Introduction

Human activities have triggered a series of critical environmental problems, one of which is plastic pollution. Microplastics (MPs), defined as small pieces and particles of plastic less than 5 mm in diameter1, were first reported in inland waters of Lake Huron2, Rhine River3, and Three Gorges Dam4 in 2011 and 2015, respectively. Since 2014, more studies have reported the distribution of MPs in global inland waters5,6,7,8,9,10. These MPs, with diverse characteristics of colors, shapes, sizes, and polymer types, enter inland waters through runoff, sewage discharge, and atmospheric deposition11,12.

MPs are extremely mobile in aquatic environments due to their properties of lightweight, insolubility, and durability13, making them easily ingested by aquatic animals. Consequently, animals may suffer from choking or starvation due to pseudo-satiety14. Smaller MPs readily accumulate through food chains, causing bioenrichment that results in reduced hatching rates, impaired growth, behavioral disruptions, and changed feeding preferences15. Additionally, MPs can adsorb pollutants such as heavy metals16, polycyclic aromatic hydrocarbons, and polychlorinated biphenyls17. Moreover, many substances added during plastic production, such as pigments, stabilizers, and plasticizers, can continuously release toxic or even carcinogenic additives18.

Inland waters are vital for human living and the development of industry and agriculture. However, rivers are known to transport MPs from terrestrial to marine ecosystems. Lebreton et al.19 reported that the 122 polluting rivers contributed more than 90% of total plastic inputs to the ocean19. Moreover, during the long process of transportation, some MPs are intercepted on riverbanks, while many MPs remain submerged in water, affecting the health of aquatic environments20. The MPs in lakes and reservoirs have also received considerable attention21. Due to their relatively closed characteristics, lakes also act as important sinks for MPs. Yet they can also become significant sources of MPs once being disturbed by environmental factors, such as wind exposure, and facilitate the transport of MPs to the outlet22. Artificial barriers such as dams and weirs also play a significant role in intercepting plastic into the ocean in freshwater bodies19.

MPs in inland waters have diverse potential sinks, including discharge into the sea through rivers19, deposition on riparian soils23, and settling in sediments24, or ingestion by aquatic organisms25. Being a significant sink for MPs, sediments can reflect historical inputs of plastic wastes26. However, MPs in sediments are prone to resuspension once influenced by hydraulic conditions and climate factors27,28, thereby being released into surface water. Activities of leaching, bioturbation, and cultivation in soils may influence the migration of MPs, while this process is exceedingly slow29,30. Indeed, the process of MPs deposition on riparian soils is not unidirectional. As earlier evidence suggests, water-driven soil erosion has been indicated to promote the immigration of MPs from cultivated land soils to freshwater31. Another study on MPs transport from headwaters to the estuary of the Yangtze River Basin also suggests that the immigration of MPs in the watershed is more pronounced in water rather than in soils or sediments32. Consequently, the complicated behaviors of MPs between water and riverine sediments and soils in freshwater ecosystems lead to greater harm to long-time transportation in freshwater systems.

Experiential settings, such as sampling methods, extraction, and purification methods, are the dominant factor to determine the MPs detection in the water, followed by the human and environmental factors33. A previous study has indicated that there is no significant difference in MP abundance between grab sample with 0.45 μm filter and net sample with 100 μm mesh18. Yet, a meta-analysis has divided sampling methods into volume-reduced and bulk sample categories, with the volume-reduced samples overlooking smaller-sized MPs than the bulk samples33. Human activities, such as industrial, agricultural, washing, road usage, fishing, and wastewater treatment plants34,35,36,37,38, are often direct sources of MPs, increasing their abundance in inland waters. Additionally, agricultural activities, such as plastic mulching, may produce MPs in the soil39. MPs could move in soils affected by soil erosion, and the input of plant processes as well as various animals39, and may also be exported to freshwater systems through irrigation or stormwater runoff18,34,40. Natural environmental factors, such as the hydrology of rivers with climate variability, play significant roles in affecting MPs dispersal in the water. Inherent properties of MPs, including the density, size, shape, and polymer type, can also influence their transportation and distribution patterns in aquatic environments41.

Although many studies have examined MPs in inland waters, most focus on specific locations or review global freshwater data about sampling methods, presence, and impacts12,42,43,44,45,46,47. Only a few studies have assessed MP abundance in water globally48,49. Since the definition of MPs pollution is only recently proposed, the limitation that MP pollution goes unreported in some areas leads to a lack of knowledge about the global spatial distribution of MP abundance in inland waters49. A recent review study49 summarized the MP pollution in global freshwater systems including river, lake, estuary, and wetland, but it mainly focused on the global efforts, knowledge gaps, and research priorities for MP monitoring, providing limited information on the spatial distribution and the associated driving factors of MPs in global inland waters. Machine learning, an essential technique in ecology, has the potential to establish models and predict global distributions50,51. Here, we present a comprehensive meta-analysis of MP distributions, characteristics, influencing factors, and predicting maps for MP abundance in global inland waters and China. Given that China is the world’s largest producer of plastic materials, it was imperative to explore MP abundance in inland waters within the country and apply machine learning to understand MPs distribution in other unstudied regions. Our research addresses several research questions, including the identification of MP research hotspots, the most significant factors influencing MP abundance, and regional differences in MP abundance across predicting maps. We, thus, aim to provide the understanding and predicting MP distribution that can be used to prioritize efforts to mitigate MP concentration, especially the smaller particles, in global inland waters.

Results

Distribution and characteristics of microplastics in global inland waters

The number of articles reporting MP contamination in inland waters has steadily increased since 2014, particularly from 2018 to 2021 (Fig. 1b), indicating the growing concerns about MP pollution in these environments. The abundance of MPs in global inland waters varied from 0.00 to 4,275,800.70 items m−3, with an average of 25,255.47 ± 132,808.40 items m−3 (mean ± standard deviation). We found significant differences between the tools used to capture MPs, as demonstrated by the pump > grab > net (p < 0.001) (Fig. 1c, see Note S1 for details of tools), while different chemistry reagents had no significant effects on detection results of MPs (Fig. S3). Higher sampling volumes led to smaller MP abundance for grab and pump sampling, and filtering size and sieving size only significantly affected MP abundance under pump sampling, showing that the smaller size leads to higher MPs abundance (Table S3). Interestingly, our analysis revealed no significant difference in MP abundance among different ecosystems (p = 0.13) (Fig. 1d), whereas it presented that MP abundance in dry season was higher significantly than wet season (p < 0.001) and that in autumn was higher significantly than other seasons (p < 0.001) (Fig. 1e). Additionally, our results indicated that MP abundance was higher at the depths of 0.00 to 0.50 m compared to 0.50 to 55.00 m (Fig. S4).

Fig. 1: The microplastics (MPs) distribution of our database derived from 301 publications including 5365 observations.
figure 1

a Locations of samples for MPs published in inland waters. b The dynamic diagram of article published year or sampling year. c Abundance distribution of MPs under different sampling methods. Effects of d ecosystem type and e different seasons on MP abundance in inland waters. In the boxplot (c–e), black points are the estimated values using liner mixed-effect model. Different lowercase letters indicate significant difference among characteristics of MPs at 0.05 level. The same as below. ***p < 0.001; ns not significant.

Full size image

Regarding the colors of MPs, transparent, black, and blue were the most prevalent in global inland waters, with mean values accounting for 29.27%, 9.21%, and 8.02%, respectively (Fig. 2a). In terms of shape, fibers, fragments, and films comprised 38.25%, 13.28%, and 3.84%, respectively (Fig. 2b). Small-sized MPs (≤1 mm) constituted a significant proportion of 63.72% in inland waters (Fig. 2c). Moreover, the most prevalent polymers in the waters were polyester (PE), polypropylene (PP), and polyethylene terephthalate (PT), accounting for 15.90%, 13.90%, and 5.10%, respectively (Fig. 2d).

Fig. 2: Characteristics of microplastics (MPs) in our database.
figure 2

Distribution of a color, b shape, c size, d polymer of MPs in the global inland waters. In which we classify white into transparent, pellet into granular, and liner into fiber. ***p < 0.001.

Full size image

Drivers of microplastics in global inland waters

Our analysis using random forest showed a satisfactory prediction of MP abundance in global inland waters, explaining 86% of the variation in MPs concentration across all observations (Fig. 3a). The model identified the human development index (HDI), actual evapotranspiration (AET), potential evapotranspiration (PET), cropland extent, and land surface runoff (LSR), each of which accounted for 26.18%, 21.22%, 17.27%, 19.08%, and 16.24% of the variation of MP abundance of global predicting model (R2: 0.86, RMSE: 1.24), respectively (Fig. 3b). The partial dependence plots showed the non-liner relationship between variables and MP abundance (Fig. 3c–g).

Fig. 3: Results of the predicted microplastics (MPs) based on random forest.
figure 3

a Comparison of predicted and observed MPs abundance in inland waters after log transformation (the grey dashed line indicates the 1:1 line and the blue line indicates the regression line between predicted and observed values). b The relative importance of predictors in the model. c–g Partial dependence plots showing the dependence of MPs abundance in the global inland waters on predictors after log transformation (Blue lines presented the smooth curve fitted by LOESS regression). HDI human development index, AET actual evapotranspiration, PET potential evapotranspiration, LSR land surface runoff.

Full size image

Prediction of microplastics in inland waters of globe

The global predicting map revealed the regions with high MP abundance in Asia, Europe, Africa, and eastern United States (Figs. 4a, S6). In particular, the high MP abundance was predicted for China (Fig. 4a). MP pollution was lower near the Arctic Circle (Fig. 4a). MP pollution was more severe in the Northern Hemisphere, primarily concentrated between 10°N and 50°N (Fig. 4b). According to the predicting map of three sampling methods, we basically found that although there were similar trends overall, there were some differences (Figs. S7, S8). The prediction map for China indicated the highest values in central, eastern, and northwestern regions, especially in the estuary of the Yellow River (Fig. S9).

Fig. 4: Predicted map of microplastics (MPs) in inland waters based on random forest.
figure 4

a Global prediction map of MP abundance, and b latitudinal pattern of MP abundance (red line indicates the locally weighted regressions and yellow ribbon spans the range between the 5% and 95% quantiles). The predicted value was the average value of the results of 100 iterations of random forest model. Predicting MP abundance after log transformation was reserved to draw the map to avoid skewed distribution of the MPs values, while the values of the map need to log-transformed and then added to 1 for viewing. The spatial resolution of grid cells was 3 arcmin-by-3 arcmin.

Full size image

Discussion

Distributions and characteristics of microplastics in global inland waters

The study provides comprehensive and more recent insights into the extent and distribution of MP pollution in global inland waters. This finding highlights significant MP contamination in spatial distribution on a global scale based on machine learning. In comparison to global marine environments where MP abundance ranges from 0.01 to 660,000.00 items m−3 with an average of 20,463.30 ± 98,346.24 items m−3 (mean ± standard deviation)52, our study suggested that the average MP abundance in global inland waters was higher. Given receiving MPs from terrestrial environment and acted as conduits for MPs to the marine environment, freshwaters represent the most complex system referred to MPs transport and retention53. MP behavior in freshwater depends on their properties (size, shape, density) and environmental factors (water chemistry, flow conditions, UV exposure), which together control sinking, aggregation, and transport54, ultimately prolonging MP retention in freshwater ecosystems. The average MP abundance reported by Schmidt et al.55 for global rivers (37,659.70 particles 1000 m−3; standard deviation = 272,998.10; including zero values) differs markedly (five orders of magnitude) from our data, in part because their study relied on literature prior to April 2017. Yet many articles related to MP pollution in inland waters were published between 2017 and 2024 (Fig. 1b). In addition, other ecosystem types (e.g., lake, wetland) also have higher pollution of MPs in the water49,56,57. Therefore, considering an exhaustive data collection can significantly elevate the MP pollution level in the inland waters.

In fact, the efficiency of the sampling and processing methods can affect the abundance of MPs measured in inland waters. Pump and grab sampling were observed to be more effective than net collection, the results of which are consistent with the study of Zhao et al.49. In terms of net sampling, net with 330 μm mesh was the popular approach to collect the top water samples12, due to the lower size boundary of MPs of 333 μm suggested by National Oceanic and Atmospheric Administration of USA58. Net sampling with a greater mesh size always leads to the omission of smaller MPs, underestimating MP pollution in aquatic environments59. Lozano and Mouat (2009) found that using an 80 μm mesh yielded results 100,000 times higher than those obtained with a 450 μm mesh when concentrating water samples60. Compared with net sampling, the filter meshes of grab (pore sizes mainly ranging from 0.20 to 200 μm, accounted for 77.42% of the grab samples in our database) and pump (pore sizes mainly ranging from 0.45 to 75 μm, accounted for 65.77% of the pump samples in our database) were smaller and can retain more MPs. However, pump sampling has a larger sampling water volumes and deeper sampling depth compared to grab sampling, which may contribute to the improved efficiency results. A study suggested that compared to plankton nets with 75 μm mesh, pump sampling with 25 μm mesh was regarded as an overvaluation of MP abundance resulting from lesser water volumes and accelerating water flow61. Despite that the smaller filtering size significantly improves the capture ability of MPs in water, especially for pump sampling, but it should be carefully considered that the blockages by zooplankton and algae can also occur62. A previous review has indicated that 60–90 μm mesh size is recommended to obtain samples for grab and pump sampling62. Lower volumes result in larger orders of magnitude for MP abundance values in pump or grab collections, being consistent with previous studies63,64,65. Consequently, sampling volumes should be large enough to minimize overestimation decreased by scaling up results64. Large volume sampling (at least 500 liters) is recommended for the grab sampling to effectively assess the prevalence of MP in surface water66, while 1000 liters was regarded as the adapt volumes for pump sampling64. Unlike previous studies reporting the main method in net sampling12,49, our study indicated that the grab sampling (3353 observations from 188 publications) had dominated the database, the analytical results of which demonstrate the importance of data updating and set the stage for the next step in global prediction. Consequently, our findings support that future research should use the grab sampling with larger volumes (at least 500 liters) and a 60–90 μm mesh size. This is because the grab sampling is more common in freshwater samples and easier for the individual comparisons across regions, while the pump sampling is largely dependent on sampling volumes and filter pore size, leading to an overestimation.

Contrary to our assumption, the abundance of MPs in inland waters during the dry season was higher than in the wet season. This effect may be attributed to a combination of various factors, such as hydroclimate, hydrodynamic factors, land-use types, MP properties, and sampling times. Among these factors, the frequency and intensity of rainfall may have continually changed the MPs level in water. Early in the wet season, rainfall may result in flush event to transport MPs from land to the aquatic environment67 or re-suspend and transport river bottom MPs caused by fast water flow68, resulting in the higher MP abundance in this period. With the extended rainy season, MPs in waters may be diluted or washed downstream through increasing river water volume in the wet season69,70, especially after heavy rainstorm events37,71,72,73, leading to more than input of plastic waste from surface runoff. As being evidenced by Cheung et al.74, MP abundance in waters under a storm event dropped significantly within just two hours74. Additionally, compared to high precipitation in the wet season, reducing river flow and low velocity in the dry season may decrease the river water volume (i.e., surface-water area ratio to be smaller), resulting in a higher amount of MPs in the surface water70. Physical characteristics of MPs (e.g., size, shape, and density) also contributed to MPs with smaller size to remain in the surface water during periods of low flow75. Noteworthily, land-use types are dominant in MP pollution. Fan et al. has demonstrated that seasonal differences in MP abundance in an urban river network could be explained by the combined effect of the urban plastic production and regional precipitation71. Industrials and urban areas are higher producers of plastic production, where impervious surfaces could contribute to the formation of surface runoff and aid MPs from diverse sources (including point and nonpoint pollution) in the region to the aquatic environment during rainfall periods76,77, resulting in the higher accumulation of MPs. The MP abundance in urban water bodies during dry season may also be higher than that in the wet season, which is related to the point source associated with domestic and industrial land uses (e.g., direct dumping of waste, domestic sewage effluent)78,79.

The abundance of small-sized MPs (<1 mm) tends to increase, as larger plastic waste undergoes the weathering80. Our results revealed that the small-sized MPs constituted an average proportion of 64% in global inland waters, which can pose greater harm, including facilitating their translocation in aquatic organism tissues81,82, delayed development of nauplii to cyprids83 and even caused acute inflammation of human blood cells84. Being similar to marine environments, fibers and fragments were the predominant shapes of MPs in inland waters52. The shape of MPs can be indicative of their parent material, with fibers, for instance, primarily originating from textiles and related to washing processes85. Besides, fibers and fragments may be generally more dangerous than other shapes of MPs. It is worth that fibers of MPs pose greater dangers to Ceriodaphnia dubia compared to beads86. Fragmentation processes change surface features such as cracks, fractures, scratches, pits, and grooves87, resulting in exacerbating biofouling processes and further influencing particle distribution and pollution in the waters. The colors of MPs may originate from their parent plastic products, but the differences in colors can also arise from processes such as solar radiation, microbial degradation, and mechanical forces88. Additionally, different colors of MPs may be ingested by different fish species38. Polyester (PE), polypropylene (PP), and polyethylene terephthalate (PT) were the most prevalent polymers in the waters, largely because of their widespread use in packaging, pipes, agricultural film, fishing gears, electronic appliances, and other applications89.

Divers of microplastics abundance in global inland waters

Human activities (open dumping, unreasonable waste management, sewage discharge, landfill, and agricultural activities, etc.) dominate the generation and emission of MPs. Climatic factors such as temperature, precipitation, evaporation, and land surface runoff affect the distribution of MPs in inland waters. Being consistent with our hypothesis, HDI indeed acts as the most important driver to influence the MP abundance in inland waters.

The HDI represents crucial aspect of development, including education, economy, and health90. The non-liner relationship between HDI and MP abundance indicated that the higher HDI did not mean richer MP abundance in their countries. In other words, HDI is an indispensable but not absolute driver to explain abundance of MPs in freshwater. Areas with higher development and plastic consumption tend to have higher MP abundance. Additionally, unreasonable and disturbed plastic managements and the lack of environmental awareness among the public are another important reason91, such as poorly managed waste treatment plants (WWTP)12,92. Some countries have a higher HDI, such as Britain and Canada (both with average HDI values exceeding 0.80)90, while the predictions revealed that the abundance of MPs in these regions ranked in the middle or lower range on our prediction map. These countries took strong action to ban the production of cosmetics containing plastic beads in 2017 and 201842. Despite of the United States with high HDI values, few regions still have revealed higher MP production. Higher population density and urbanization in the eastern United States are more direct input sources of MPs. Indeed, it is estimated that the 110 million people living within 50 km of the Unite State coasts can produce 0.1 million metric tons of plastic litter to the nation surface waters each year93. Noteworthily, tire wear particles are the major contributor of MPs in urban land use94. It is reported that the United State emitted an estimated 1,797,480 metric tons of tire wear and tear annually, resulting in a per capita emission rate of 4.70 kg per year and ranking as the largest contributor of tire wear and tear globally35. However, early strict policies on MPs in some cities of the United State have appropriately mitigated pollution. For example, California (revealing lower MPs prediction abundance) banned use of the plastic bags in 2007 and California also became the first state to ban plastic bag related to retail stores in the entire state in 201495.

Evapotranspiration is related to diverse climate variables such as surface temperature, ultraviolet radiation, and precipitation. Raising air temperature leading to one of nonnegligible results is increasing of freshwater evaporation96, giving birth to the enlargement of MP concentration in the water. Temperature is also considered as an influencing factor of the bacterial diversity on plastic debris, as it alters the average growth rate of biofilm97, then affecting biodegradation progress of MPs. Chang et al.48 has also reported that global warming is projected to potentially increase MP abundance and alter the share of small-size MPs by impacting on evaporation, precipitation patterns, and ultraviolet radiation intensity48. Additionally, the change of temperature and precipitation may also disturb adsorption between MPs with sediments or other pollutants and then impact the MP concentration in surface water.

In general, the MP abundance in water showed a decreasing trend with the increase of land surface runoff (mainly including urban runoff, industrial runoff and agriculture runoff), supporting our previous statement that dilution of MPs with heavy rainfall in water bodies might be greater than the input of plastic waste from land surface runoff. Urban runoff is the great contributor of MPs transfer to aquatic environments98, and the larger potential cause is the poor plastic waste management99. Additionally, urban land has the greater potential to product wear-derived MPs particles, which are easily transported into aquatic water from roadsides and stormwater runoff94. Indeed, treated or untreated industrial runoff including effluent of WWTP are the importance pathway for MPs, as being early demonstrated100. In the agricultural areas, biosolids and compost, wastewater irrigation, mulching film, polymer-based fertilizer and pesticides, and atmospheric deposition are the sources of MPs in agricultural soils101, then entering surface water through stormwater runoff or irrigation. Among them, plastic mulching, sewage sludge, and application of fertilizers and compost are the main sources of MPs in agricultural soils39, and runoff in storm events have been regarded as the important channel of MPs into surface water18,34. It is reported that 1.44 million metric tons of plastic mulch in China are used to contribute to increase the crop yield, the remaining plastic mulch in soils of which varies from 7190 to 25,910 kg·km−247, and partly can be transported into aquatic water through runoff. Besides, approximately 50% of sewage sludge in Europe and North America was utilized as economical fertilizer on farmlands, contributing to an influx of up to 850 tons of MPs per million inhabitants into European agricultural soils102. Zhang et al.103 found that the content of MPs in soil was significantly higher after adding sludge compost every year103, while another study reported that in Germany alone, approximately 0.035–2.20 trillion MPs are added to agricultural soil through organic fertilizer annually104. Although irrigation has been shown to intensify differences in MPs between different compartments, leading to a significant difference of MP abundance in surface water which is significantly higher after irrigation than before irrigation. This migration process also occurs through runoff40.

Global patterns of microplastics in inland waters

Being consistent with our hypothesis, the variation of global patterns of MP abundance in inland waters was clearly presented, especially in Asia. This contrasting distribution variation in global inland waters was attributed to the high demand and production of plastics and the poor management of plastic waste, especially in underdeveloped or developing countries. Indeed, the production and consumption of plastics, particularly in packaging, building, and construction applications, significantly contribute to global plastics production, therefore leading to their pollution in terrestrial and aquatic ecosystems. As recently estimated, 6.10 Mt of plastic waste had leaked into rivers in 2019 alone. This highlights that the plastic waste will maintain for decades to leak into the ocean even if not improving the ability to manage waste in the future105.

Our findings clearly illustrate that Asia was a dominant region regulating the global patterns of MP abundance in inland water. This corresponds with the previous study of Lebreton et al.19, analytically estimating that plastic releases from Aisa rivers accounted for 86% of the total global input, and followed by Africa19. Lin et al.106 have also evidenced the high MPs abundance of rivers within some Asia countries, exceeding 10,000 items m−3106. Asia is home to about 60% of the world’s population due to about 30% of Earth’s total land area, while nearly most of them are developing countries, experiencing rapid development and simultaneously facing growing environmental problems107. Consequently, plastic production in Asia has been at a high level, accounting for 52% of the world’s total production in 2021 (390.70 million tonnes)89. Especially, the recycling rate of plastic waste in these developing countries is much lower than that of developed countries. Meanwhile, these developing countries usually have a high percentage of mismanaged plastic waste, 12 Asian countries (China, Indonesia, and Philippines ranked the top three) of which had ranked by the mass of mismanaged plastic waste among the top 20 countries for 201093.

China plays a substantial role in the global plastics market, accounting for approximately 32% of total world production in 202189. Despite playing an important role in the production of the world, the per capita consumption in China is low. According to the reports from 2014, the per capita consumption in China was 45 kg, comparing to 65 and 109 kg per person in Europe and the United States108. However, poor waste management in China may be a key reason, leading to a considerable amount of plastic waste, especially in the form of widely used packaging plastics47. Compared to the western regions, the eastern regions of China generally have higher population density because of faster social and economic improvement109. This disparity results in more sources and opportunities for plastic production and leakage into the environment in the eastern regions, leading to higher MP abundance. The lower section of Yellow River was the most prosperous region in ancient China, and its catchment goes through rapid urbanization, industrialization and agricultural activities recent decades110,111. Additionally, natural processes in rivers, such as turbulent diffusion and MPs resuspension, can transfer the resuspended MPs from sediment to surface water, contributing to MP abundance in water bodies92,112, these processes of which may be more pronounced in the Yellow River because of high sediment in the river.

The production of the European Union, as well as the United Kingdom, Norway, and Switzerland, only accounted for 15% of the world plastic production in 2021, slightly decreasing 4% compared to 201789. Siegfried et al.113, combining with many MP sources, calculated river export of MPs from the European rivers to coastal seas and indicated the high yields in Israel and Lebanon, Turkey, southern France, western Portugal, northern as well as eastern Spain, and the United Kingdom113. Our findings are consistent with MP pollution in western Asia, as well as Central and eastern Europe. These subtle differences are attributed to a variety of natural factors, such as climate (i.e., rainfall). Yet plastic pollution in Europe and the Middle East should be taken seriously. In 2020, the ratios of recycling in Poland, Slovakia, Latvia, and Hungary were 27%, 30%, 34%, and 22%, respectively, but for landfill, 41%, 45%, 64%, and 57%, respectively89. In contrast, Germany, Denmark, Norway, and the Netherlands had over 40% recycling and less than 2% landfill89. A study mapping the mismanaged plastic waste (MPW) distribution in the Carpathians of East-Central Europe has suggested that most of MPW accumulation hotspots occur along rivers in Romania, Hungary, and Ukraine114. Russian-Ukrainian war may aggravate the MP pollution in rivers from untreated waste115. Additionally, tire and road wear particles constituted the largest source of MPs in European rivers, accounting for 42% of the total measured MP load, while plastic polymer-based textiles ranked second at 29%113. Previous study estimated that Africa contributed 24% of the global emissions of mismanaged plastic waste116. Despite of its relatively low production and consumption of plastics (5% and 4% respectively) compared to that of China, North America, Western Europe, and India, but Africa has the highest global environmental emissions of plastics due to poor waste management, contributing 24% of the global emissions of mismanaged plastic waste116. Many cities in Africa have experienced rapid urban development117. Indeed, the sub-Saharan urban expansion averaged 140% between the 1960s and 1990s, which is ten times faster than in OECD countries and 2.5 times faster than in the rest of the developing world118. This, to a certain extent, accelerates the MP pollution. Besides, it is not overlooked that most African countries have insufficient budgets to plan and execute governmental projects including research activities107, overlooking the MP pollution to some extent.

Human activities in the Arctic are relatively low, greatly reducing the source of MP pollution. The Northern Hemisphere (10°N-50°N) accounts for most of the world’s population and urban development119,120, significantly increasing the potential sources of MPs. Asia notably contributes to MP pollution in the 10°N-50°N latitude region. At the same time, we excluded the desert and glacier areas, including northern Africa, Arabian Peninsula, and Greenland, which may have introduced a considerable error in this range.

We also predicted three global maps of MPs through different sampling methods to assist the global tendency. The trends of grab sampling and net collection are consistent with the map patterns generated by the three methods. However, the pump collection method is more efficient in collecting MPs. Additionally, the samples are small and extremely uneven. Observations from pump sampling are concentrated in Central Asia, while there are no observations in Africa, making the higher trend in Asia and lower in Africa, leading to differences in the maps generated by the three methods.

Limitations and prospects

Despite our considerable effort to obtain data and predict MP abundance in global inland waters, some limitations remain. The research on MP abundance in inland waters in Canada, Australia, South America, and Africa is still insufficient, which may affect the accuracy of our predicting map. For example, Aragaw et al.121 have reported that Africa, as a rapidly developing continent, has consumed a substantial amount of plastic production121, but contributes 24% of the global emissions of mismanaged plastic waste116. However, the studies based on MPs in African freshwater ecosystems are still insufficient. Additionally, our database exhibits a severely right-skewed pattern with some extreme maxima, such as those observed in the Yellow River and small-scale rivers in Japan, potentially leading to lower prediction accuracy for these high-abundance areas.

In our prediction model, data related to WWTP, mismanaged plastic waste, and atmosphere deposition, are not included due to the lack of globally available and published distribution data on these. Therefore, these factors were not incorporated into the study and are only mentioned in the Discussion Section. Additionally, the lack of unified standards for experimental methods, including sampling, purification, extraction, and identification, has impacted the range distribution of MP abundance. Our analysis clearly demonstrated that MP abundance is influenced by the sampling method and tools used, which was consistent with previous studies122,123. Harmonizing these methods is essential for accurately assessing MP pollution globally. Our results suggested that grab sampling with larger volumes (at least 500 liters) and a 60–90 μm mesh size would be recommended, since grab sampling is more common in freshwater samples and comparisons across regions are easier, and pump sampling is too affected by sampling volumes and filter pore size, leading to overestimation.

Recently, mass concentrations of MPs in water have emerged as a reasonable unit, particularly due to the breakdown of microplastics in the environment. For example, Kooi and Koelmans124 and Koelmans et al.125 proposed a method to convert items to mass of MPs based on three key components: (1) a power-law distribution of MP-particle sizes, (2) form-ratios representing the height and width of MP particles, which are expressed as triangular distributions with specified lower and upper ratio-boundaries for forms-various, and (3) triangular distributions of polymer-densities that take into account the observed density ranges of polymers in the environment. However, the research is currently unable to estimate the mass concentration, because of the varying sampling methods employed in our database leading to discrepancies in MP abundance. Additionally, different studies of our database define MP size classifications differently. Estimating global mass concentrations of MPs is indeed a worthwhile direction for future research, which can further promote the knowledge for us to grasp the MPs global distribution.

According to our results, the world should make collective efforts to manage MPs pollution in freshwater ecosystems, particularly in Asia, East and Central Europe, Africa, South America, and the eastern United States. Developing a circular plastic economy (i.e., where plastics are reused, recycled, and redesigned to minimize waste) and stricter waste management are key to mitigating MP pollution. Among them, the primary mitigation strategy is to improve plastic waste management, such as restricting open dumping, improving plastic recycling, raising public awareness and education, implementing strict waste management and waste sorting, and enhancing the management of landfills and incineration plants. Besides, it needs to strengthen the management of urban surface plastic waste to reduce the ratio of MPs entering freshwater bodies through land surface runoff, particularly in the dry season. Given the low flow during the dry season, it is essential to enhance the collection of plastic waste in water bodies and their surrounding areas. Moreover, constructed wetlands in urban could be emphasized to consider the movement of MPs in surface waters by a combination of promoting gravitational settling, microbial adhesion, plant retention, and sediment adsorption56. In terms of cropland, organic (e.g., grass clippings, straw, compost, sawdust, dry leaves, bark and even paper) and inorganic mulches (e.g., consisting of stones and gravel) can be substituted for conventional plastic mulch films126. Additionally, there has also been some progress in experiments related to the removal of MPs from water bodies127. For example, using 1.3 g L−1 of Fe3O4 magnetic nanoparticles and letting it sit for 150 minutes can remove over 80% of MPs128. Additionally, some colloidal silica minerals (e.g., kaolinite, smectite, montmorillonite, or Al/Fe-oxyhydroxides) can adsorb dissolved organic matter via electrostatic/covalent interactions129. Since oxidized or biofilm-coated MPs often have negative charges (-COOH/-OH) like that of organic matter, this may suggest that these colloidal silica minerals may act as similar adsorption processes in aiding MP removal from aquatic systems. Yet this still needs to be further validated in future.

Conclusion

By synthesizing global datasets, this study was to evaluate the differences in MP abundance among different ecosystem types, seasons, and sampling processes, revealing the characteristic distribution of MPs to fill the gap in the spatial distribution of MP abundance in global freshwater bodies. Our findings demonstrate that MP abundance ranges over six orders of magnitude (0.00 to 4,275,800.70 items m−3), with small-sized MPs (<1 mm) dominating freshwater systems. This is a previously underquantified threat that demands urgent attention due to its potential to disrupt aquatic ecosystems.

In contrast to the initial hypotheses (1), the MP abundance in inland waters during the dry season was significantly higher than that in the wet season. This variation is attributed to the alteration of hydrodynamic factors (such as rainfall duration and intensity), land-use types, and sampling time. This underscores the need for time-sensitive monitoring strategies, particularly in regions with pronounced climatic variability. Methodologically, we identify that standardized grab sampling (≥500 liters, 60–90 μm mesh size) is essential for accurate cross-regional comparisons, because of pump-based methods risk overestimation induced by the sampling volumes and filter pore size dependencies. As expected, our analytical results highlight that the human development index plays a crucial role in the distribution of MPs in global inland waters, being consistent with our hypothesis (2). Despite that HDI does not directly cause plastic waste or emissions, it serves as a proxy for development patterns that drive these environmental impacts. The distribution of MPs in inland waters is further regulated by climate factors such as temperature, precipitation, evaporation, and land surface runoff. In agreement with our hypothesis (3), Asia, Europe, Africa, and eastern United States face the highest burdens of MPs in inland waters, particularly China, exemplifying the nexus of high plastic consumption and inadequate waste management.

Indeed, human activities and social development have brought about the varying degrees of MPs pollution. To manage plastic waste, some effective measures should be implemented to reduce emissions into the environment. Our research highlights the importance of enhancing waste management policies and public awareness and improving plastic waste management in urban areas and freshwater ecosystems, particularly during the dry season, as well as addressing agricultural MP emissions (such as by replacing plastic films with organic or inorganic mulches). These measures are crucial for reducing environmental emissions and mitigating MPs pollution in inland waters. This finding will serve as a valuable resource for further studies and the development of effective management measures for inland waters.

Methods

Data source and processing

We conducted a thorough search for peer-reviewed literature on microplastics (MPs) in inland waters using the Web of Science and China National Knowledge Infrastructure (CNKI) in April 2024. The search theme included terms such as “microplastic,” “microplastics,” and “MPs,” combined with keywords related to inland waters (“watercourse,” “stream,” “river,” “creek,” “lake,” “reservoir,” “dam,” “inland water,” “watershed,” and “freshwater”). We applied the following criteria to filter the articles: (1) The study was conducted on inland waters; (2) At least one record of MPs abundance was reported; (3) The unit of MPs abundance in water was measured in items m−3 or items L−1; and (4) Data from wastewater treatment plants (WWTP) were excluded.

Our database included a total of 5365 observations from 301 articles (Fig. 1a; Appendix 1). In addition to MP abundance, we also collected the following information: (1) Longitude and latitude of sampling sites, article publication year, sampling year, sampling season, sampling volume, and ecosystem type; (2) Methods used for obtaining MPs, including sampling, extraction, and purification methods (Note S1); (3) Characteristics of MPs, including particle colors, shapes, sizes, and polymers; and (4) The abundance unit of MPs was uniformly converted to items·m−3. We hypothesized that (1) the abundance of MPs in inland waters during the wet season was higher than in the dry season, due to the input of MPs from rainfall through land surface runoff; (2) human development index (HDI) is the most important drivers to promote MP emissions of inland waters; and (3) MP abundance would show a clear global patterns across inland waters, which was higher in areas with higher human activities, especially in Asia. Our study consists of two modeling parts: assessing the differences in microplastics at different taxonomic levels and global predictions. The overall process is shown in the supplementary file (Fig. S1).

Modeling of assessing the differences within the different division

Prior to analysis, we conducted a log transformation (log (abundance+1)) on our data to obtain normal distribution data (Fig. S2). First, we used univariate linear mixed models in the lme4 package130 to assess the effects of ecosystems, different seasons, experimental settings on MP abundance in global inland waters, and also explore the differences within the different characteristics of MPs. Next, we will conduct global predictions, and the specific process is as follows.

Global predicted variables

Based on our database coordinate information, we extracted 23 candidate predictor variables from the HydroSHEDS data131 (https://www.hydrosheds.org) (Table S1) to establish and predict MP abundance in global inland water environments using machine learning. This study resampled all variables in EPSG:4326 (World Geodetic System 1984) at 0.05-degree grid resolution. We also conducted log(abundance+1) transformation for the response variable to facilitate modeling in the predicting map. To avoid collinearity, we eliminated predictors with low contributions one by one until the variance inflation factors (VIFs) of all predictors were less than five. After that, we conducted the VSURF package for model selection132, which used the random forest permutation-based score of importance and conducted using a stepwise approach, adding variables only if they significantly reduced the out-of-bag error (OOB) by more than the average change after adding the noise variables. Finally, there are the remaining 5 variables used for modeling (Fig. S5).

Modeling of prediction

Based on our database, we conducted random forest regression model based on randomForest package133 to explore the relationship between environmental predictors and MP abundance. As a supervised learning algorithm, random forest combines multiple decision trees to make predictions, reducing the risks of overfitting and improving the generalization performance. To determine the best-performing model, we tuned a series of parameters to optimize model performance134, using the train function from the caret package135. We fulfilled repeated random 10-fold repeated cross-validation. The process was trained 10 times, each time using 90% of the data for training and leaving the remaining 10% for validation. Finally, we select the optimal parameters to build the final model.

Model test and predictors importance assessment

In each model, we evaluated the root-mean-square error (RMSE) and R2_Accuracy to summarize the cross-validation results and assess the predictive performance of the model. R2_Accuracy (R2_Accuracy = 1 – SSE/SST) represents the amount of variation explained by the model136, where SSE is the squared error between observation and prediction, and SST is the total sum of squares137. The model with the lowest RMSE and highest R2_Accuracy was considered the best performance for predicting MP abundance in specific environments. We assessed the variable importance using “caret” package for the best parameter random forest model, and the results of variable importance with normalization within a range of 0–100% in the model indicated the effects of predictors on the model results.

Global prediction and uncertainty

We extracted data on these five predictor variables for the global scale, excluding glaciers and desert regions. Based on this data, the model with best parameter was conducted to predict and generate the global map of MPs, the process of which were performed 100 iterations. Notably, we generated 100 global maps related to MPs, and by averaging the results over these 100 iterations, we obtained the per-pixel mean values and standard deviations. Afterwards, we calculated the coefficient of variation (standard deviation divided by mean value).

Analysis of differences MP abundance among sampling methods

In terms of MPs in water, sampling methods largely affect the abundance values of MPs in water samples, which can be easily overestimated or underestimated. We divided all the methods into three categories (grab, net, and pump), and modeled as well as predicted world maps for the three methods with the same processes as above. Additionally, owing to major sampling points in China, we also modeled and predicted these properties for China using the same processes.