Drivers of rising monthly water temperature in river estuaries

River estuaries are habitats for a variety of organisms, including many temperature‐sensitive species; water temperatures in estuaries are affected by several factors as they are influenced by both terrestrial and marine environments. Therefore, understanding the factors that influence river estuaries are essential for environmental management. However, no information exists on temperature change at high temporal resolution and over a wide area; moreover, the relationship with environmental factors has not yet been clarified. Here, we have described the actual status of water temperature change in riverine estuaries and its relationship with environmental factors for a wide area of the Japanese archipelago. Our results indicated that 217 of 294 rivers showed a significant monthly increase in water temperature. The average annual change in rates of water temperature increase was the highest in October at 0.090°C yr−1 and lowest in February at 0.068°C yr−1. Furthermore, snowfall, air temperature increase, sea water temperature, and land use were identified as factors increasing the rate of water temperature increase. Factors influencing the increase in river estuarine water temperature varied from month to month—with meteorological factors being the strongest influencers from spring to fall, and anthropogenic factors in winter. Our findings emphasize the importance of considering not only meteorological and oceanographic phenomena, but also anthropogenic influences and topographic features to understand the pattern of water temperature changes in river estuaries.

Estuaries are essential for ecosystem services and biodiversity; however, like other biomes, they are strongly affected by climate change.Climate change impacts on the estuary include sea level rise (Cayan et al. 2008), changes in freshwater inflow (Yang et al. 2015), changes in sediment dynamics (Wang et al. 2010), changes in nutrient inflow (Statham 2012), and changes in water temperature (Oczkowski et al. 2015;Suzzi et al. 2022).These phenomena cause a change in water quality, including salinity (Levinton et al. 2011;Wan et al. 2022), topographic changes (Ganju and Schoellhamer 2010), and changes in food resources (Colombano et al. 2021), leading to major changes in the ecosystem.
Among the impacts of climate change, changes in coastal and estuarine water temperatures will alter biota, including an increase in thermotolerant species, with significant consequences for ecosystems (Madeira et al. 2012;Ohtsuki et al. 2023).Changes in biota associated with coastal climate change have been reported worldwide, including the Brazilian coast (Araújo et al. 2018), Gulf of Mexico (Gericke et al. 2013), English and Bristol Channel (Henderson 2007) and elsewhere in the world.These changes have also been reported in estuarine areas, although to a lesser extent than in coastal areas.For example, as per an analysis of finfish abundance in Long Island Sound based on seasonal trawl survey data, the annual abundance of cold-adapted species was negatively correlated with mean bottom water temperature, while warm-adapted species were significantly positively correlated (Howell and Auster 2012).In South African estuaries, the abundance of the dominant three groups of mugilids (tropical, warm-water, and cool-water) is associated with sea surface coastal temperatures and have been identified as key species for climate change impact assessments (James et al. 2016).A comparison of fish species distributions in the Atlantic European seaboard between the 1970s and present shows a northward shift of populations in 11 species, suggesting that many estuarine fish species have migrated northward over the past 30 yr as water temperatures have increased (Nicolas et al. 2011).
Estuarine areas, including river estuaries, are habitats for a variety of species; however, changes in biodiversity have been observed in many of these areas due to changes in water temperature, caused by climate change and other factors (Howell and Auster 2012;Gericke et al. 2013;Araújo et al. 2018;Itsukushima, 2023).Some species dwelling in river estuaries are sensitive to temperature changes and thus have limited acclimation to increased water temperatures (Hoegh-Guldberg et al. 2007;Brown et al. 2016).Hence, there is worrying concern that their populations might decline with increasing changes in water temperature.In addition, many species have been reported to temporarily use estuarine areas as part of their life cycle, or only at certain times of the year (Elliott 2007;Tewksbury et al. 2008).However, most studies on river temperatures, including those in river estuaries, report annual averages and do not account for the time resolution necessary for the survival of species sensitive to temperature changes or for the conservation and restoration of species that use the river estuaries only during certain times of the year.Therefore, identifying the factors that affect water temperature and extent of monthly temperature increase will greatly contribute to the environmental management of river estuaries, as this information can help to determine which species are at risk at which time of the year and which influencing factors need to be mitigated to manage the increasing water temperatures in river estuaries.
Although changes in water temperature are extremely important for estuarine ecosystems, information on the response of estuaries to climate change is lacking (Suzzi et al. 2022).A case study of estuarine water temperature changes found that the annual average water temperature in estuaries in the northeastern United States has increased by $ 1.2 C over the past 50 yr, indicating that shallow intertidal areas are important sites for studying climate change impact (Oczkowski et al. 2015).Furthermore, the frequency of lethal temperatures for Chinook salmon (> 16 C) is increasing in the Sacramento River estuary due to increased temperatures and decreased snowmelt runoff (Cloern et al. 2011).The East Australian coastal estuaries are rising in temperature faster than predicted by global oceanatmosphere models (Collins et al. 2013), with a reported rate of change of 0.2 C yr À1 (Scanes et al. 2020).
Among the several factors that influence water temperature in rivers, atmospheric conditions are considered to be the most significant contributors in the heat exchange processes occurring at the water surface (Livingstone and Lotter 1998;Erickson and Stefan 2000).The degree of heat exchange in water with the riverbed and river discharge are also known factors that influence water temperature (Sinokrot and Stefan 1994).In addition, river water temperatures are affected by various anthropogenic factors (Caissie 2006); major anthropogenic factors include climate change (Liu et al. 2020), land use changes such as urbanization (Kaushal et al. 2010), heating runoff from power plants (Webb et al. 2008), and sewage effluent from cities (Kinouchi et al. 2007).Owing to the mixing of land and sea waters in river estuaries, there is considerable influence of sea water temperature on estuary water temperatures (Suzzi et al. 2022).Therefore, in order to monitor water temperature in river estuaries, it is important to clarify, which factors have a significant influence on each target site.
Although the decrease in freshwater inflow and the increase in seawater temperature are considered the leading causes of estuarine water temperature change, the factors that influence estuarine water temperature change are extremely complex, including anthropogenic impacts, such as land use change and dam construction; natural factors, such as topographical factors; and ocean currents and tides, in addition to climate change.However, only few cases have examined these factors.No studies have discussed the relationship between various environmental factors based on long-term observational data on the impacts of climate change in river estuaries.Although the relationship between watershed-specific environmental factors and water temperature is dynamic and suggests the importance of studies with diverse temporal resolutions (Lisi et al. 2015), no examples of monthly relationships exist for many river estuaries.This is the first study that attempts to elucidate the complex factors influencing water temperature changes in river estuaries by using monthly nationwide water temperature data in the Japanese archipelago, which traverses multiple climatic zones and is influenced by various ocean currents, including both cold and warm currents.

Water temperature data
We used water temperature data from 294 river estuaries in the Japanese archipelago that have been observed at least once a month for the past 30 yr or more.Four major ocean currents in the seas around the Japanese archipelago play important roles in establishing the water temperature diversity in river estuaries; these are the Kuroshio Current, which enters the Pacific Ocean from the East China Sea and flows northward along the Japanese archipelago coast; the Tsushima Current, which comprises of water from the Kuroshio Current and coastal waters from the East China Sea that flow together into the Sea of Japan via the Tsushima Strait; the Kuril Current, which flows from the North Pacific Ocean and the Sea of Okhotsk; and the Liman Current, a cold current that flows southward from near the Mamiya Strait, along the Eurasian continent, to the Sea of Japan (Kume et al. 2021) (Fig. 1A).Years with a missing period of even a month were regarded as missing years.To avoid the effects of missing measurements, river estuaries with missing years of < 10% of the total observation period were considered.The total years of recording  was 11,866 (average 40.4 yr), and the number of missing years was 2.1% of the total.The data sources were the Water Information System managed by the Ministry of Land, Infrastructure, Transport and Tourism (http://www1.river.go.jp/), the comprehensive site for water environment information managed by the Ministry of the Environment (https://water-pub.env.go.jp/water-pub/mizu-site/mizu/kousui/dataMap.asp), and the water quality measurements for public water bodies in each prefecture.

Environmental data
As environmental factors, 24 indicators were selected for meteorological, oceanic, anthropogenic, and geomorphologic factors.The meteorological factors were calculated using data from 252 stations with observation records of river estuarine temperatures for similar time periods.Among the environmental factors, the rate of air temperature change was calculated, as it is considered the most important influencing factor for river temperature change (Livingstone and Lotter 1998;Erickson and Stefan 2000), and this data were available for the time period corresponding to the years of observation of river estuarine water temperature (REWT).For the other factors, average values within the observation period were used.The indicators were the average of each of the following years.
(1) annual precipitation, (2) annual average temperature, (3) annual maximum temperature, (4) annual minimum temperature, (5) annual sunshine hours, and (6) annual snow depth.The time-series trend variation of annual average temperature was tested using the Mann-Kendall method, and the slope of temperature change ( C yr À1 ) was calculated as (7) air temperature change; air temperature change for meteorological stations where no significant trend variation (p < 0.05) was identified was set to 0. To calculate the basin average of the target river estuaries for the above indicators, the dominant area of each station was geometrically obtained by Voronoi partitioning, and the indicator values of each station were weighted and averaged by the area of the dominant area.Meteorological data were from the Japan Meteorological Agency (https://www.data.jma.go.jp).
As the sea area influences REWT, three indices for seawater temperatures were selected: (8) annual average sea water temperature, (9) annual maximum sea water temperature, and (10) annual minimum sea water temperature.Sea surface water temperature data were obtained from the Japan Meteorological Agency's Sea Surface Temperature Information for Coastal Areas of Japan (https://www.data.jma.go.jp/gmd/ kaiyou/data/db/kaikyo/series/engan/eg_areano.html) and from water quality measurements in public waters of the Ministry of the Environment.Of the 66 sea surface temperature data points organized based on these data sources, the sea surface temperature data in the nearest vicinity of the targeted river estuaries was used.In addition, because the closure of a bay is thought to affect the exchange of seawater, which in turn affects the temperature of river brackish water, (11) the degree of closure of the bay was calculated based on the following equation (https://www.emecs.or.jp/): W : width of the entrance to the bay， S : area of the bay， D1 : depth of the deepest part of the bay，and D2 : deepest depth at the entrance to the bay.
The value of ( 11) for the degree of closure of the bay for rivers flowing directly into the open ocean was set to 0. The topographic information needed to calculate (11) the degree of closure of the bay was obtained from GSIMAP (https:// maps.gsi.go.jp/) of the Geospatial Information Authority of Japan.The values of D1 and D2 were obtained from the website (https://mar-nets.com/contents/#contents1), which is based on bathymetric maps created by the Japan Coast Guard.
Indicators were calculated for anthropogenic disturbance, focusing on land use, population, and dams.The percentage of land use in each watershed was calculated for (12) forests, (13) agriculture, and ( 14) urban area.Data on land use were obtained from the 1 km mesh land use data of 2016, recorded in the digital national land information (https://nlftp.mlit.go.jp/index.html).Next, (15) population in the watershed and ( 16) population density in the watershed were calculated using the 2010 population statistics.Data were obtained from the 500 m mesh demographics in digital national land information (https://nlftp.mlit.go.jp/index.html).Information on dams with a bank height of 15 m or more was organized based on the World Conference on Dams definition (International Commission on Large Dams, 1998); (17) the number of dams in the watershed and (18) the percentage of the dam catchment area to the watershed area were used as indicators.Information on dam location and catchment area was collected from the Dam Handbook (http://damnet.or.jp/Dambinran/ binran/TopIndex.html).
The geomorphological factors are as follows: (19) watershed area (km 2 ), (20) watershed average elevation, (21) relief amount, ( 22) average inclination angle, (23) percentage of flat field, and (24) percentage of steep slopes (slope is 15 or higher) of the target river.Elevation and inclination angle data were obtained from the slope fifth order mesh (250 m mesh) in digital national land information; (20) watershed average elevation was the average value of 250 m elevation mesh data; (21) Relief amount was calculated by the difference between the maximum and minimum elevation in the watershed; (22) average inclination angle was the basin average of the average slope of the 250 m mesh data; (23) percentage of flat field was the percentage of meshes in the watershed with an average slope of 0-5 in the 250 m mesh data; and, (24) percentage of steep slopes was the percentage of meshes in the watershed with an average slope angle of 15 or greater in the 250 m mesh data.
The above 24 environmental factors showed variation among the target watersheds and a high correlation among multiple factors (see Supporting Information Fig. S1).In the statistical analysis, variables were removed for correlation coefficients with absolute values greater than 0.7, and finally 15 environmental factors were used in the statistical analysis; the following factors were excluded from the analysis: (2) annual average temperature, (5) annual sunshine hours, (8) annual average sea water temperature, (12) forest, ( 16) population density in the watershed, ( 17) the number of dams in the watershed, (20) watershed average elevation, (23) percentage of flat field, and (24) percentage of steep slopes.

Statistical analysis
The Mann-Kendall method was used to test the time-series trend of water temperature fluctuations in the 294 targeted river estuaries.The Mann-Kendall method was used to test 15 items: the average water temperature for each month from January to December, the annual average water temperature (Tavg), the annual maximum water temperature (Tmax), and the annual minimum water temperature (Tmin).Based on the Mann-Kendall test, REWTs judged to be significant (p < 0.05) were considered to have trend changes, and the slope of the linear approximation was calculated as annual change rates in water temperature ( C yr À1 ).
To identify environmental factors that influence the trend of water temperature change, a random forest was conducted with the trend of water temperature change (increase/no change/decrease) of the 294 river estuaries based on timeseries trend variation analysis as the objective variable and environmental factors as explanatory variables.First, feature selection of the Boruta algorithm was applied to select critical environmental factors for each objective variable (Kursa and Rudnicki 2010).Random forests were conducted on the selected environmental factors.Variable importance of environmental factors was assessed by mean decrease gini (Han et al. 2016), and out-of-bag (OOB) error was used to evaluate the goodness of fit of the model (Mitchell 2011).
In addition, we conducted a non-metric multidimensional scaling analysis (nMDS) to summarize the composition of occurrence trend of the month of temperature increase in each river estuary using the Bray-Curtis similarity.River estuaries that showed an increase in water temperature in any month of the year and had no months with a decreasing trend in water temperature were included in the analysis.In addition, environmental factors with a significant effect (p < 0.05) on the NMDS axis were identified by the Monte Carlo permutation test.To investigate the statistical significance in the environmental factors among groups classified by NMDS, multiple comparisons were conducted.First, Bartlett's test was performed to test the equal variances, followed by the Kruskal-Wallis test and one-way ANOVA for unequal and equal variances, respectively.When the differences between groups were significant based on the Kruskal-Wallis test, a Steel-Dwass test was performed for multiple comparisons.When the differences between groups were significant based on the one-way ANOVA, Tukey's HSD test was performed for multiple comparisons.The significance level was set to p < 0.05.
Finally, a generalized linear model (GLM) analysis was conducted using the annual change rates in water temperature of Tavg, Tmax, and Tmin per month as objective variables and environmental factors as explanatory variables.GLM is an extended model of the linear model, which allows the incorporation of non-normal distributions of the response variables and transformations of the dependent variables linearly (McCullagh and Nelder 1989).We compared the Akaike information criteria (AIC) of each model obtained from the method by increasing and decreasing the variables.Finally, we adopted the lowest AIC model as the best model for each species.GLM was conducted with the MASS (version 7.3-50) package.All analyses were performed using R (version 4.0.4)software.

Time-series trends of REWT fluctuations
Mann-Kendal test for water temperature changes for each month in the 294 river estuaries in the Japanese archipelago (Fig. 1A) showed that 217 of the 294 sites had a significant increase in water temperature in any month.The largest number of rivers showed a rise in water temperature in October, with 120 rivers showing an increasing trend.However, in September, the number of rivers showing a growing trend in water temperature was small, and the number of rivers showing an increasing trend was 38.The average annual water temperature (Tavg) in 154 rivers, the annual maximum water temperature (Tmax) in 100 rivers, and the annual minimum water temperature (Tmin) in 75 rivers were significantly increased (Fig. 1B).The annual change rates in water temperature were 0.045, 0.067, and 0.063 C yr À1 for the Tavg, Tmax, and Tmin, respectively (Fig. 1C).The average annual change rates in water temperature were found to be the highest in October at 0.090 C yr À1 and lowest in February at 0.068 C yr À1 (Fig. 1D).Of the 294 rivers, 43 rivers had no months in which a change in water temperature was observed.In 34 rivers, there was no month in which an increasing trend in water temperature was observed, and there were months in which only a cooling trend was observed.

Temperature fluctuations in river estuarine water
The results of a random forest with the categories of the fluctuation trend of REWT as objective variables (increasing trend/ no change/decreasing trend) and 16 environmental factors as explanatory variables showed an OOB error of 21.7 AE 2.8% (average AE standard deviation) for the case covering each month.The predictive accuracy was higher in the winter season, with the lowest OOB being 17.9% in January.In contrast, the OOB for Tavg, Tmax, and Tmin were 34.0%, 25.5%, and 26.0%, respectively.The most frequently implicated environmental

High
Importance Low    , Feb, Mar, and Apr), indicators related to the anthropogenic influence of (13) agriculture, ( 14) urban area, and (15) population in the watershed were identified as the main factors explaining the trend in water temperature fluctuations.In addition, meteorological and oceanic factors were attributed for fluctuations in water temperature from May to August (Fig. 2).

Classification of REWT by monthly change trend
As the trend of monthly water temperature varied between sites, classifications were developed and relationships with environmental factors were examined.Non-hierarchical clustering classified only the 198 rivers with an increasing trend in water temperature into four groups (Fig. 3).As a characteristic, Group A, placed around the third quadrant of the nMDS plane, tended to increase the water temperature in May and June; Group B, placed around the fourth quadrant, tended to increase the water temperature in winter; Group C, placed around the first quadrant, tended to increase the water temperature in the fall season; Group D, placed around the second quadrant, showed a tendency to increase the water temperature in the summer season (Fig. 4A-D).
Annual change rates in ( 7) air temperature change and (10) annual minimum sea water temperature (positive on the first axis), ( 13) agriculture and ( 4) annual minimum temperature (positive on the second axis), and (6) annual snow depth (negative on the first axis) were identified by nMDS as environmental factors strongly related to the ordination of water temperature change trends (Fig. 3).The test of differences in the average of the environmental factors for each group showed significant differences ( p < 0.05) in five aspects: (6) annual snow depth, (7) air temperature change, (10) annual minimum sea water temperature, (13) agriculture, and (15) population in the watershed.Group B, which showed a trend of increasing water temperature in winter, had a significantly larger (15) population in the watershed than the other groups.Group C, with a trend of increasing water temperature in the fall, showed a trend of higher (13) agriculture, (10) annual minimum sea water temperature, and (7) air temperature change.In group D, where the water temperature tended to increase during the summer season, a trend toward greater (6) annual snow depth was observed compared to the other groups (Fig. 5A-E).

Annual change rate in REWT and environmental factors
The results of the GLM with the annual increasing rates in water temperature for each month, Tavg, Tmax, and Tmin as objective variables and environmental factors as explanatory variables showed that for meteorological and oceanic factors, (7) air temperature change, (6) annual snow depth, and (9) annual maximum sea water temperature were selected in many months as the factors that increased the annual increasing in water temperature.However, (1) annual precipitation was identified as a factor negatively impacting the annual increasing rates in water temperature.Anthropogenic factors, (13) agriculture and ( 14) urban area, were identified as factors increasing the annual increasing rates in water temperature.However, ( 14) urban area negatively affected the (9) annual maximum sea water temperature and August water temperature.Among topographic factor, (21) relief amount negatively affected the annual increasing rates in water temperature.The (18) percentage of dam catchment area to the watershed area was not an influencing factor (Table 1).

Discussion
Water temperature change trends in river estuaries Estuarine temperatures have risen in the northeastern United States by 0.024 C yr À1 over 50 years (Oczkowski et al. 2015), and in Australia's wave-dominated back-barrier coastal estuarine lake, they have increased by approximately the same rate as ocean warming (+ 0.015 C yr À1 ) from 1970 to 2010 (Suzzi et al. 2022).In this study, covering about 40 yr of observation records, an annual average water temperature (Tavg) of 0.045 C yr À1 and monthly trends of increasing water temperature were observed, with the highest change in October (0.090 C yr À1 ) and lowest change in February (0.068 C yr À1 ).The annual average water temperatures in estuaries in the northeastern United States have been reported to increase from winter to early spring (Suzzi et al. 2022).This analysis shows that the greatest number of REWT increases were reported in October, suggesting that the dominant factors influencing the formation of increasing water temperatures are regionally different.
Random forest results showed that the rate of misinterpretation of categories of the fluctuation trend of REWT was lower at each monthly water temperature than the annual average.The importance of environmental factors affecting the fluctuation trends varied from month to month (Fig. 2).The results suggest that temporal resolution is essential for accounting for changes in water temperature.This study is the first attempt to address water temperature changes and complex environmental factors in river estuaries, focusing on seasonality.The importance of considering seasonality in the analyses of river water temperature has been emphasized (Mantua et al. 2010).Furthermore, since many species inhabiting estuaries change their lifestyle depending on water temperature, knowledge of water temperature changes with high temporal resolution is vital for conserving and restoring estuarine ecosystems.For example, the salmonid Salmo trutta, an important fishery resource, is known to migrate upstream and spawn in response to changes in water temperature (Jonsson and Jonsson 2009;Haraldstad et al. 2017).In addition, many individuals of the endangered Plecoglossus altivelis ryukyuensis that hatch in estuaries at temperatures above 20 C do not return home, and the longer the temperature above 20 C is experienced at the juvenile stage, the lower the growth rate (Kishino et al. 2008).For the conservation and restoration of species with a narrow range of permissible water temperatures at specific stages of life, elucidating the relationship between the various time-resolved information on water temperature changes in river estuaries and complex environmental factors is essential.

Meteorological and oceanic factors
Atmospheric conditions are among the most important factors in river water temperature because they are primarily involved in the water surface heat exchange process (Caissie 2006).This study revealed that among the meteorological and oceanic factors, (6) annual snow depth, (7) air temperature change, and (9) annual maximum sea water temperature had positive effects on the annual change rates in REWT (Table 1).The GLM results also showed that (7) air temperature change had a positive influence on REWT in several months, suggesting that climate change has a greater influence on the overall increase in REWT.
In addition, (9) annual maximum sea water temperature and (10) annual minimum sea water temperature were identified as factors influencing the increase in REWT during the summer and spring, respectively (Fig. 2).Additionally, (9) annual maximum sea water temperature was observed to increase the rate of REWT in winter (Table 1).In areas with higher maximum water temperatures, the summer months are accompanied by formation of more water vapor, which increases summer runoff and relatively lowers winter runoff, thereby influencing the increase in REWT in winters.Decreases in winter precipitation has been reported on the Pacific Ocean side of eastern Japan and the Sea of Japan side of western Japan, where warm currents flow (Japan Meteorological Agency 2017); therefore, changes in the annual water circulation system associated with changes in sea water temperature may have an impact on the increased rate of REWT.In the Japanese archipelago targeted in this study, climate change has increased the flow of the warm Tsushima Current (Kida et al. 2020) and changed the flow regime of the Kuroshio Current (Sakamoto et al. 2005).More future changes in heat transport mechanisms are assumed to occur in the Japanese islands (Nishikawa et al. 2020).Furthermore, water temperatures in brackish lakes in Australia are rising at the same rate as ocean temperatures (Suzzi et al. 2022).Since the influence of the ocean on the formation of REWTs is significant, it is necessary to accumulate and monitor the effects of ocean currents and changes in seawater temperatures in the future.
The nMDS showed that (6) annual snow depth tended to be larger in streams with a greater tendency to increase water temperatures during summer (Fig. 3).The GLM analysis also revealed that snowfall increases the water temperature rate in July and September (Table 1).In general, snowmelt runoff causes less water temperature change in snowfall-dominated rivers (Mohseni and Stefan 1999;Lisi et al. 2015).However, this analysis showed that snowfall tended to increase water temperature during summers.This may be attributed to the fact that, in rivers where snowmelt runoff is dominant, the summer runoff tends to be lower than other areas and the water temperature tends to increase during the summers.In particular, in the heavy snowfall areas of the Japanese archipelago, a large difference between cold and warm temperatures exists, and summer maximum temperatures almost always exceed 30 C. The combination of lower summer flow rates and higher temperatures is thought to have contributed to the increased summer water temperatures.In the future, climate change is expected to lead to earlier onset of snowmelt and reduced winter snowfall, which will have significant responses to changes in stream temperature (Diffenbaugh et al. 2013;Berghuijs et al. 2014).

Anthropogenic factors
As for the anthropogenic impacts, ( 14) urban area and (15) population in the watershed contribute significantly to the degree of increase in REWT (Fig. 2, Table 1).Many studies have confirmed that urbanization is the driving force behind the rise in river water temperature (Cao et al. 2016).Factors contributing to urbanization-dependent increases in stream water temperatures include loss of riparian vegetation and reduced shading (Dugdale et al. 2018), heat addition from wastewater discharges (Kinouchi et al. 2007), and urban runoff heated by pavement (Li et al. 2013).In addition, estuarine areas are sites of high anthropogenic impact because of the presence of considerable ecosystem services and ancient human use (Costanza et al. 1997).Hence, many cities are located near estuaries (Edgar et al. 2000), which in turn have significant consequences on REWT.In contrast, GLM results showed that urbanization negatively impacts the water temperature increase rate of Tmax and T8 (Table 1).One possible reason for this is that in rivers flowing through overcrowded cities, water temperatures rise significantly during the summer months, exceeding the treated sewage water temperature.Such inversions between treated sewage and river water temperatures in urban rivers during summer have been reported in large cities (Miyamoto et al. 2006;Han et al. 2021), and the selection of ( 14) urban area as a negative factor for the degree of water temperature increase in August and Tmax in the GLM results may be a reflection of the above phenomenon.
In addition to urban areas, agricultural land is also a factor that accelerates the rate of increase in REWT (Table 1).Agricultural lands were identified as a factor in increasing water temperature in April and May, when paddy fields are puddling and transplanting in many parts of the Japanese archipelago (Wakai et al. 2005;Kitamura et al. 2011).The water temperature is thought to increase due to the decrease in river flow caused by the withdrawal of water for puddling and transplanting and the discharge of water heated by soil heat from the rice paddies.Knowledge of the impact of agricultural land on river temperatures is limited compared to urbanization; however, the possibility of agricultural land increasing water temperatures has been pointed out (Fierro et al. 2017).To evaluate the impact of farming on river water temperature, detailed data, including regional variances in water withdrawal timing and soil temperature, must be collected.
Dams strongly impact water temperature in downstream rivers (Kędra and Wiejaczka 2017;Tao et al. 2020) and the correlation between air and water temperatures is low because of their limited interaction (Erickson and Stefan 2000).However, our results suggest that dams do not significantly affect REWT changes.This may be due to the fact that the river estuary is located at the end of the watershed, at a distance from the dam catchment, and therefore is not as dominant as other factors.

Geomorphologic factors
Geomorphologic factors of a watershed have been noted as factors controlling the relationship between air temperature and river water temperature (Lisi et al. 2015).In this of river estuaries, several geomorphologic factors were also identified to influence water temperature.The (19) watershed area is an important factor that explains the water temperature variability trends (Fig. 2).River estuaries with larger watershed areas have larger alluvial areas and lower flow velocities in the downstream area, which may be attributed to the fact that water temperature is more likely to increase due to sunlight.
The (21) relief amount was a factor that negatively affected the rate of increase in river estuarine temperature for many months (Table 1).Although it has been shown in glaciercovered watersheds, the mean elevation of the watershed is a factor that decreases water temperature (Fellman et al. 2014), a similar trend was observed in the analysis for the Japanese archipelago, where glaciers are virtually non-existent.This may depend on higher elevations with lower temperatures and rainfall temperatures (Lu et al. 2022).Another important finding is that the relief amount in the watershed also affects the water temperature in river estuaries at a distance from the headwaters.
In addition, the topography of the estuary is a factor that influences the exchange and retention of water with seawater and therefore affects water temperature.The (11) degree of closure of the bay was identified as a factor that decreased the rate of water temperature increase in July.Scanes et al. (2020) analyzed the relationship between water temperature and topography in Australian estuaries and highlighted that estuarine water depth and flushing times control water temperature and showed that lagoons and river estuaries warm at the fastest rate because of their shallow mean water depth and limited interaction with the ocean (International Commission on Large Dams 1998).To construct a more accurate model of water temperature changes in river estuaries, including topographic changes, such as the period of estuary blockage may be necessary.
The results of this study indicate that REWTs are influenced not only by climate change but also by various anthropogenic influences.In addition, basin-specific topographic factors influence the formation of REWTs.As the effects of climate change are intricately related to other anthropogenic impacts and are likely to accelerate the increase in REWT, predicting changes in water temperature according to the characteristics of the basin and the associated changes in the environment and ecosystem and considering appropriate adaptation measures is necessary.

Seasonal changes in factors influencing REWT
Among the environmental factors that drive REWT changes, those that directly increase REWT through heat exchange and other processes are called meteorological factors.Although the factors that influence river water temperature are diverse, the influence of heat exchange processes occurring at the water surface is known to be significant (Livingstone and Lotter 1998;Erickson and Stefan 2000).On the contrary, factors that indirectly increase REWT are called anthropogenic influences, which include changes in the water cycle system and external water supplies, such as wastewater from sewage systems.Marine influences can be considered as both direct and indirect factors.In river estuaries, heating and cooling occur via direct contact with sea water, and changes in land temperature due to sea water temperatures can indirectly affect REWT.Geomorphic factors are different from these abovementioned factors and can be considered potential determinants of REWT.
This variation in the mechanism by which different factors affect changes in REWT is assumed to be responsible for the month-to-month fluctuations in the factors' effects on REWT.For instance, climatic factors that have a direct influence on REWT are not regarded as important factors for river water temperature increase in winters (November-February) (Fig. 2).Moreover, in winter, cooler temperatures, less sunlight, and less rainfall have considerably less effect on raising REWT.In contrast, temperature of treated wastewater discharged from cities is not expected to fluctuate significantly in either summer or winter; therefore, the influence of cities is expected to be relatively more in winters.

Conclusion
We describe the actual status of water temperature change in riverine estuaries and its relationship with environmental factors nationwide.As per the results, 217 of 294 rivers showed a significant monthly increase in water temperature.
Our analysis showed significant monthly increases in water temperature in 217 of the 294 rivers.The highest average rate of water temperature change in the target rivers was 0.090 C yr À1 in October, and the lowest average rate was 0.068 C yr À1 in February.As factors that increase the rate of increase in REWT, the (6) annual snow depth, (7) air temperature change, and (9) annual maximum sea water temperature were selected as meteorological and oceanic factors, while the land use of (13) agriculture and ( 14) urban area was selected as an anthropogenic factor.However, (1) annual precipitation and (21) relief amount were found to be factors that had a negative influence on the rate of increase in REWT.However, the degree of influence of these factors varied significantly from month to month, indicating that annual averages are not sufficient to understand changes in REWT.Our results also reveal the importance of considering not only meteorological and oceanographic phenomena, including climate change, but also anthropogenic influences and geomorphological features in order to understand changes in REWT.Future studies at various temporal and spatial scales will enable us to predict future REWT and the resulting changes in the ecosystem, thereby making progress toward adaptation in river estuaries.

Fig. 1 .
Fig. 1.Study area and the water temperature fluctuation trends.(A) Location of the study area.All 294 river estuaries are represented on a map of the Japanese archipelago showing the sample sites as black dots.(B) Water temperature fluctuation trends in river estuaries.Fifteen items are shown: water temperature for each month, annual average water temperature (Tavg), annual maximum water temperature (Tmax), and annual minimum water temperature (Tmin).Water temperatures are color-coded according to the degree of water temperature change.The number of river estuaries where no significant water temperature change was observed is shown in gray.(C) Boxplot of annual change rates in water temperature of Tavg, Tmax, and Tmin.(D) Boxplot of annual change rates in water temperature of each month.