2026-01-29_water-resources-podcast_dilong
Bridget Scanlon: [00:00:19] I am pleased to, to welcome Di Long to the podcast. Di is a professor in hydrology and remote sensing at the Department of Hydraulic Engineering at Tsinghua University in Beijing. And prior to this position, Di was a postdoc in our group at the University of Texas from 2011 to 2014, and he did his PhD at Texas A&M prior to that.
So thank you so much Di for joining me today.
Di Long: It's my great pleasure, Bridget.
Bridget Scanlon: Today I think we're going to talk about Di and his team have worked on many different research topics and today I think we're going to talk a little bit about the drought work he did at the University of Texas because he was here during the 2011 drought. Very extreme drought in Texas.
And then after that we would like to talk about the, his work on the Tibetan Plateau, Asian Water Tower, and then his research on changes in water storage in the North China Plain as a result of the South to North water diversions. And then Di has used a variety of different tools and artificial intelligence and machine learning then to conduct the research and talk a little bit about that also.
So let's first start about your work when you were here at the Bureau of Economic Geology at the University of Texas, Di your postdoc year after completing PhD at A&M. And, Texas was experiencing one of the worst one-year droughts on record in 2011. And you applied the GRACE satellite data to estimated water storage depletion as a result of that drought.
And the GRACE satellites are great because they provide total water storage. But one of the aspects also that you did was try to figure out what was contributing to the storage losses, whether it was reservoir storage, soil moisture, or groundwater. Maybe you can describe that work a little bit.
Di Long: Okay. That period really shaped how I think about the droughts and the water systems. When I was a PhD graduate at Texas A&M and a postdoc at UT Austin, Texas experienced the mostly extreme one-year droughts on record in 2011. Precipitation dropped to about 40% of the long-term mean, and the impacts were enormous.
Agriculture long suffered around. 7.5 billion US dollars in losses affecting both crop and livestock across the state. And what made this job particularly interesting scientifically, was that it wasn't just a surface phenomenon. This was one of the first events where we could really use GRACE satellites to look at total water storage with the TWS, meaning surface water, soil moisture, and ground water together.
Rather than trying to infer impact component by component using GRACE data from CSR and GRGS centers, we found that during the peak of the drought, Texas lost about 62 cubic kilometers of total water storage with uncertainty of roughly about 18 cubic kilometers to put that in perspective, that's about 50 million acre feet or nearly three times a Texas current annual water use, which is on order of 15 million acre feet.
From the meteorological perspective, the drought was very clear as well. The Palmer Drought severity index dropped below minus two in February, 2011 and stayed there until September, 2012 with extreme minimum close to minus eight in late 2011. And that help us define the core periods and connect atmospheric force into hydrological response.
And a key question, we are trying to address: where did all that water loss come from? Surface water did decline, but surprisingly, they explained owning a small fraction of the TWS signal using in site reservoir storage data we estimated about 7.6 cubic kilometers of reservoir storage loss, which is only about 12% of the total GRACE observed depletion.
So then we look at soil moisture. We use multiple land surface models from GLDAS and also individual models varied quite a bit. They all showed the strong depletion. In fact, GRACE TWS anomalies were very highly correlated with modeled soil moisture anomalies with correlations between about 0.86 and 0.95, suggesting soil moisture plays a dominant role. And in terms of ground water, it is another important piece of information based on ground water pumping statistics from the Texas Water Development Board. Groundwater used in 2011 was around 12 to 13 cubic kilometers or roughly 10 million acre feet.
And in regions like the High Plains aquifer, we estimated groundwater storage losses of around five cubic kilometers. So overall ground water likely accounted for about 8 to 16% of the TWS decline during the drought. So putting all of this together, our syntheses suggested that soil moisture depletion accounted for roughly 70 to 80% of the TWS loss with ground water and surface reservoirs contributing smaller, but still important shares.
Bridget Scanlon: Yeah. So that was a really nice synthesis and when we experienced droughts, we're always wondering what the impacts of the droughts are. And you mentioned, the huge economic losses during that drought, and I, I don't know if you remember Di, but when you went out that summer, it was like going into a convection oven.
It was so hot and windy and,
Di Long: Yeah, still vivid, yeah.
Bridget Scanlon: Yeah, it's difficult to disaggregate the GRACE satellite data into the components. And we are fortunate at the University of Texas to have the Center for Space Research, CSR who were one of the main groups that provide the GRACE satellite data globally.
And then, the reservoir storage the Texas Water Development Board has a pretty good handle on that, and they monitor the water storage in the reservoirs. So that's a pretty good number. And then also they estimate how much groundwater water is used in the state. And so you can estimate, even if you assume that there was no recharge and that all that groundwater was a net loss in storage is still a small fraction of the total water storage depletion.
So that was about 50 million acre feet and Texas uses about 13 to 15 million acre feet a year. So even if you assume there was no recharge that still doesn't account for all the total water storage lost. So a lot of it has to be in soil moisture, or it could be some shallow groundwater also.
Di Long: Yeah. yeah right, yeah
Bridget Scanlon: So with that big depletion in soil moisture, it explains why it takes so long then to recover from a drought. So 2011 was the most extreme part of the drought, but rainfall after that maybe was close to normal conditions. But you have that huge soil moisture deficit that you have built up that you have to overcome before you can get run off into the reservoirs.
So it wasn't until 2015 when we had the highest rainfall on record in the state that we really ended that drought. So the drought, there was some recovery in 2012 and Nelun Fernando at the Texas Water Development Board and Rong Fu looked at the impact of North Atlantic oscillation on some recovery at that time.
But then really the drought persisted until 2015 when we had a lot of flooding Wimberley floods in May and other floods. 2015 was a very wet year, and you almost need a flood to end the drought, particularly one that has built up all that depletion.
Di Long: Yes.
Bridget Scanlon: And so you've been looking at the US Drought Monitor provides a great record of droughts and we've been seeing more droughts in Texas since then I guess you have worked on droughts a lot and your work at, during your PhD was looking at evapotranspiration and looking at all of these different products.
What do you think of the more recent droughts in, I guess we can look at those now with GRACE satellite data and we are developing a dashboard then to provide the GRACE data. And I guess one of the advantages of GRACE, again, give us a statewide picture of water storage because it's a coarse resolution.
And so you can look at the large scale.
Di Long: Yeah. I think GRACE fundamentally changed how we monitor drought. Instead of agreeing about uncertainties in soil moisture models or ground water observations, GRACE gave us a direct integrated estimate of how much water the entire state was losing. That's incredibly powerful for drought assessment and water resources management, especially in a place like Texas where water is managed across multiple sectors and jurisdictions.
So this work feels even more relevant today, I think. Texas has since experienced the 2021 winter freeze followed by severe droughts in 2022 and 2023 with cost impacts on agricultural ecosystems. And GRACE and now GRACE Follow On gives us a way to consistently track these compound extremes and understand how the system responds over time.
Yeah.
Bridget Scanlon: And I guess also, Di, the important aspect with that water storage is how much water you need to recover from these long-term droughts. And so you get an idea then from that. And there is in your paper on the 2011 drought, you did show that soil moisture storage there were huge variations in the soil moisture storage from the models.
But we look at a lot of different components and try to estimate the uncertainty. But it showed that we needed a lot of water to overcome that drought. And another colleague, Mike Inger wrote a paper called something like Drought Busting Floods. And it was talking about the atmospheric rivers in the Western US and how they oftentimes bust long-term droughts in California and other places. Di, you went back to China in 2014 and you're at university and I was listening to a book recently by Dan Wong called Breakneck and comparing China and the US and indicating that many of the politicians in China have an engineering background, and contrasted to the US where most of politicians have a legal background or went to law school.
And I remember when I visited your university there were photos of past presidents that went through your department and stuff. So maybe you can describe that a little bit.
Di Long: Okay I'm really grateful for your support over the years, Bridget, first, so including during my time at UT Austin, returning to China was a big decision for me, but in many ways it felt like a natural step. What you mentioned from Breakneck really resonates with my experience. Engineering and science have a very special status in China, historically and culturally. Many national leaders have engineering backgrounds at Tsinghua University in particular has long been associated with the training engineers who later take on leadership roles. And President Xi is also a graduate. And if you walk through our department buildings you will still see photos of the former president and leaders in various sectors who studied or taught here.
So at a more practical level, this emphasis translates into strong and sustained investment in education and research. Most graduate students in China have their tuition largely covered by the government, which removes the financial burden and allows students to focus on learning and research.
This is quite similar to the US I think in my case. Much of my student funding comes from competitive national grants, especially from the National Natural Science Foundation of China, as well as programs supported by the Ministry of Science and Technology. These funds allow us to support PhD students, postdocs, and long-term research that might take many way years to mature. And also beyond the national funding, there are also many additional sources. Some of my colleagues collaborate closely with industry, particularly in areas like infrastructure and AI. In addition, local and the municipal governments invest heavily in research, especially in innovation. Folks, the cities like Shenzhen, where research finding is often closely linked to real world applications and rapid technology transfer.
What I find especially exciting is that this engineering driven cultural encourages problem-oriented science. There's a strong expectation that research should eventually inform decision making, whether it's water security, climate adaption, or infrastructure safety. So that mindset has strongly shaped how I approach my work since returning to China. So in many ways, my past reflects a broader system, strong training in China and abroad, combined with domestic environment that values engineering, support students and invests heavily in long-term scientific capacity.
Bridget Scanlon: Yeah. So it is great that you are doing so well at Tsinghua University, and that's one of the premier universities in China. And I guess the current Chairman Xi was a graduate from Tsinghua University.
Di Long: Yeah, chemical engineering, yes.
Bridget Scanlon: Yeah, that's pretty interesting and the fact that the government supports the students fees and that frees you up then, that you don't have to try to get grant funding then to cover their tuition and all of that sort of thing.
So it's, it is a big help. So you have been working in a variety of different topics since you returned to China, Di and one of the first ones I'd like to discuss with you is the Tibetan Plateau and its role as the Asian Water Tower and providing the headwaters for many of the major rivers in China and surrounding countries.
And then also providing water with the electricity energy production, but also irrigation downstream, supporting about a billion people downstream. maybe you can describe for us a little bit the geomorphology, the physiology, the climate, and of the Tibetan Plateau.
And then as your research findings on the role of the Tibetan plateau in the bigger water system surrounding it.
Di Long: [00:17:19] Okay, so thank you, Bridget. The Tibetan Plateau really is one of the most fascinating and important regions I've ever worked on, both scientifically and personally. So to start with a sense of scale, the Tibetan Plateau is enormous. It covers roughly 3 million square kilometers above the size of the Mississippi River Basin in the US.
So the Mississippi river basin is around the 3.2 million square kilometers. The Tibetan Plateau has an average elevation of over 4,000 meters above sea level. So because of the extreme elevation it is often caused the third pole of the earth, and hydrologically, the plateau is quite incredibly diverse in the central Tibetan Plateau.
Many basins are endorheic lakes, meaning they are internally draining water flows into lake, but doesn't reach the ocean. This is why you see such a dense network of lake there, many of them large, remote and minimally affected by human activity. And in contrast, along the margin of the plateau, the basins are exoreic and form the headquarters of some of the Asias large rivers.
So from west to south to east. This includes the Indus, Ganges, Brahmaputra, Salween, Lancang (Mekong), Yangtze, and Yellow rivers. So together these rivers supply water to more than 1 billion people downstream, making the plateau critical for water security across much of Asia. And the plateau is also extraordinary in terms of surface water.
They are more there and 7,000 glaciers on the plateau itself, covering about 83,000 square kilometers. And lakes are just as remarkable. The plateau is one of the most densely distributed lake regions on earth. With round 1400 lakes greater than one hundred square kilometers.
These lakes occupy about 50,000 square kilometers, representing about 60% of China's total lake area. And many of these lakes have been expanding rapidly in recent three decades, largely driven by changes in glacier melt. And in terms of climate, it plays a central role in shaping all of this.
The plateau acts as a massive elevated barrier in the atmosphere. And moister air actually is forced upward by the terrain correctly. And they release precipitation, a process known as the orographic effect. And what makes the plateau unique is that it's not just a mountain range, but a vast elevated land mass.
Precipitation over the plateau is primarily controlled by two major systems. The South Asian summer monsoon brings moisture from the Indian Ocean, and meanwhile, the mid latitude westerlies transport moisture from regions like the Mediterranean and the Iranian plateau dominating precipitation over the Western and the Northern plateau.
And from the scientific perspective, the plateau is especially valuable because it is highly sensitive to climate change, yet relatively free from direct human disturbance, especially in the lake interior. This means that changes in various variables like lake area and lake water storage provide a very clean signal of how the regional hydrological system is responding to climate change.
We often think of the plateau as a natural lab. By combining satellite remote sensing, hydrological modeling, and now AI based data fusion, we're able to track how glaciers snow, lakes and rivers are evolving and what that means, not only for the plateau itself, but for downstream regions that depend on it.
So our research probably relies on using multi source satellites, remote sensing, including GRACE satellites and optical remote sensing images to check changes in glacier lakes. And we incorporate this sort of information into hydrological modeling to understand the changes in hydrological processes, including snow melt, glacial melt, and the rainfall runoff processes.
Bridget Scanlon: Yeah. it plays a huge role in the hydrology then of Asia. And as you mentioned the third pole and so much of it above four kilometers in elevation and such a large area. So it would be very difficult to evaluate the hydrology from ground-based approaches.
It shows the value of remote sensing then to try to understand how the system is evolving. And very interesting work you did on looking at the lake water storage and its response to climate. And you mentioned, I guess two of the main drivers of climate there. The westerlies, which means water coming from the west. And then also the South Asian monsoon system from the Indian Ocean. And the role of orographic precipitation, then supplying water to downstream users. So you're working at the lake water storage, you we're covering the period from about 2000 to 2017 and did a detailed analysis of water levels and storage changes in the lake using GRACE and altimetry data.
And focusing on about 50 very large lakes, greater than a hundred square kilometers. And talking about expanding water storage in these lakes over that time period, and some of the plots in the paper show five to 10 meter increases in lake level height.
And then also there are concerns about potential glacial lake outbursts like, Lake Kusai and Lake Salt, and the potential impact on downstream populations. So maybe you can describe the results of that work a little bit.
Di Long: Okay, thank you Bridget. We actually conducted a detailed analysis of lake water levels. The surface is changed and the storage changes between 2000 and 2017 focusing on large lakes were satellite signals are most reliable. So in total we analyzed about 50 lakes larger than a hundred square kilometers.
And what we found was quite striking. The majority of the lakes showed persistent expansion, both in surface area and in water storage. Over this 17 year period in terms of what levels a typical increase was on the order of five to 10 meters, which is very large when you think about the lake systems at this scale.
They are also these are not short-term fluctuation. They represent a sustained accommodation of water over nearly two decades. And we, when we translate those level increases into storage, the signal becomes even clearer. Many lakes gain substantial volumes of water driven primarily by increase the precipitation and enhance the glacial melt and a warming climate.
Because these lakes are mostly located in the lake basins, the water essentially has nowhere to go. It accumulates in the lake system and evaporates to back to the air. And at the same time, this changes with important water storage, particularly related to Glacier Lake outbursts, the floods or growth.
And this is something I have been experiencing very personally. In Spring 2023, I joined the field expedition to the remote source region of the Youngs River. During that trip, my former student Dr. Han, who is now associate professor at Dalian University of Technology and two other team members were traveling in the same vehicle when they accidentally slide into the river.
They were trapped in the water for nearly an hour before being rescued by others. So fortunately, everyone survived, but the instance really highlighted how remote, harsh, and dangerous these environments can be. So access is extremely limited, where the changes rapidly, and even basic field work carries significant risk.
This experience reinforced for me why we can't rely solely on in situ observations in regions like the Plateau. And this is where remote sensing becomes indispensable. So satellite geometry, optical imagery and SAR allow us to continuously monitor lake levels, surface expansion, and sudden changes even in places that are nearly impossible to reach safely.
So in a sense Tibetan Plateau Lakes serve two roles at once. They are first indicators of climate achieving hydrological changes, and they also are potential sources of hazards. So understanding both aspects, long-term storage change and short term instability is critical for risk assessment and downstream water management.
Yes.
Bridget Scanlon: Yeah, that, that really highlights how difficult it is to do field studies in those regions. I remember going, visiting glaciers in northwest China, and people warning us because we drove up to the glacier and they said, you have to walk very slowly because the oxygen levels are so low and everything.
I was with an Australian colleague and she was much younger than I was, and she ran out there and the next minute she was keeling over because there wasn't enough oxygen. So it's great that we have remote sensing data, so GRACE water storage data, and also synthetic aperture radar for altimetry data.
And you have done that type of analysis for all of the reservoirs globally using altimetry data, very helpful in understanding the system. And so you mentioned, this is internally draining system, so it's not impacting downstream except potentially glacial outbursts. But if it's sourced from glacier melting, then this expansion, how long before the glaciers would be potentially completely melted, and then that source will go away and maybe you won't have continued expansion of the lakes forever.
Di Long: Yes. The lake will be, so the storage and the water level of the lake will be dependent on a number of factors including temperature and also the evaporation of the lakes' water storage. If the temperature continues to rise, I think evaporation from the lakes will also increase.
The lakes even though they are currently expanding they were not as such kind of expansion may not be persistent for decades ahead, yeah.
Bridget Scanlon: so if the source glacial melt, then that could be a transient pulse over decades. I can know the river flows northwest China, but there were sourced from glacial melting. There was high river flows in response to the glacier melting, but that won't last forever.
So important to consider these factors when you are thinking about infrastructure and water resources management. Di, one of the aspects of your research that I really appreciate is how thorough and comprehensive you do the analysis and you use a variety of different sources and you approach the problem from the top down, for example, in the Tibetan plateau with the satellite data, and then also from the bottom up look.
And then you combine and fuse all the satellite data and then use models and AI to better understand the system. So I guess your work at the Plateau, you showing that total water storage increased about 10 cubic kilometers per year. And so for the US listeners 1 million acre feet is about 1.2 cubic kilometers.
So the units are fairly similar, and that's over the entire plateau. And in the endoreic region, then the internally drained region about five and a half cubic kilometers. But decreasing trends in the exoreic basins, about minus 16 cubic kilometers. Maybe you can describe the exoreic basins and their role then an impact on downstream river flows.
Di Long: Okay. So yes, Bridget at the scale of the entire Tibetan Plateau, total water storage showed a net decline of about 10 cubic kilometers per year. As you just mentioned but this overall number masks strongly contrasting regional behaviors. One we separated the plateau into the lake basins and exoreic lake basins that feed major rivers.
The contrast was striking. So the regions, mainly across the central and the Northern Plant plateau, we observed a significant increase in water storage on the order of five to six, six cubic kilometer per year. So this increase was driven primarily by widespread lake expansion as we just discussed which along contributed nearly 5.8 cubic kilometers per year.
So combined with localized the glacial mass in regions such as the, and the Western Karakoram Mountains, and in contrast, lake basins, especially those joining South and west. Including the Indus and Ganges and Brahmaputra headwaters experienced substantial water storage losses around the minus 16 cubic kilometers per year.
So these losses were dominated by negative glacier mass balance associated with regional warming and the changes increased where snowfall increasingly turns into rainfall, glaciers in region such as the Hindu Kush have been retreating, contributing roughly about 10 cubic kilometers per year to water storage to decline.
What's particularly important is that these spatial patterns align well with large scale atmospheric changes. So over the recent decades, south Asian monsoon intensity has weakened, but mid latitude westerlies have strengthened. And this shift helps explain why southern Monsoon influenced the basin losing water while northern and interior basin influenced by westerlies are, at least for now, gaining water.
So in terms of the implications, those contrasting trends matter quickly. Changes in total water storage directly affect water availability, particularly melt water that supports irrigation and ecosystems downstream. And at the same time, they influence natural hazard risks such as glacial collapses and Glacier Lake outburst floods.
So these processes can amplify compound risks and geopolitical tensions in densely political downstream regions that depend on water originating from the plateau. So taking together this study highlights how the Asian Water Tower is not responding uniformly to climate change. Instead, it's fragmenting into regions of gain and loss and understanding that heterogeneity is essential for both signs and water management.
Bridget Scanlon: I think, oftentimes we just want a simple answer, Di. We just, is it increasing? Is it decreasing? And are we in trouble or whatever. But , what your studies highlight is the variability in what's happening throughout the Tibetan plateau. And that's very important to understand.
So it's not all decreasing, so it's increasing in the north and decreasing in the south, and then understanding what's causing those increases and decreases glacier melting and the atmospheric circulation pattern. So you really, your work really links all of those things together.
And if we want to manage things in the future, we need to understand the processes that are operating and have been operating in the past. To develop appropriate water management solutions. Kudos to you and your team for doing very detailed analysis and not just coming with a simple answer.
One of the things with satellite data is that many people think that we didn't know anything or do anything before we had satellites. And so that gets a bit frustrating, I think, for people who have been working a long time in hydrology. But I really appreciated the work you did on reconstructing stream flow over past centuries using tree ring data and also proxies for the drought PDSI and things like that.
Maybe you can describe that a little bit for us, those long-term reconstructions of stream flow over past centuries , Di and what you learned from that.
Di Long: Okay. Thank you, Bridget. Because instrumental stream flow records on the Tibetan Plateau are very short, mostly since the 1960s, they don't capture the full range of natural variability. So to address this we constructed annual stream flow back to around 1200 CE using tree ring based PDSI data. Technically we were using canonical analysis combined with log linear regression. We were able to explain about 64 to 70% of the variants in observed stream flow at five key head water stations, including the Yellow, Yangtze, and Lancang (Mekong) rivers. So what we find is that the past 800 years include many wet and dry periods that exceed the five, the fifth, and 95 percentile range of the instrumental record.
In particular in the Northern part of the Plateau, we identify several extremes and prolonged drought episodes that are not captured by mode observations. In other words, what we often think of as a stream today is not unprecedented when viewed over longer time scale, but the context matters.
And one notable example from this study occurs in the mid 1600's corresponding to the Baroque period. So when we constructed the stream flow in northern rivers, which is exceptionally low levels and persists for couple of years, this extreme drought conditions coincide with well documented historical mega drought in China, and are widely recognized as one of the major natural stressors contributing to the collapse of the Ming dynasty through impacts on agricultural food security and social stability.
And a particular striking result is a strong north south contrast with a dividing line around 32 to 33 degree in the north. North of this line, the Yellow, Yangtze, and Lancang headwater conditions tend to be drier when the south is wetter and vice versa. For example, the headwaters were generally wetter in earlier centuries, while many northern rivers were drier.
This opposing behavior persists across about 10 major wet dry cycles. And in terms of drivers, ENSO is dominant at inter-annual timescale, especially three to eight years while PDO and AMO and the Indian Ocean play a larger role at decadal scales. So together these results highlight a strong nonstationarity in the plateau hydrology, and show that recent changes must be interpreted against the backdrop of very large natural variability, something that is crucial for future water management during the climate change.
Yes.
Bridget Scanlon: It was very interesting analysis and it's great to go back to 1200 period using the tree rings and then the gridded data on the Palmer drought severity index from the tree ring data. And then you had stream gauge data from the sixties, and then you were able to look at the relationship between the droughts and wet cycles over centuries. And compared that then with what we have seen from the instrumental record, the fifth to 95th percentiles of stream flow. And I think one of your graphs in the paper was very striking in showing that you had a lot of periods that were much drier than what we have seen in the instrumental record below the fifth percentile, and then some periods that exceeded the 95th percentile.
And another plot in the paper that I found very interesting was the contrast in wet and dry periods. So when it was wet in the north, the dry in the south and vice versa. And it's great that you were able to link it to history and the Ming Dynasty and their fate, it's really nice to be able to combine all sorts of data to increase the reliability of your analysis and enforce what you are looking at.
I really enjoyed that paper. Di, you and your team have done a lot of work and there's a lot of discussion these days on the Mekong River and the term, I guess it's called the Lancang River in China. Looking at the relationship between climate, hydrology, and hydropower with the reservoirs along the Lancang and maybe you can describe your analysis of the impact of those reservoirs and the different phases of reservoir development filling stage and then when it's when the reservoirs are full and then afterwards when they're operating the reservoirs, their impact on downstream water availability.
Di Long: Okay. I would like to start with briefly introducing the attributes of the Lancang/Mekong River Basin. This river is one of Asia's most important transboundary river systems. It originates on the Tibetan Plateau and flows through six countries before discharging into the South China Sea.
The basin has a large topographic gradient dropping by more than five kilometers from source to mouth, and along the Chinese reach along the river descends about 4.5 kilometers over roughly 2200 kilometers. So this steep relief creates enormous hydropower potential, but also makes the system hydrologically complex.
And in terms of the climate, the basin is dominated by the Indian Ocean monsoon and the westerly precipitation increases sharply from northwest to southeast from about seven hundred to 800 millimeter per year in the upper basin to over 3000 millimeter in parts of Laos and Cambodia in terms of the hydrological regime around 75 to 85% of annual runoff occurs during the wet season from June to November with peak flow, typically in August or September.
And in terms of the hydropower development a cascade of large dams has been constructed in the river in China, with installed capacity of about 5.9 gigaWatt (GW) of electricity with 1.75 GW associated with Jinghong Dam. And so the Mekong/Langcang mainstem contributes on the order of about 1% of the China's total installed electricity capacity.
And this is a classic example where climate forcing, extreme topography, and the large-scale hydropower interact, understanding the system requires separating natural monsoon driven variability from human induced flow regulation. And in terms of our study we combined hydrological modeling under the tree ring reconstruction together with remote sensing dataset to try to understand the impacts of dam construction on the flow for three different periods.
So we actually we explicitly separated the river hydrology into three phases. The first is a pre-impact period, the transition period during dam construction and this is the second stage. And the third phase is a post impact operation period. During the pre-impact period from 1980 to 1986, they observed the stream flow closely followed the natural climate driven variability. So using data combined with our hydrological model, the RS hydrological model, we constructed the natural runoff, which represents the response to climate.
And for the transition period from 1987 to 2007 it marks dam construction and the reservoirs filling So during this period, we see the clear recent negative impact on total stream flow, especially at upstream station, such as changes in Thailand and the reservoir filling, reduce the annual discharge, creating a growing divergence between observed and natural flow. And this reflects the water being temporarily stored to fill that and active storage.
And in the post impact operational period especially from 2008 to 2014 and after the signal changes and the total annual stream flow gradually returns towards nearly natural levels, indicating that the strongest impact occurs during filling rather than long-term observation.
But the signal redistribution of flow is substantial water flow operations reduce wet season flows and increased dry season flows. This has benefits flood mitigation and increases in the dry season which can be especially valuable during recent extreme drought.
However, there are also negative consequences. One key concern is the sediment trapped that would normally nourish the flood plains and the delta. So reduce the sediment delivery increases risks of delta subsidence, coastal erosion, and ecosystem degradation.
So in summary, dam filling causes temporary reduction in total stream flow and the dam operation stabilizes annual volumes, but fundamentally alters signal timing. The system experiences clear trade-offs, improving flood and drought regulation upstream, but growing downstream concerns relative to subsidence, ecosystems health, and long-term delta sustainability.
Bridget Scanlon: Yeah, that's fascinating and really nice that you disaggregated temporally then the construction and filling of the reservoirs and their impact on flow. And then you were able to compare with the pre reservoir period to know what the natural system looked like. And then also quantified the impacts of dam operations then on downstream flow and providing more flow during droughts and retaining water during flood periods, evening out the water supply in order to support hydroelectricity, then you want a fairly uniform discharge and that elevation difference along the Lancang/Mekong River is amazing. And I guess the hydropower potential is huge, as you said, one to five gigawatts. Very nice study.
I guess the last topic I would like to discuss a little bit, and we are running out of time, but is your work on the North China Plain. I mean there are a lot of discussion on global hotspots of groundwater depletion in many regions.
And the Upper Ganges, North China Plain was one of them, parts of the High Plains Aquifer in the US and the California Central Valley. But in China then you have built the south to north water diversion system to transfer water from the Yangtze River in the south, the humid south to the North China Plain which is a semi-arid region to support municipalities, you have huge population density there in the North China Plain, and also a lot of agriculture.
So would really appreciate if you can describe that a little bit for us.
Di Long: Okay, so let me start with describing the geography and the climate for the North China Plain. So it is one of the world's most severe groundwater depletion hotspots, but it is also one of the creators of Chinese agriculture. And this region is characterized by deep aquifer and alluvial soils formed by long-term deposition from the Hai, Yellow, and Huai rivers and their tributaries.
So because of this exceptional soil fertility, the Plain has been a green producing region since ancient times supporting this population and intensive framing well before modern irrigation existed. So the fundamental challenge in today arises from a structural mismatch between climate and agriculture.
About 60% of annual rain falls during the summer monsoon, while the dominant crop, winter wheat, followed by summer maize requires irrigation in the dry spring. For decades, this gap was filled almost entirely by groundwater pumping. Well, the plain is bounded by the Taihang Mountain to the west and the Yanshang mountains to the north with rivers heavily dammed since the 1950s for flood control.
As a result, surface water availability declined sharply and cities like Beijing and much of the Plain became overwhelmingly dependent on groundwater. Over the past four decades, groundwater levels fell by 20 to 60 meters, representing the cumulative decline of roughly 180 cubic kilometers.
So a turning point came with a large scale human intervention, especially the South to North Water Diversion project. The central route since 2014, water from the Dangjiangkou Reservoir on the Yangtze system has been delivered along 1200 kilometer canal, supplying Beijing, Tianjan, and major cities.
So between 2015 to 2024, this route delivered about 45 cubic kilometer of water comparable to the long term Yellow River imports. Well, in terms of our study, by combining GRACE satellites, and large scale source water and groundwater coupled modeling, we found that the increased precipitation accounted for about 45% of recent recovery and water diversion and ecological replenishment account for roughly 30%. And water conservation and demand management contribute the remaining 25% and the results are striking. After decades of decline, groundwater storage shifted from losses of up to minus two to minus five cubic kilometers per year to a net gain of about 6.5 cubic kilometers per year after 2020.
And groundwater levels have been recovering at a rate of about 0.7 to 0.9 meters per year. So recovery is the strongest in urban areas like Beijing and mainly in shallow aquifers. The system comes with real energy and treatment costs. There are ongoing concerns about ecological impacts, especially in the water source region, particularly the impact of a flow reduction on the lower reaches of the Han River.
So the key lesson is not that diversion alone fixed the depletion, but that large scale aquifer recovery is possible when climate variability, infrastructure and governance are aligned through integrated water accounting. Yes.
Bridget Scanlon: [00:54:19] Yeah, I mean it's amazing, a huge project, with three separate routes and you described the central route from the reservoir on the Yangtze to the North China Plain. And I recall visiting Shijiazhuang in the North China Plain many years ago. And they were monitoring groundwater levels declining by a meter and a half per year.
And the winter wheat is when you really need all of that irrigation and that heavy reliance on groundwater and now bringing in surface water. And it's somewhat similar to what they've done in California with the State Water Project and the Central Valley Project with the aqueducts taking water from the humid north to the Central Valley to support irrigation there and conjunctive management of surface water and groundwater.
But I guess the difference is you have huge populations in Beijing, Tianjin, and most of the diverted water is going to those municipalities, but then that frees up water for agriculture. And so I really appreciate that you were able to disaggregate the impacts of climate variability and increased precipitation since 2020. And also conservation and other things rather than just simply saying, they're diverting the water and, the groundwater is recovering,
Di Long: Yes. Yeah, right. Yeah.
Bridget Scanlon: And that's it. So it's very important to link the policies and the infrastructure to the impacts because it's occurring within a context of many other things with climate and other processes.
Di I guess the last thing is, your team has been using, you've described a lot of applications of remote sensing and also, instrumental data monitoring and modeling analysis and I guess you increasingly using artificial intelligence and machine learning.
Maybe you can describe how you see the use of AI and ML and how you see things going forward. How you will use AI and ML to improve your workflow.
Di Long: Yeah. Thank you, Bridget. In the remote sensing area we have been facing key challenges. That is the tradeoff between the spatial and the temporal resolution provided by different satellite platforms. For example, let say, one provides long-term records, but suffers from sparse temporal coverage and data gaps and other sensors like MODIS offer higher temporal resolution but lower resolution.
So the tradeoff has limited our ability to detect the seasonality, long-term change and extremes of the surface water bodies consistently at the global scale. So what we do is a few multiple satellite data sets using the U-Net model operating on both cloud platform such as Google Earth Engine, and on supercomputers at our university to explore both the spatial and the temporal resolution. So AI helps us optimize across sensors, constrain uncertainties, and generate for example consistent lake surface extent globally. So even under the cloudy conditions or during rapid hydrological transitions, and based on the technological framework described above we decompose lake dynamics into three distinct components.
The long-term trends reflecting decadal expansion. The second is the inter-annual variability driven largely by wet-dry climate cycles. And the third is the seasonality capturing inter-annual transitions. So what we found at the global scale, the seasonality dominates the lake surface dynamics.
Seasonally dominated lakes account for about two thirds of global what we tested is based on the hydro basin lake data sets covering 1.4 million lakes globally. And approximately 6% of all lakes worldwide is also dominated by the seasonality. At slightly more than 93% of the population reside in basins with more than half of the lakes dominated by seasonality.
More broadly, this work shows how AI power to remote sensing combined with big data and the physical understanding, allows us to move from static maps to dynamic process-oriented monitoring of inland waters at the planetary scale. That is essential if we want to understand how lakes respond to climate extremes and long-term change in the coming decades.
Bridget Scanlon: That's amazing. Di and I think in NASA and other groups are using AI a lot to to fuse data. And I guess we've always known some satellites have high temporal resolution with low spatial resolution MODIS maybe 250 to 500 meters, and then high revisit times the high frequency revisits, but then LandSat, maybe 30 meter spatial, but then every 15, 16 days revisits. Being able to fuse these different products then helps you to optimize spatial and temporal resolution and then address new issues. And your analysis of global reservoirs is really exciting. And also the different types of satellites you mentioned, cloud cover impacting optical satellites, and then radar satellites not impacted.
So the more different types of data that we can combine together and AI is really key to that. And I think in the past, maybe we spent half our time downloading data and then now being able to do it in Google Earth Engine, and avoiding that step. So it really is advancing a lot. Thank you so much for taking the time to describe your research today.
Our guest today is Di Long, who is a professor in hydrology and remote sensing at the Department of Hydraulic Engineering at Tsinghua University. And we look forward to seeing your future advances in hydrology, Di, with your team. Thank you so much.
Di Long: Thank you Bridget for your thoughtful guidance and your suggestions are really helpful. Thank you, and looking forward to talking and collaborating with you.
