Timing of Global Surface Water Transitions Reveals Anthropogenic Dominance of Recent Growth in Water Extent

doi:10.21203/rs.3.rs-3215886/v1

Download PDF

Article

Timing of Global Surface Water Transitions Reveals Anthropogenic Dominance of Recent Growth in Water Extent

https://doi.org/10.21203/rs.3.rs-3215886/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The extent of coastal and inland surface water resources is constantly varying in response to complex interrelated processes, driven by natural and anthropogenic factors. Recent advance in satellite technology and cloud computing have enabled global-scale monitoring of the changing occurrence and extent of these surface water resources. However, until now, no previous study has sought to estimate the timing of these surface water changes at the global-scale. Here we introduce the first global-scale identification of the year when water advanced or receded within a given pixel, using a 38-year Landsat time series. Our methods focus exclusively on persistent changes in water features, filtering out seasonal or short-lived fluctuations. We use the new algorithm to map the timing of water advance and/or recession events globally, encompassing both inland water bodies and coastal dynamics. Additionally, the timing of water transitions enabled the identification of the primary drivers behind these changes. As a result, we identified that most of the large-scale water change events are related to human influence, such as damming, infrastructure failures and even conflicts. These combined factors contributed to a global shift, with accumulated water advancing surpassing water receding over time.

Earth and environmental sciences/Hydrology

Earth and environmental sciences/Environmental sciences/Environmental impact

Earth and environmental sciences/Climate sciences/Hydrology

Earth and environmental sciences/Climate sciences/Climate change/Climate-change impacts/Environmental health

Earth and environmental sciences/Ocean sciences/Physical oceanography

Surface water is an indispensable resource that plays a fundamental role in shaping our planet's landscapes and supporting human development ^1–6. The dynamic interplay between wet and dry conditions has driven profound changes throughout Earth's history, sculpting diverse landforms, influencing ecosystem dynamics, and impacting human civilizations. These changes can be attributed variously to human activities such as urban expansion and the construction of dams and levees, as well as to natural processes such as river meandering ^2,7−15. Climate change also influences the distribution of surface water through its impacts on precipitation variability, temperature changes and meting of ice ^16–18. Establishing the occurrence and timing of transitions of surface water state (i.e., transitions between wet and dry conditions, or vice versa) is, therefore, crucial for understanding the complex processes underlying landscape evolution and assessing their implications for sustainable human development ^6,19.

To address the challenge of studying surface water on a global scale, in recent years researchers have turned to satellite imagery and advances in cloud computing to enable the detection and estimation of various water resource variables at the global scale. Examples of such prior studies include mapping the presence or absence of surface water, assessing its temporal occurrence, classifying the behavior of surface water changes (e.g., do quantified surface water extents vary seasonally, are they permanent?), monitoring reservoir dynamics, and identifying areas of permanent water loss and gain ^12,19,20. Pekel, et al. ²⁰ and Donchyts, et al. ¹⁹ have made significant contributions to this field by mapping global surface water extent and changes using Landsat imagery, with their studies shedding new light on the spatial distribution and interannual variability of Earth’s surface water. However, whilst these previous approaches have focused on identifying areas of changing water occurrence, they ignore the timing of such changes, potentially limiting their utility for evaluating causation.

In this study we present the first global analysis of the timing of surface water transitions using Landsat satellites images from 1984 to 2022 within the Google Earth Engine (GEE) cloud computing environment (Fig. 1). Herein we define such temporal transitions as occurring either when a pixel was dry, and subsequently becomes persistently flooded or with seasonal water presence (inundated during the flooding period of the region), or when a pixel was wet, and subsequently becomes persistently dry or experiences only scarce flooding events (which occur less frequently than inundation periods). Our algorithm was specifically designed to minimize the impact of short-lived events (lasting one year or less) and, instead, it focuses on identifying transitions that substantively impact the 38-year history of each pixel. This approach lends confidence to the identification of shifts between dry and wet conditions. The derived dataset is publicly available on GEE and is also accessible through a GEE app that aids visual interpretation (https://gustavoonagel.users.earthengine.app/view/water-change-time-detection). The novelty of our research lies in its emphasis on understanding the timing of surface water transitions, and the implications of such transition timings for landscape evolution and human development.

Our surface water transition dataset provides a comprehensive representation of surface water temporal variability, capturing the timings and intricate patterns of change influenced by diverse natural and anthropogenic processes (Fig. 1). For example, in regions like China (Fig. 1.b) and Dubai (Fig. 1.g), distinctive patterns²¹ shaped by human land reclamation and food production activities stand in stark contrast to the natural patterns of change observed on rivers and floodplains (Fig. 1.c, 1.e, 1.h). Additionally, our dataset reveals that human interventions tend to occur over relatively short time intervals, such as the artificial infilling of the Três Irmãos Reservoir in Brazil that occurred abruptly in 1991 (see histogram below Fig. 1.f), the construction of the Palm Islands in Dubai, which experienced its peak of land reclamation in 2005 (histogram below Fig. 1.g), or the fish lakes that were filled in 2015 following Amazonian deforestation (Fig. d). In contrast, natural influences on water dynamics operate over longer time spans, such as in the Ganges delta (Fig. 1.h), which displays a more distributed and similar rates of water advance and recession over time (histogram below Fig. 1.h), the gradual evolution of an Amazonian River (Fig. 1.e), or the long-term droughts affecting a Turkish reservoir (Fig. 1.a). These natural processes are indicated in Fig. 1 by smooth transitions from darker red and blue colors (indicating earlier timings of transition) to lighter colors (representing more recent transitions). Furthermore, different scales of change are also captured, as exemplified in Fig. 1.a, which shows a lake of approximately 70.2 km² that experienced drought and a nearby lake of only 0.08 km² that was filled in 1994.

Major and rapid events, such as reservoir filling, coastal expansion, and severe droughts, can lead to substantial advance or recession of water bodies within just a few years. Figure 2 illustrates the timing of such transitions, at a global scale, in the period from 1984 to 2022, highlighting portions of the globe where transitions have occurred over spatially-coherent regions. Furthermore, the years in which the biggest water change was related to direct human interventions, such as dam construction, flooded agriculture expansion, poor water management, and poor infrastructure planning, is also highlighted (this information was gathered using visual interpretation and based on the literature).

In regions where water is receding, 21% of the recession years were driven by direct human interventions or human-induced events, such as the Aral Sea disaster. The Aral Sea disaster is a prominent example of major water recession caused by the Korakal dike construction in the 1960s by the Soviet Union to divert water to agriculture projects ^22–24. Considering the period since 1984, the most significant water recession event in the Aral Sea region occurred in 2004, surpassing any other water recession events in the world in that year and throughout the entire analysis period. In 1996, the Aral Sea, albeit on a smaller scale, was also the largest area of global water recession. Other lakes with recognized human intervention also experienced water recession in other periods, such as the Colhué Huapi in Argentina (1994) and Lake Urmia in Iran (2009), both affected by the over exploitation of water ^25,26. Furthermore, although on a smaller scale, in East Asia, particularly the coasts of China and South Korea are expanding, with a general peak in 2012, due to the construction of ports and expansion of cities. The highlighted Saemangeum Seawall, for example, is the largest dike on Earth, and although completed in 2006, the most significant impact was observed in 2012 due to land reclamation behind the 33-km long sea-wall ²⁷. Moreover, the identification of the year of water transitions has proven valuable in understanding the strategic manipulation of water resources during conflicts. A notable example is the Russian annexation of Crimea in 2014, where Ukraine's response of blocking the North Crimean Canal led to the drying of the canal and associated reservoirs in Crimea ²⁸ (peak of water recession in 2014).

In regions where water advanced, 66% of the years that experienced water change events were driven by direct human intervention, predominantly attributed to the construction of dams, which accounted for 55% of the water advance years. In the Amazon region, the Balbina Dam, filled in 1989, stands out as a significant water advance event that occurred within just one year, resulting in the highest infilling of water anywhere globally in 1989. This example serves as a reminder of the rapid and substantial impact that reservoirs can have on the environment, highlighting the importance of careful planning for additional dams in the region^29,30. Other examples include the Serra da Mesa Dam (1997) and Porto Primavera Dam (1999) in Brazil, the Merowe Dam in Sudan (2005), the Bui Dam in Ghana (2006), and the Murum Dam in Malaysia (2012). However, in a similar way that the Aral Sea disaster was attributed to poor infrastructure planning, the largest recorded water advance event was also linked to another disaster that occurred in the Garabogazki Basin Lake in Turkmenistan, which displays peak water advances in 1993 and 1994. A breach in 1992 of a dam between the lagoon and the Caspian Sea led to a rapid influx of water into the basin³¹, substantially increasing the water surface area by around 2772 km². Overall, these examples demonstrate the significant impact that human activities have on the advance of surface water globally.

Besides human-influenced events, the analysis also identified natural or indirectly human-influenced occurrences of water advance and recession. However, even in the absence of direct human influence, these areas might still be susceptible to the effects of climate change in driving drought and flooding regimes. In Southeast Australia, for example, prolonged droughts are completely drying up many reservoirs ^32–34, such as in 2001 when the Australian Lake Menindee experienced the world’s largest water recorded recession transition. In Afghanistan, many lakes in the Registan desert completely dried up in 2005 (collectively representing the largest recession of water in the world in that particular year). In South America, the Pantanal wetland region and the downstream Paraná River also experienced severe droughts ^35–38, with a water retreat peak displayed in the year 2000 in the Pantanal highlighted region. California in North America also faced reduced rainfall events, leading to a decline in reservoir capacity ^39–41 (with a water retreat peak in 1991 in the highlighted Lake Malheur). Changes in the climate can also lead to an increase in water extent. The lakes in the Tibet Plateau, for example, experienced the most significant water advance in the world in 2005. These lakes are growing in response to the impacts of climate change, driven by rising precipitation levels, glacier loss, and permafrost thawing in the region ^42,43. Higher precipitation rates also contributed to the filling of the highlighted Quill Lakes and surrounding prairie potholes in North Dakota^44,45 (USA), with a peak of water advance in 2012, and the Toshka Lakes in Egypt, resulting in significant water expansion in 2020 (the largest advance of water in the world in that year). The Toshka Lakes are closely connected to the Nile River and were influenced by increased flow from the upstream regions of Sudan, which experienced higher precipitation rates ⁴⁶.

Considering the accumulated amount of water change, global areas of water advance and recession initially grew at similar rates until 2012, when the total areas of water advance began to surpass the total areas of water recession (Fig. 3.a). The cumulative area of water advance from 1984 to 2022 was estimated as 323,326,000 km², while the extent of water recession was 236,892,000 km². Of the total area of water advance, 56% was observed in high residence time inland water areas, such as reservoirs, lakes, wetlands, and flooded agriculture, while 29% occurred in rivers, 12% in coastal regions, and 3% in deltas. For water recession, the proportions were 60%, 26%, 11%, and 3%, in inland waters, rivers, coasts and deltas, respectively. However, the timings of the transitions of water advance and recession vary substantially across continents. For example, South and Central America and Europe had similar rates of water advance and recession, despite having slightly higher water recession areas overall. In South and Central America, water recession areas were on average 12% higher than areas of water advance, which aligns well with estimates by MAPBIOMAS for Brazil during the same period (15%)⁴⁷. North America, characterized by the highest percentage of water change occurring in lakes and reservoirs (73%), initially had more areas of water recession, but experienced a shift in 2001 when areas of advancing water surpassed areas of receding water. Similar patterns are observed in Africa (lake dominated), where a crossover from water recession to water advance occurred in 2002. Likewise South Asia (the only river dominated region) crossed from water recession to water advance dominant in 2011, while East and Southeast Asia (with more balanced proportions) crossed in 2022 and Central Asia (reservoir dominated) did so in 1993. In contrast, the South Pacific, and especially the Middle East, were the regions with the highest relative proportions of water recession, with a discrepancy of 19.8% and 43.6% between water recession and water advance, respectively.

The new global water transition timing dataset captures a wide range of natural and anthropogenic processes, providing valuable insights into global water dynamics. The emphasis on timing of water transitions allows the identification of human-induced rapid events that changed large swaths of water in a short time interval (one year), revealing the construction of dams around the world, major droughts, infrastructure failures, and even the effects of conflict on water extents. Direct human influence was responsible for the largest water change events in 21% of the years for recession events and 66% of the years for water advances (55% of the years were related to dam construction). The infrastructure failures were responsible for the largest historical water advance, related to a breach in the Caspian Sea³¹ in 1992, that inundated large areas in 1993 and 1994, and water recession events, related to the Aral Sea disaster^22–24, with a peak of recession in 2004. These events show how poorly planned human interventions can have huge impacts on the environment. Furthermore, the construction of large dams surpassed any other water advance event in more than half of the analysed years (21 years). The Balbina Dam, for example, which filled its reservoir in 1989 in the Amazon, was the largest artificial reservoir build during the analysed period. This is a stark example of how dams in the low-gradient lowlands of the Amazon produce large reservoirs, which in turn submerge large areas of forest.

These dam constructions certainly played a role in the surpassing of the accumulated areas of advance over accumulated areas of recession in the different continents of the world, such as North America, which transitioned to this state in 2002, Africa (transitioned in 2000), South Asia (transitioned 2011), East and Southeast Asia (transitioned in 2002), and Central Asia (transitioned 1993). In East Asia, in addition to the surge of dams⁴⁸ and the spread of flooded agriculture (more than 90% of the global rice is produced in the Asia–Pacific region⁴⁹), climate change is also impacting the rates of precipitation and ice melting, which are filling up the natural lakes of the Tibetan Plateau^42,43. Furthermore, in North America, despite the water retreat observed on the US West Coast (mainly before 2010), the filling of prairie potholes in North Dakota in 2012, mainly caused by higher precipitation^44,45, also contributed to the positive surface water balance of the continent. In South and Central America, despite the great spread of dams, the severe droughts of Pantanal, with peak in 2000 and 1994, and Northwest Brazil (peak in 2012) might have contributed to equilibrate the rates of water advance and recession. However, the Middle East and South Pacific had higher rates of water recession, which means that the overall water surface greatly decreased between 1999 and 2011 in both regions. In both desertic regions, many lakes were completely depleted by droughts ^{32–34, 50,51}, such as the Menindee Lake and many others in Southeast Australia (peak of recession in 2001) and in some lakes of Afghanistan (2005).

Nonetheless, the continents with positive water balance are not necessarily in a better water security condition, since the spatial distribution of water transition is not even, as shown in the previous examples. Areas of water advance can also have the opposite effect (on water security), such as the expansion of areas of flooded agriculture, which can reduce water security since it increases water evapotranspiration rates⁵². Water security and conflicts in general can significantly impact water dynamics, as in the case of the Crimean Canal blockage. While currently on a smaller scale, the impact of conflicts on water may escalate in the future due to population growth and climate change, leading to decreased water availability in certain regions ^53,54. Overall, the new dataset presented in this paper is able to identify globally important water transitions across the planet, related to drought, reservoir construction and infrastructure projects, and allows comparison of the timing and rates of water transition between continents. The attribution of water changes to human influence was focused on large-scale events, as global information on whether water changes are shaped by humans or natural influences is not readily available. Identifying infrastructure failures and overexploitation of water was even more challenging, and thus we relied on existing literature for these cases. We anticipate that the dataset will be of considerable interest to researchers seeking to decipher the intricate relationships between hydrological processes, erosion rates, landform evolution and the development and sustainability of human activities. The dataset is publicly available on Google Earth Engine (GEE) and an Earth Engine App has been developed for convenient data visualization and retrieval (https://gustavoonagel.users.earthengine.app/view/water-change-time-detection ).

The unique combination of its high spatial resolution (30 m), long duration of archive (spanning from 1984 for Landsat 5 to 2023 for Landsat 9), and a revisit time of 16 days positions the Landsat program as a highly suitable tool for undertaking a comprehensive analysis of global inland and coastal surface water. To accurately detect the specific year of change, we developed two algorithms utilizing Landsat Time Series from 1984 to 2023 and the cloud computing platform Google Earth Engine (GEE). The first algorithm, the Water Change Detection (WCD) Algorithm, estimates areas of surface water transition by leveraging the Modified Normalized Difference Water Index (mNDWI)^55–64, while the second algorithm, the Water Time Track Detection (WTTD), identifies the year of change through the use of the Green_Red Normalized Difference Water Index (grNDWI) proposed herein (see Fig. S1 in Supplementary Information, SI). Regions with a Landsat dataset starting only after 1990 were excluded from the analysis (see Fig. S5 in SI)

4.1 Water Change Detection Algorithms

To delineate areas of changing water extent, we implemented the Water Change Detection (WCD) algorithm on annual Modified Normalized Difference Water Index (mNDWI) images derived from pre-processed Landsat 5, 8, and 9 data, covering the period from 1984 to 2023 (see Fig. S4 in SI for details and Fig. 4.a for a summary). Through a pixel-wise analysis, areas of potential change were identified based on the crossing of a pre-determined mNDWI threshold (see Fig. S3 in SI), indicating either water advance (upward crossing) or water recession (downward crossing) (Fig. 4.b). Recognizing the dynamic nature of surface water, where regions may experience recurrent episodes of flooding or drying, we employed a further analysis of each pixel's mNDWI trend, the trend being determined from Ordinary Least Squares Regression as computed using annual mNDWI over time. Pixels that exhibited both an upward crossing of the threshold and had a regression line slope exceeding 0.005 mNDWI/year were considered as areas with a transition to water advance. Conversely, pixels that displayed both a downward crossing of the threshold and a trend of less than − 0.008 mNDWI/year were classified as regions of water recession (Fig. 4.c). The determination of these trend values was obtained by extracting slope samples from areas previously identified as undergoing erosion and sedimentation processes (see Fig. S3 in SI for details).

To determine the timing (here we resolve timing at annual resolution) of water advance or recession, we employed the Water Time Track Detection (WTTD) algorithm (Fig. 4). This algorithm was specifically designed to filter out short-lived events, such as seasonal flooding and droughts. In areas identified by the Water Change Detection (WCD) algorithm as experiencing water advance or recession, we constructed a new time series using the Green_Red Normalized Difference Water Index (grNDWI). Figure 4.d to 4.k illustrates the processing steps applied to a single pixel identified as water receding, with similar procedures followed for water advancing pixels. The Landsat 5, 8, and 9 images underwent pre-processing procedures including cloud filtering, cloud and snow masks, and the generation of annual grNDWI composites (see Fig. S4 in SI for details). However, even with the use of annual composites, cloud cover can result in missing data for several years (Fig. 4.d), particularly in regions prone to cloudiness like the Amazon. To address this issue, we developed a technique to fill in data gaps in the time series by utilizing the preceding and subsequent grNDWI years (Fig. 4.e – details are provided in Fig. S4). With the new, uninterrupted, time series we extracted the maximum and minimum grNDWI values to calculate the mean grNDWI, which served as the pixel-specific water-land threshold (Fig. 4.2.f). This pixel threshold represents a locally optimized water/land threshold, acknowledging that a pixel transitioning from forest to turbid water would have a different threshold than a pixel transitioning from sand to clear water, even if they were in close proximity. We note that, by relying on the entire time series, the pixel threshold method is susceptible to potential outliers, such as unfiltered clouds and atmospheric constituents, which can affect the accuracy of the results. To address this, we computed the mean between the individual pixel threshold and a regional threshold (calculated using the OTSU method, see Fig. S2 in SI), resulting in the creation of the Optimal Threshold that incorporates both local and regional influences (Fig. 4.g).

The newly derived Optimal Threshold serves as the basis for determining the transition year of water advance or recession. For any individual pixel to be defined as experiencing water advance, it must cross the Optimal Threshold in an upward direction, while water recession pixels must cross it in a downward direction, as depicted in Fig. 4.h. Considering the influence of short-lived flooding events, an individual pixel may cross the Optimal Threshold multiple times, resulting in several candidate transition years (Fig. 4.h). To select the actual year of transition used in the analysis, we calculated the mean grNDWI using a 5-year moving window and determined the year at which it crossed the Optimal Threshold, establishing a reference year (Fig. 4.i). By comparing the candidate years with the reference year, we selected the candidate with the smallest difference (Fig. 4.j). In the provided example, we encountered two candidate years and chose the more recent one (Fig. 4.k). The transition year is then stored as part of the pixel dataset that covers the entire Earth.

To summarize the water change dataset, we employed a hexagonal grid system with each hexagon covering an area of 7,179 km². Within each hexagon, we determined the year with the most significant water change and quantified the extent of both water recession and advance. This approach allowed us to identify major recession and/or advances and their corresponding years of occurrence. Additionally, we calculated the accumulated area of water change for different continents. To assess the proportion of change occurring in various water environment types, such as high residence time inland waters (reservoirs, lakes, wetlands, and flooded agriculture), rivers, coastal regions, and deltas, we further subdivided the world into smaller hexagons measuring 1,010 km². By estimating the percentage of each water environment within these smaller hexagons (using datasets provided by Google Earth Engine), we associated these percentages with the accumulated areas of water recession and advance within each hexagon. Then, we summed up the water change associated with each water environment type to determine the percentage of change associated to lakes, rivers, coastal, and deltas. Fig. S7 in the Supplementary Information (SI) provides a visual representation of this process.

4.2 Validation

We conducted a validation process by collecting 700 geographically diverse data points across the globe. To facilitate this, we developed a user-friendly Google Earth Engine (GEE) App that produced a random distribution of points and displayed a Landsat Time Series with the NIR, Red, and Blue band combination to highlight distinctions between water and land. Through careful visual analysis, we determined the most appropriate year of transition in relation to water advance and recession and compared that year with the year of change as estimated using the WTTD algorithm. The determination of the most appropriate year was performed by identifying the year that water advanced at each point and remained flooded, and the year that the water receded and remained dry. This validation procedure encompassed a range of different types of water bodies such as rivers, floodplains, flooded agricultural areas, lakes, coasts, and delta regions. Key statistical metrics were derived to evaluate the performance of the WTTD algorithm. These metrics include the Mean Absolute Percentage Error for combined water advance and recession points (MAPE), MAPE for water advance points (MAPEadv), MAPE for recession points (MAPErec), the Coefficient of Determination (R²), and the Critical Success Index (CSI), thereby establishing a comprehensive assessment framework (see Fig. 5) to ensure the accuracy and reliability of our findings. An overall error distribution can be visualized in Fig. 5.h, and for each water feature in Fig. S6 in the SI.

Our analysis yielded a Mean Absolute Percentage Error (MAPE) of 13.9%, indicating a high level of forecasting accuracy, as supported by previous studies (Montaño Moreno et al., 2013). The coefficient of determination (R²) was 0.82, demonstrating a strong correlation between the predicted and observed data points, while the overall CSI was 95.7% when considering all water bodies in a single regression model (Fig. 5.a). These statistics indicate the robustness of the WTTD algorithm, considering the intricate and dynamic nature of global water surfaces. Notably, the accuracy of the WTTD algorithm is seen to vary across different water features, with lakes and coasts exhibiting the best results (Fig. 5.e and 5.f), characterized by similar MAPE values of 11.1%, CSI values of 96% and 98.9%, and high R² values of 0.85 and 0.93, respectively. Rivers (MAPE: 15.5%, CSI: 95.5%, and R²: 0.79) and deltas (MAPE: 18.5%, CSI = 100%, and R²: 0.85) also demonstrated good accuracy (Fig. 5.b and 5.g). However, floodplains and flooded agriculture areas, while showing good MAPE values (15.6% and 13.2%, respectively) and good CSI values (88.2% and 100%, respectively), exhibited substantial disparities between the estimated and observed years, resulting in decreased R² values (0.67 and 0.69) (Fig. 5.c and 5.d). Remarkably, floodplain regions showed a concentration of misclassifications, with 6.7% of the predicted points not correctly experiencing persistent changes. Moreover, it is important to highlight that regions of water advance generally exhibited better estimations compared to water receding regions.

The variations in accuracy can be attributed to the inherent complexity of different water features and the extent to which water colour and water extend change over time, which directly impact the performance of both the WCD and WTTD algorithms. Coastal regions, for instance, are influenced by a combination of factors such as ocean currents, wave energy, tides, and coastal morphometric characteristics ^65,66. However, these influences tend to persist over longer durations, and it is unlikely for an eroding coast to transition into a sediment-filled one within the time span of our analysis. Additionally, coastal areas generally exhibit lower complexity in terms of their water colour when compared to inland water bodies, conditions that increased the accuracy of the water change algorithms. On the other hand, rivers and floodplains are subject to interannual rainfall variations, seasonality effects, and water turbidity fluctuations, which introduce uncertainties in the analysis. Floodplains, in particular, have relatively shallower inundated areas compared to rivers, thus with a greater colour influence from the submerged surface bottom. Furthermore, they are highly influenced by seasonal variations and short periods of drought and flooding, which adds complexity and potentially leads to reduced accuracy in the assessment of floodplain regions. In the case of flooded agriculture, although they tend to be as shallow as floodplains, they are less affected by rainfall variability as their areas of inundation are artificially controlled. This artificial manipulation contributes to greater stability, resulting in higher accuracy compared to floodplain regions.

Lakes generally exhibit higher overall accuracy, primarily attributed to the rapid global construction of new dams ⁶⁷. The water change algorithms can easily detect the year when a reservoir is filled, as it undergoes a transition from a dry to a persistently flooded state. However, lakes that are highly dynamic and sensitive to rainfall conditions tend to have lower accuracy. Nonetheless, certain characteristics are common to all water features. In all water features, pixels representing water advance tend to have higher accuracy compared to pixels indicating water recession. This is primarily because inundation or erosion processes tend to occur more rapidly, making it easier to identify the transition year. In the case of meandering rivers, for instance, erosion typically occurs in the deeper outer bank, which experiences lower seasonality. Conversely, sedimentation gradually accumulates over time in the shallower inner bank, which is more influenced by seasonal fluctuations, a more complex condition that reduces the water change accuracy. However, the development of the WCD and WTTD algorithms aimed to mitigate these complexities, enabling the monitoring of surface water bodies worldwide.

Konapala, G., Mishra, A. K., Wada, Y. & Mann, M. E. Climate change will affect global water availability through compounding changes in seasonal precipitation and evaporation. Nature Communications 11, 3044, doi:10.1038/s41467-020-16757-w (2020).
Sengupta, D. et al. Mapping 21st Century Global Coastal Land Reclamation. Earth's Future 11, e2022EF002927, doi:https://doi.org/10.1029/2022EF002927 (2023).
Salerno, L. et al. Satellite Analyses Unravel the Multi-Decadal Impact of Dam Management on Tropical Floodplain Vegetation. Frontiers in Environmental Science 10, doi:10.3389/fenvs.2022.871530 (2022).
Cai, X., Feng, L., Hou, X. & Chen, X. Remote Sensing of the Water Storage Dynamics of Large Lakes and Reservoirs in the Yangtze River Basin from 2000 to 2014. Scientific Reports 6, 36405, doi:10.1038/srep36405 (2016).
Li, Y. et al. An index and approach for water extraction using Landsat–OLI data. International Journal of Remote Sensing 37, 3611–3635, doi:10.1080/01431161.2016.1201228 (2016).
Luijendijk, A. et al. The State of the World’s Beaches. Scientific Reports 8, doi:10.1038/s41598-018-24630-6 (2018).
Ielpi, A., Lapôtre, M. G. A., Gibling, M. R. & Boyce, C. K. The impact of vegetation on meandering rivers. Nature Reviews Earth & Environment 3, 165–178, doi:10.1038/s43017-021-00249-6 (2022).
Ma, Y. et al. Identifying Dike-Pond System Using an Improved Cascade R-CNN Model and High-Resolution Satellite Images. Remote Sensing 14 (2022).
Liman Harou, I., Whitney, C., Kung’u, J. & Luedeling, E. Mapping Flood-Based Farming Systems with Bayesian Networks. Land 9, 369 (2020).
Camporeale, C., Perucca, E. & Ridolfi, L. Significance of cutoff in meandering river dynamics. Journal of Geophysical Research 113, doi:10.1029/2006jf000694 (2008).
Shields Jr, F. D., Simon, A. & Steffen, L. J. Reservoir effects on downstream river channel migration. Environmental Conservation 27, 54–66, doi:10.1017/S0376892900000072 (2000).
Donchyts, G. et al. High-resolution surface water dynamics in Earth’s small and medium-sized reservoirs. Scientific Reports 12, 13776, doi:10.1038/s41598-022-17074-6 (2022).
Bussi, G. Impact of dams and climate change on suspended sediment flux to the Mekong delta. Science of the Total Environment 755, 142468 (2021).
et al. Analysis of Meandering River Morphodynamics Using Satellite Remote Sensing Data—An Application in the Lower Deduru Oya (River), Sri Lanka. Land 11, 1091 (2022).
Nagel, G. W. et al. Impacts of meander migration on the Amazon riverine communities using Landsat time series and cloud computing. Science of The Total Environment 806, 150449, doi:https://doi.org/10.1016/j.scitotenv.2021.150449 (2022).
Taylor, C., Robinson, T. R., Dunning, S., Rachel Carr, J. & Westoby, M. Glacial lake outburst floods threaten millions globally. Nature Communications 14, 487, doi:10.1038/s41467-023-36033-x (2023).
Tabari, H. Climate change impact on flood and extreme precipitation increases with water availability. Scientific Reports 10, doi:10.1038/s41598-020-70816-2 (2020).
Mukherjee, S., Mishra, A. & Trenberth, K. E. Climate Change and Drought: a Perspective on Drought Indices. Current Climate Change Reports 4, 145–163, doi:10.1007/s40641-018-0098-x (2018).
Donchyts, G. et al. Earth's surface water change over the past 30 years. Nature Climate Change 6, 810–813, doi:10.1038/nclimate3111 (2016).
Pekel, J.-F., Cottam, A., Gorelick, N. & Belward, A. S. High-resolution mapping of global surface water and its long-term changes. Nature 540, 418–422, doi:10.1038/nature20584 (2016).
Tarolli, P., Cao, W., Sofia, G., Evans, D. & Ellis, E. C. From features to fingerprints: A general diagnostic framework for anthropogenic geomorphology. Progress in Physical Geography: Earth and Environment 43, 95–128, doi:10.1177/0309133318825284 (2019).
Micklin, P. The Aral Sea Disaster. Annual Review of Earth and Planetary Sciences 35, 47–72, doi:10.1146/annurev.earth.35.031306.140120 (2007).
Small, I. & Bunce, N. The Aral Sea disaster and the disaster of international assistance. Journal of International Affairs 56, 59–73 (2003).
Lioubimtseva, E. in Biological and Environmental Hazards, Risks, and Disasters (Second Edition) (eds Ramesh Sivanpillai & John F. Shroder) 435–447 (Elsevier, 2023).
Rahimi, A. & Breuste, J. Why is Lake Urmia Drying up? Prognostic Modeling With Land-Use Data and Artificial Neural Network. Frontiers in Environmental Science 9, doi:10.3389/fenvs.2021.603916 (2021).
Zucca, C., Middleton, N., Kang, U. & Liniger, H. Shrinking water bodies as hotspots of sand and dust storms: The role of land degradation and sustainable soil and water management. CATENA 207, 105669, doi:https://doi.org/10.1016/j.catena.2021.105669 (2021).
Moores, N., Rogers, D. I., Rogers, K. & Hansbro, P. M. Reclamation of tidal flats and shorebird declines in Saemangeum and elsewhere in the Republic of Korea. Emu - Austral Ornithology 116, 136–146, doi:10.1071/mu16006 (2016).
Shumilova, O. et al. Impact of the Russia–Ukraine armed conflict on water resources and water infrastructure. Nature Sustainability, doi:10.1038/s41893-023-01068-x (2023).
Latrubesse, E. M. et al. Damming the rivers of the Amazon basin. Nature 546, 363–369, doi:10.1038/nature22333 (2017).
Tundisi, J. G., Goldemberg, J., Matsumura-Tundisi, T. & Saraiva, A. C. F. How many more dams in the Amazon? Energy Policy 74, 703–708, doi:https://doi.org/10.1016/j.enpol.2014.07.013 (2014).
Kosarev, A. N., Kostianoy, A. G. & Zonn, I. S. Kara-Bogaz-Gol Bay: Physical and Chemical Evolution. Aquatic Geochemistry 15, 223–236, doi:10.1007/s10498-008-9054-z (2009).
Squire, D. T. et al. Likelihood of unprecedented drought and fire weather during Australia’s 2019 megafires. npj Climate and Atmospheric Science 4, doi:10.1038/s41612-021-00220-8 (2021).
Holgate, C. M., Van Dijk, A. I. J. M., Evans, J. P. & Pitman, A. J. Local and Remote Drivers of Southeast Australian Drought. Geophysical Research Letters 47, e2020GL090238, doi:https://doi.org/10.1029/2020GL090238 (2020).
et al. Adapting Urban Water Systems to a Changing Climate: Lessons from the Millennium Drought in Southeast Australia. Environmental Science & Technology 47, 10727–10734, doi:10.1021/es400618z (2013).
Marengo, J. A. et al. Extreme Drought in the Brazilian Pantanal in 2019–2020: Characterization, Causes, and Impacts. Frontiers in Water 3, doi:10.3389/frwa.2021.639204 (2021).
de Barros, A. E. et al. Wildfires disproportionately affected jaguars in the Pantanal. Communications Biology 5, 1028, doi:10.1038/s42003-022-03937-1 (2022).
Lázaro, W. L., Oliveira-Júnior, E. S., Silva, C. J. D., Castrillon, S. K. I. & Muniz, C. C. Climate change reflected in one of the largest wetlands in the world: an overview of the Northern Pantanal water regime. Acta Limnologica Brasiliensia 32, doi:10.1590/s2179-975x7619 (2020).
Correa, D. B., Alcântara, E., Libonati, R., Massi, K. G. & Park, E. Increased burned area in the Pantanal over the past two decades. Sci Total Environ 835, 155386, doi:10.1016/j.scitotenv.2022.155386 (2022).
Allen, R. J. & Anderson, R. G. 21st century California drought risk linked to model fidelity of the El Niño teleconnection. npj Climate and Atmospheric Science 1, 21, doi:10.1038/s41612-018-0032-x (2018).
Wang, S. Y. S., Yoon, J.-H., Becker, E. & Gillies, R. California from drought to deluge. Nature Climate Change 7, 465–468, doi:10.1038/nclimate3330 (2017).
Diffenbaugh, N. S., Swain, D. L. & Touma, D. Anthropogenic warming has increased drought risk in California. Proceedings of the National Academy of Sciences 112, 3931–3936, doi:10.1073/pnas.1422385112 (2015).
Zhang, G., Chen, W. & Xie, H. Tibetan Plateau's Lake Level and Volume Changes From NASA's ICESat/ICESat-2 and Landsat Missions. Geophysical Research Letters 46, 13107–13118, doi:https://doi.org/10.1029/2019GL085032 (2019).
Zhang, G. et al. Response of Tibetan Plateau lakes to climate change: Trends, patterns, and mechanisms. Earth-Science Reviews 208, 103269, doi:https://doi.org/10.1016/j.earscirev.2020.103269 (2020).
Kharel, G. & Kirilenko, A. Considering Climate Change in the Estimation of Long-Term Flood Risks of Devils Lake in North Dakota. JAWRA Journal of the American Water Resources Association 51, 1221–1234, doi:https://doi.org/10.1111/1752-1688.12300 (2015).
Huang, S. et al. Simulating the water budget of a Prairie Potholes complex from LiDAR and hydrological models in North Dakota, USA. Hydrological Sciences Journal 58, 1434–1444, doi:10.1080/02626667.2013.831419 (2013).
Abd Ellah, R. G. Morphometric analysis of Toshka Lakes in Egypt: A succinct review of geographic information systems & remote sensing based techniques. The Egyptian Journal of Aquatic Research 47, 215–221, doi:https://doi.org/10.1016/j.ejar.2021.05.005 (2021).
Souza, C. M. et al. Reconstructing Three Decades of Land Use and Land Cover Changes in Brazilian Biomes with Landsat Archive and Earth Engine. Remote Sensing 12 (2020).
Miao, C., Borthwick, A. G. L., Liu, H. & Liu, J. China’s Policy on Dams at the Crossroads: Removal or Further Construction? Water 7, 2349–2357 (2015).
Sun, C. et al. Twenty-meter annual paddy rice area map for mainland Southeast Asia using Sentinel-1 synthetic-aperture-radar data. Earth System Science Data 15, 1501–1520, doi:10.5194/essd-15-1501-2023 (2023).
Sidiqi, M., Kasiviswanathan, K. S., Scheytt, T. & Devaraj, S. Assessment of Meteorological Drought under the Climate Change in the Kabul River Basin, Afghanistan. Atmosphere 14 (2023).
Qutbudin, I. et al. Seasonal Drought Pattern Changes Due to Climate Variability: Case Study in Afghanistan. Water 11 (2019).
Wang, X., Chen, Y., Li, Z., Fang, G. & Wang, Y. Development and utilization of water resources and assessment of water security in Central Asia. Agricultural Water Management 240, 106297, doi:https://doi.org/10.1016/j.agwat.2020.106297 (2020).
Kundzewicz, Z. W. & Kowalczak, P. The potential for water conflict is on the increase. Nature 459, 31–31, doi:10.1038/459031a (2009).
Schillinger, J., Özerol, G., Güven-Griemert, Ş. & Heldeweg, M. Water in war: Understanding the impacts of armed conflict on water resources and their management. WIREs Water 7, doi:10.1002/wat2.1480 (2020).
Xu, H. Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery. International Journal of Remote Sensing 27, 3025–3033, doi:10.1080/01431160600589179 (2006).
Acharya, T. D., Subedi, A. & Lee, D. H. Evaluation of Water Indices for Surface Water Extraction in a Landsat 8 Scene of Nepal. Sensors 18, 2580 (2018).
Du, Y. et al. Water Bodies’ Mapping from Sentinel-2 Imagery with Modified Normalized Difference Water Index at 10-m Spatial Resolution Produced by Sharpening the SWIR Band. Remote Sensing 8, 354, doi:10.3390/rs8040354 (2016).
Herndon, K., Muench, R., Cherrington, E. & Griffin, R. An Assessment of Surface Water Detection Methods for Water Resource Management in the Nigerien Sahel. Sensors 20, 431, doi:10.3390/s20020431 (2020).
Deoli, V., Kumar, D. & Kuriqi, A. Detection of Water Spread Area Changes in Eutrophic Lake Using Landsat Data. Sensors 22, 6827, doi:10.3390/s22186827 (2022).
Chen, F. et al. Open water detection in urban environments using high spatial resolution remote sensing imagery. Remote Sensing of Environment 242, 111706, doi:https://doi.org/10.1016/j.rse.2020.111706 (2020).
Worden, J. & de Beurs, K. M. Surface water detection in the Caucasus. International Journal of Applied Earth Observation and Geoinformation 91, 102159, doi:https://doi.org/10.1016/j.jag.2020.102159 (2020).
Gautam, V. K., Gaurav, P. K., Murugan, P. & Annadurai, M. Assessment of Surface Water Dynamicsin Bangalore Using WRI, NDWI, MNDWI, Supervised Classification and K-T Transformation. Aquatic Procedia 4, 739–746, doi:https://doi.org/10.1016/j.aqpro.2015.02.095 (2015).
Sarp, G. & Ozcelik, M. Water body extraction and change detection using time series: A case study of Lake Burdur, Turkey. Journal of Taibah University for Science 11, 381–391, doi:10.1016/j.jtusci.2016.04.005 (2017).
Rokni, K., Ahmad, A., Selamat, A. & Hazini, S. Water Feature Extraction and Change Detection Using Multitemporal Landsat Imagery. Remote Sensing 6, 4173–4189, doi:10.3390/rs6054173 (2014).
Davis, R. A. & Hayes, M. O. What is a wave-dominated coast? Marine Geology 60, 313–329, doi:https://doi.org/10.1016/0025-3227(84)90155-5 (1984).
Coca, O. & Ricaurte-Villota, C. Regional Patterns of Coastal Erosion and Sedimentation Derived from Spatial Autocorrelation Analysis: Pacific and Colombian Caribbean. Coasts 2, 125–151, doi:10.3390/coasts2030008 (2022).
Zhang, A. T. & Gu, V. X. Global Dam Tracker: A database of more than 35,000 dams with location, catchment, and attribute information. Scientific Data 10, 111, doi:10.1038/s41597-023-02008-2 (2023).

There is NO Competing Interest.

SupplementaryInformation.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Timing of Global Surface Water Transitions Reveals Anthropogenic Dominance of Recent Growth in Water Extent

Status:

Version 1

Abstract

Figures

1. Introduction

2. Results

3. Discussion

4. Methods

4.1 Water Change Detection Algorithms

4.2 Validation

References

Additional Declarations

Supplementary Files

Status:

Version 1