a method that works around it. That method found at least 54 percent of the measured sinking was permanent, and 98 percent by the team's preferred estimate. Those two results are the source of the 5 and 9 percent of pumped water. [S04] The authors add that the 9 percent is a regional average, and in the clay-heavy southwest the share of pumped water that came from compaction was likely much higher. [S04] Estimates from the 1970s ran from 20 to 60 percent in one west-side area, and Smith's team notes that those did not separate permanent compaction from the springy kind. [S04]

One discrepancy sits inside a single paper: Ojha's abstract says "~2%" while its results give 0.16 to 1.3 percent, and we use the results. [S05] For the 2012–15 drought, the same group put the permanent loss in the San Joaquin Valley at 0.4 to 3.25 percent of storage capacity, and added the two droughts together for 0.6 to 5.25 percent. That sum takes the first drought as 0.5 to 2 percent, a different range from the one in the 2018 paper's results. [S06]

Every one of these studies agrees on scale. Smith's team calls the loss "just a small fraction of the total water stored underground," and USGS calls it "a small permanent reduction in the storage capacity of the aquifer system." [S01, S04] The sand and gravel layers still store water and still refill.

Most of the spread comes from what each number is divided by.

Texas stopped the sinking by changing its water

Around Houston, pumping for industry and cities along the Ship Channel drew down the coastal aquifers through the mid-1900s. By 1979 up to 10 feet of subsidence had occurred, and almost 3,200 square miles had sunk more than a foot. [S03] Some of the earliest sinking had a different cause, the oil, gas and brine pulled from fields such as Goose Creek. Later sinking followed groundwater pumping as it spread inland. [S03]

On a low coast, that much sinking moved the shoreline. The San Jacinto Battleground, where Texans led by Sam Houston defeated Mexican forces under Santa Anna on April 21, 1836, became partly submerged. USGS reported in 1999: "About 100 acres of the park are now under water due to subsidence," and a road to the San Jacinto Monument had been closed by flooding. [S03] In 2016 the state pumped sediment dredged for a nearby industrial project onto part of the battleground marsh, raising its bottom about 10 to 20 inches so marsh grass could grow back near Boggy Bayou. [S10] That ground went back up because material was added on top.

In 1975 the Texas legislature created the Harris-Galveston Coastal Subsidence District "for the purpose of ending subsidence," and the coastal area switched much of its supply to surface water from Lake Houston and Lake Livingston. [S03] By 2025, water levels in the converted coastal and central areas had risen more than 200 feet since 1977, and the coastal area showed "minimal subsidence to uplift." [S09] The ground stopped falling. We see nothing in how clay compacts that would bring the 10 feet back, because most of that compaction happened in clay.

North and west of Houston, where growing suburbs still rely mostly on groundwater, water levels have declined significantly, and one GPS station near Katy and Fulshear recorded more than 3 centimeters of sinking a year. Conversion to other water sources there is not due to finish until 2035. [S09]

About 100 acres of the park are now under water due to subsidence

A city on a lake bed shows where the process ends

Much of Mexico City stands on the clay of the former Lake Texcoco. Chaussard and colleagues combined 115 years of levelling surveys, 24 years of satellite radar and 14 years of GPS. They found the city sinking at rates mostly constant since at least 1950, reaching 50 centimeters a year. [S07]

They found no significant springy, seasonal movement, so the sinking is almost fully permanent. Its rate tracks the thickness of the upper clay layer, about 3 millimeters a year faster for every extra meter of clay. [S07] On average that clay is about 17 percent compacted. The team estimates about 150 more years until it is fully compacted, with up to 30 meters more sinking in places. Even if water levels were raised, they write, "there is no hope for recovering the majority of the already-lost elevation and the lost aquitard [clay layer] storage capacity." [S07]

The squeezed clay causes a second problem. With water levels now below most of it, the clay's salty, polluted pore water drains down into the aquifer the city pumps from, and shrinking clay cracks, opening paths for sewage. [S07]

there is no hope for recovering the majority of the already-lost elevation and the lost aquitard [clay layer] storage capacity

The old low sets the limit

Across these places, one number decides whether the damage is permanent: the lowest water level a basin has already reached. Pumping that stays above it squeezes the clay a little and lets it spring back, which shows up as seasonal rises and falls of the land of up to an inch or more. [S01] Pumping below it spends storage that will not return. That is our reading of the process USGS and the satellite studies describe, and Houston's coast shows the stopping half of it. [S03, S09]

The bill for crossing that line comes late. After the 2012–15 drought, Ojha's group found the clay still compacting for 0.5 to 1.5 years. [S06] A model of 65 years in one part of the San Joaquin Valley, by Lees and colleagues, suggests residual compaction there runs for decades to centuries, so slowing subsidence needs some recovery of water levels, and holding them steady is not enough. [S08]

A well's lowest reading usually gets filed as a bad year. In clay country it also marks how far the next drought can go before it spends storage for good. Smith's team expects the clay-rich subsidence bowl where the Tulare Lake, Kaweah and Tule subbasins meet to compact permanently again whenever heavy pumping and little recharge coincide. [S04]

In clay country it also marks how far the next drought can go before it spends storage for good.