A 15-member research team led by Shijie Wang, of Yunnan University in China and Johannes Gutenberg University Mainz in Germany, among other institutions, has used tree rings to reconstruct 343 years of flow for the Selenga River, which flows into Lake Baikal, and the Angara River, which flows out of it. According to a paper published on November 3 in the journal Geophysical Research Letters (GRL), a combined index of the two rivers' flow fell by about 90% from a high level during a period in the 20th century, then turned to an increase of about 80% over the past decade. The team assesses the recovery since 2017 as the most pronounced in the 343-year record.

However, Lake Baikal's water level has been affected by dam regulation since 1959. In addition, the flow estimates for 2015–2024 fall outside the period calibrated against river observation data. The team explains that a local slowdown of warming in the Arctic altered atmospheric circulation and increased the inflow of water vapor, but other research suggests that internal climate variability is involved in the stall in Arctic sea ice loss.

AD

What was reconstructed: flow of the two rivers entering and leaving Lake Baikal

According to UNESCO's World Heritage Centre, Lake Baikal is the world's oldest lake, at about 25 million years old, and also the world's deepest, at about 1,700 m. It covers 3.15 million hectares (about 31,500 km²) and holds roughly 20% of the planet's unfrozen fresh water. Wang and colleagues' paper identifies the Selenga River, flowing into the lake, and the Angara River, flowing out of it, as the main rivers governing the water balance of this vast freshwater lake.

An environmental report on the Lake Baikal basin says that surface water, including rivers, accounts for 57.77 km³ per year, or 82.4% of the long-term average inflow to the lake. Meanwhile, 60.89 km³ per year, equivalent to 86.8% of outflow from the lake, is discharged through the Angara River. In other words, most of the water entering the lake is supplied by rivers, and most of the water leaving it is concentrated in the Angara.

What the study reconstructed is a combined index averaging the observed flows of the Selenga and Angara rivers. The team used data from the Mostovoy station on the Selenga and the Irkutskaya GES station on the Angara, and tried to capture the water flow of the lake as a whole from the two rivers' average flow. The "roughly 80% increase" reported in the paper therefore does not mean that total inflow to the lake or its stored water volume grew by 80%. The result is that an index combining the flow of the river entering the lake and the river leaving it rose substantially over the past decade.

The study used 52 tree-ring cores newly collected in autumn 2024 at two sites around the lake, along with existing records from 12 sites registered in the International Tree-Ring Data Bank. The paper reports that the recent flow recovery is consistent with greening of vegetation in the basin and rising lake levels. Not only the rivers but also plant growth and the state of the lake surface show trends indicating wetter conditions.

How 343 years of flow are estimated from tree rings, and how this differs from a 2006 study

There is precedent for reconstructing past river flow from tree rings. In a study published in the Journal of Climate in 2006, Nicole Davi and colleagues used ring-width records from five sites in west-central Mongolia to reconstruct Selenga River flow for 1637–1997. The Selenga is known in Mongolia as the Selenge River and flows into Lake Baikal as the country's largest river.

According to a dataset published by the US National Oceanic and Atmospheric Administration (NOAA), Davi and colleagues estimated mean April–October flow and explained 49% of the observed flow variability.

Reconstructing flow from tree rings is done mainly in two stages: calibration and estimation. Trees in arid regions tend to grow more, producing wider rings, in years when more water is available. Researchers first fit a statistical model relating ring width to flow, using the period in which river-flow observations and ring data overlap. This period is called the calibration period.

The resulting statistical model is then applied to ring data from eras with no observational records to estimate past flow. The 49% figure in Davi and colleagues' study represents the share of observed flow variability that the ring-based model could explain. The remaining roughly half of the variability is not explained by the model.

Wang and colleagues' study used a similar method. The reconstruction covers 1682–2024, and a regression model based on 1959–2014 observation data explained 47.0% of flow variability. After verifying the relationship between the ring records and observed flow, the team estimated past flow.

However, because river observation data were available only through 2014, the estimates for 2015–2024 extend beyond the calibration period. To check the reliability of this period, the team also examined relationships with lake level data for 2008–2023 and gridded runoff data for 1902–2019. Even so, the paper says the 2015–2024 estimates carry uncertainty and need to be verified against longer observation records.

The 2006 study and the present one differ in target, samples, and period. Davi and colleagues used tree rings from west-central Mongolia to reconstruct April–October mean flow for the Selenga alone. Wang and colleagues, by contrast, combined ring records from around Lake Baikal and targeted the combined flow of the Selenga and Angara. Whereas the earlier reconstruction ended in 1997, this one is extended to 2024, making it possible to assess the rapid changes since the 2010s.

AD

Dam-regulated water levels since 1959, and past low-water periods

To understand the "rising lake levels" the paper cites in support of the recent recovery, the influence of dam regulation of water levels has to be taken into account. Since 1959, Lake Baikal has been affected by the Irkutsk hydropower plant built on the Angara River, so its present water level is not determined by natural conditions alone.

According to the basin environmental report, the Irkutsk hydropower plant's dam was completed in 1956–1958 and began impounding water. The backwater, the water held up on the upstream side of the dam, reached Lake Baikal in 1959 and raised the lake's long-term level by 1.3 m by 1964. The lake has since gone through multiple low-water and high-water periods, and plant operations and government regulation influence its level.

Year / period Main event
1956–1958 Irkutsk hydropower plant dam completed; impoundment begins
1959 Dam backwater reaches Lake Baikal
1964 Long-term lake level raised by 1.3 m by this year
1976–1981 First major low-water period after the plant began operating
1996–2017 Second prolonged low-water period. Comparing 1996–2020 with 1971–1995, the Selenga's share of inflow to the lake is said to have fallen by about 20%
2001 Russian government sets the lower limit of the lake level at 456 m (Pacific datum)
February 2015 Special decree issued permitting water use below 456 m only during autumn and winter 2014–2015
2020–2021 High-water period, prompting a review of lake level regulation
November 3, 2026 Wang and colleagues publish the 343-year flow reconstruction in GRL

The dam history draws on the basin environmental report, the division into low-water periods on a 2022 study in Geography and Natural Resources, and the regulatory history on Russian government documents and news reports.

During the 2015 low-water period in particular, the government faced difficult decisions over the lower limit it had set for lake levels. Regulations from 2001 set the minimum lake level at 456 m, but in February 2015 water use below that level was permitted as a special measure. The Russian government decree allowed use below the lower limit only during autumn and winter 2014–2015, in order to secure water needed for residents and economic activity downstream of the Irkutsk hydropower plant, and attached a condition that the level be restored during the 2015 flood season. According to an Interfax report, the lake level fell to 455.99 m on February 25, 2015.

A study published in Water in 2024 reconstructed 123 years of a "conditional natural level", the level that would have prevailed had the dam not been built. It found that natural lake levels showed a rising trend of about 15 cm per century. At the same time, it concluded that operations limiting discharge from the plant, together with legal level regulations, have greatly constrained the range of lake level fluctuations.

This study shows that dam release operations and level regulations affect the state of the lake. The combined flow reconstructed by Wang and colleagues also includes the Angara River, which is affected by operations at the Irkutsk hydropower plant. However, based on comparisons with upstream observation records, Wang and colleagues explain that the dam's influence on the natural flow variability signal used in the reconstruction is small.

Slowing Arctic warming as a possible driver of the recovery, with natural variability also affecting sea ice loss

Wang and colleagues focused on Arctic climate change as a factor behind the recent rapid flow recovery. The Arctic has been warming faster than the global average, but according to the paper, warming slowed in some Arctic Ocean regions, and atmospheric circulation changed accordingly. The team explains that this change increased water vapor transport to the Lake Baikal basin and led to the recovery in river flow.

On the other hand, some research holds that internal climate variability has been heavily involved in recently observed changes in the Arctic Ocean. A study published in GRL in 2025 by Michael England and colleagues reported that no statistically significant decline was seen in September Arctic sea ice extent since 2005.

That study analyzed large-scale simulations from the CMIP5 and CMIP6 climate models and concluded that it is not unusual for the pace of sea ice loss to slow for several decades even as greenhouse gas emissions rise. The team believes that natural variability arising within the climate system has substantially offset human-caused sea ice loss in recent years.

However, England and colleagues examined September sea ice extent in the Arctic Ocean, and their work did not directly test the local slowdown of Arctic warming that Wang and colleagues pointed to. Even so, natural variability on the Arctic side is an important element in considering the explanation that moisture supply to the Lake Baikal basin is governed by changes in Arctic atmospheric circulation.

Also, a temporary slowdown in Arctic Ocean sea ice loss does not mean sea ice is not declining in all surrounding seas. According to a diagnosis published by the Japan Meteorological Agency on June 30, 2026, the annual maximum sea ice extent in the Sea of Okhotsk is declining at a rate of 59,000 km² per decade. The 95% confidence interval is 34,000 to 84,000 km², and the decline is equivalent to 3.7% of the Sea of Okhotsk's total area. Because the sea area and the metric differ from England and colleagues' study, the figures cannot be compared directly, but they show that climate change around the Arctic varies by region.

AD

Will the roughly 80% flow recovery continue? A new perspective for water management

What Wang and colleagues' study shows is that Lake Baikal's flow has fluctuated greatly over several centuries. A long-term drying trend progressed in the 20th century, while a particularly pronounced recovery appeared after 2017 even within the 343-year record. In light of this change, the team points out that water resource management based only on conventional observation records may not adequately capture rapid changes in the future.

The paper stresses the need for adaptive water management that takes into account "non-stationarity", that is, it does not assume that past flow and climate conditions will retain the same statistical properties in the future. Non-stationarity refers to the property that long-term means and the range of variability change over time. In the case of Lake Baikal, dam operations and legal level regulations, as well as climate change, shape the conditions for water use.

Indeed, the 2024 study in Water showed that level fluctuations could lead to regulatory violations or unfavorable water-use conditions in both low-water and high-water periods. In 2015, use below the lower level limit was permitted as an exception, and the subsequent 2014–2017 low-water period and 2020–2021 high-water period led to reviews of the regulatory system. Management that can respond not only to water shortages but also to excess water is needed.

On changes in the Arctic, the study by England and colleagues offers an outlook. Climate model analysis indicates that the stall in Arctic sea ice loss could continue for another 5 to 10 years, while the probability that sea ice declines faster than average in the near future is also rising. However, this is a projection for Arctic sea ice and does not directly show how long the wet period in the Lake Baikal basin will last.

In addition to the fact that the 2015–2024 flows in this reconstruction are estimates outside the calibration period based on observation data, the paper's abstract gives the roughly 90% decrease and roughly 80% increase without specifying the baseline years or periods for the comparison. The team says that updated observation data and reconstructions combining multiple independent records will be needed to assess how long the current flow recovery will continue.