Changes in Major Global River Discharges Directed into the Ocean

Under the influence of global climate change, the discharges of major global rivers directed into the ocean have undergone significant changes. To study the trends and causes in discharge variation, we selected 40 large rivers and analyzed their annual discharges near their estuaries from 1960 to 2010. The method of runoff variation attribution analysis based on the Budyko hypothesis for large-scale basins was developed, in which influencing factors of human activities and glacial melting factors were added to the formula. The contribution rate of climate factors and human activities to changes in discharge were quantitatively identified. Climatic factors include precipitation, evapotranspiration and glacial melting. Human activity factors include underlying surface and artificial water transfer. The contribution rate is determined by the elastic coefficient, which is obtained by the ratio of change rate of each factor and the change rate of runoff. The results indicated that the discharges predominantly showed downward trends with a few upward trends. Rivers in North America and Africa showed downward trends, and those in Europe principally showed upward trends. Climate was the main influencing factor of discharges changes, and only approximately 25% of river discharges were greatly affected by human activities. River discharges in 75% of the basins which mainly contains subtropical monsoon humid climate and savanna climate zones showed upward trends. In the four basins which are mainly contains tropical rainforest climate and tropical monsoon climate, they all showed downward trends. The trend of discharges in the temperate monsoon climate, temperate continental climate, and temperate maritime climate cannot be accurately judged because of irregular variation. The discharges in the mid-high latitudinal zones predominantly showed upward trends, while those in the mid-low latitudinal zones with the influence of human activities showed downward trends.


Introduction
The discharge of large rivers into the oceans constitutes an important component of the global hydrological cycle and an important link for the coupling between terrestrial and oceanic hydrological cycles. In recent years, the global mean surface temperature (GMST) has continued to increase [1,2]. The frequency of extreme events such as heavy rainfall and drought has increased significantly in various regions of the world [2][3][4][5][6]. Combined with the impact of human activities, discharges may change significantly [7,8]. At present, the environmental degradation of many estuaries in the world is a serious issue that poses a threat to water supply security, ecological environment safety, and sustainable socioeconomic development. Discharge directed into the ocean represents the final process in the complex water cycle, and its changes are comprehensive manifestations of climate change When analyzing the changes of discharge in some large rivers, most of the previous studies select only one or several river basins, and the time series of data are different, which makes it difficult to compare runoff changes in multiple basins in the same time period. Method of attributive analysis of runoff variation based on the Budyko hypothesis is rarely used in the analysis of runoff variation at large basin, it is mostly used to study the source region of rivers. In this research, the collected data are interpolated to the same time series by using various methods, and the processed runoff data is analyzed to obtain the variation and influencing factors in the major river basins in the same period, which will provide a basis for studying changes in basin runoff and water cycle over a certain period.

Study Area
This study considers factors such as data richness, geographical representation, natural and social development of the basin and the extent of the basin area. Forty major rivers in the world are selected, in Asia, Africa, North America, South America, Europe, and Oceania and cover a wide range of climatic zones and latitudes. The drainage area, flow regions, and natural conditions are different. The total drainage area is approximately 60×10 6 km 2 , accounting for approximately 45% of the global land area. The distribution of the selected river basins is shown in Figure 1.

Data
The discharge data used in this paper is collected from the Global Runoff Data Center (GRDC), the U.S. Geological Survey (USGS) Hydro-Climatic Data Network (HCDN), the Canada HYDAT Database (HYDAT), the Australian Bureau of Meteorology (ABM), the South American Observation Service SO HYBAM (formerly the Environmental Research Observatory) (http://www.orehybam.org/), and other research projects. Considering the representativeness of the geographical locations of the sites and the reliability and completeness of the data, we selected the observed and simulated annual discharge data for the major rivers at the stations near the estuaries for the past 50 years  and then processed the data through interpolation and prolongation to make a complete series. The data sources and references are shown in Table 1.

Data
The discharge data used in this paper is collected from the Global Runoff Data Center (GRDC), the U.S. Geological Survey (USGS) Hydro-Climatic Data Network (HCDN), the Canada HYDAT Database (HYDAT), the Australian Bureau of Meteorology (ABM), the South American Observation Service SO HYBAM (formerly the Environmental Research Observatory) (http://www.ore-hybam.org/), and other research projects. Considering the representativeness of the geographical locations of the sites and the reliability and completeness of the data, we selected the observed and simulated annual discharge data for the major rivers at the stations near the estuaries for the past 50 years  and then processed the data through interpolation and prolongation to make a complete series. The data sources and references are shown in Table 1. The global digital elevation (DEM) data used in this study are derived from ASTER GDEM data with a resolution of 30 m × 30 m. Considering that the coverage area of the studied basins is large, we resampled the DEM data at a resolution of 1 km × 1 km to generate the river basins in ArcGIS.
The precipitation and temperature data are monthly data from 1900 to 2014 with a resolution of 0.5 • × 0.5 • , which was derived from the Global Precipitation Climatology Center (GPCC). On this basis, the monthly precipitation and temperature raster data of 40 watersheds from 1960 to 2010 were extracted, and their annual values were calculated.
The actual evapotranspiration and potential evapotranspiration data from 1948 to 2010 with a monthly scale and a resolution of 0.25 • × 0.25 • were derived from the Global Land Surface Data Assimilation System (GLDAS). According to the research, the monthly data of actual evapotranspiration and potential evapotranspiration from 1960 to 2010 in 40 watersheds -were extracted, and the annual values were calculated.

Linear Tendency Estimation
In this method, x i represents a measured variable for a total number of samples n, and t i represents the time corresponding to x i . A linear regression equation can be established between x i and t i : where a is the regression constant and b is the regression coefficient. The sign of b indicates the trend tendency of the climate variable x.

Mann-Kendall Test
A Mann-Kendall (hereinafter referred to as M-K) test is a nonparametric statistical testing method that can be used to analyze the trend and mutation of a time series. In an M-K trend test [49], the statistical value Z is defined as: where n is the length of the sequence, S is a statistical value: where sgn() is a symbolic function, when (X i − X j ) is less than, equal to, or greater than zero, sgn(X i − X j ) is −1, 0, or 1 respectively. For a random sequence X = (x 1 , x 2 , . . . , x n ), at the given significance level α, the critical test value Z α/2 can be found from a normal distribution table. Then, |Z| and Z α/2 are compared to determine the significance of the change.

Cumulative Departure Curve
The cumulative departure curve is a commonly used method for judging the trend change of a curve. For a sequence x, the cumulative departure at time t is expressed as follows: where: All of the cumulative distances are calculated, after which we can plot the cumulative departure curve to conduct trend analysis.

Method of Runoff Variation Attribution Analysis Based on the Budyko Hypothesis for Large-Scale Basins
The water balance mechanism in the natural water system is the amount of input water equal to the amount of output water plus the change in the water storage capacity. The formula is expressed as: Further expressed as: where R represents the observed runoff in the natural system, that is, the natural runoff; P represents the precipitation; E represents the actual evapotranspiration; ∆S represents the change in the storage capacity.
It is generally considered that in the natural state, when the large basin scale is studied and the research period is long, the change in the water storage capacity can be neglected, that is, ∆S = 0. The expression is as follows: The coupled water-energy balance based on the Budyko hypothesis and named the Choudhury-Yang formula [50] characterizes the long-term water-energy balance relationship of the basin under certain climatic and underlying conditions. There are two preconditions for applying this formula. One is that the basin must be a closed basin, that is, the surface and the underground watershed are consistent; the second is that the change in the water storage capacity of the basin for many years is negligible [29]. The concrete expression is as follows [51]: where E is the average annual actual evapotranspiration (mm), P is the average annual precipitation (mm), E 0 is the average annual potential evapotranspiration (mm), and n is the underlying surface coefficient, which can reflect the characteristics of the underlying surface of the watershed and can be calculated from the known parameters in the formula. If we consider the impact of human activities on the water balance of the water system, then equation (8) can be written as: where R s is the measured runoff and Rr is the amount of artificial water withdrawal and water transfer or glacial meltwater caused by global warming. When R r indicates the amount of artificial water withdrawal, it is negative; when Rr indicates the amount of artificial water transfer or glacial meltwater, it is positive. P is the precipitation, and E is the actual evapotranspiration. Then, bringing formula (9) into formula (10), the following expression is obtained: From formula (11), the measured runoff R s can be expressed as a function of P, R r , E 0 and n, that is, R s = f (P, R r , E 0 , n). The elastic coefficient refers to the ratio of the growth rate of two interconnected indicators in a certain period of time. For example, the precipitation elastic coefficient refers to the change in the runoff caused by the change in precipitation, expressed as: Similarly ε E 0 , ε R r and ε n can be calculated; thus, the total differentiation form of the runoff variation is obtained as follows:

Trends
Through the annual discharge variation lines and M-K trend analysis in the 40 rivers from 1960 to 2010 ( Figure 2, Table 2), the Mississippi River, Orinoco River, and five other rivers reached a significance level of α = 0.05 with a significant upward trend. Ten rivers, such as the Rio Grande and Murray-Darling Rivers, reached a significance level of α = 0.05 with a significant downward trend. The Columbia River and the Sao Francisco River reached a significance level of α = 0.1 with a significant downward trend. The trends in the other rivers did not pass the significance test of α = 0.1, indicating that their trends are not obvious. According to the comprehensive analysis, the discharges of 7 rivers increased significantly, and those of 12 rivers significantly decreased. Eleven of the other 21 rivers showed upward trends, and 10 showed downward trends, but their trends were not significant. a * indicates a clear trend; spaces indicate that the trend is not obvious. b * represents a significance level of α=0.1, ** α=0.05, *** α=0.01, and **** α=0.001. c ++ represents a significant upward trend, --represents a significant downward trend, + represents an upward trend that is not significant, and -represents a downward trend that is not significant.
Among the nine selected rivers in North America, six rivers (two-thirds of the North American rivers) showed downward trends. Similarly, four of the six rivers (a large proportion) in Europe showed an upward trend. In Africa, one of the five selected rivers showed an upward trend, and thus, downward trends accounted for 80% of the total. In South America and Asia, the proportion of upward and downward trends was generally consistent.

Figure 2. River discharge trends
Through the cumulative departure curve and M-K mutation test methods, the results for the segment trends were different. The results obtained by a comprehensive analysis of the two methods are shown in Table 3. Among the 40 rivers, there are 16 rivers showing the trend of "wet-dry", 13 rivers showed the trend of "dry-wet", eight rivers showed the trend of "dry-wet-dry", and one river showed the trend of "wet-dry-wet". The two rivers showed a trend of "wet-dry-wet-dry", were the Pearl and Rhine rivers, but the mutation time of the Pearl River relatively lagged behind the Rhine River. Among them, the North American rivers mainly showed the trend of "wet-dry" with a mutation time around the 1980s; the five rivers in Africa all showed the trend of "wet-dry" with a mutation time around the 1970s; the mutation time of the European and Asian rivers was mainly in the 1980s and 1990s. The mutation time varies among rivers in South America. Table 3. Mutation and segmentation trends of the river discharge changes.

River
Cumulative anomaly
Among the nine selected rivers in North America, six rivers (two-thirds of the North American rivers) showed downward trends. Similarly, four of the six rivers (a large proportion) in Europe showed an upward trend. In Africa, one of the five selected rivers showed an upward trend, and thus, downward trends accounted for 80% of the total. In South America and Asia, the proportion of upward and downward trends was generally consistent.
Through the cumulative departure curve and M-K mutation test methods, the results for the segment trends were different. The results obtained by a comprehensive analysis of the two methods are shown in Table 3. Among the 40 rivers, there are 16 rivers showing the trend of "wet-dry", 13 rivers showed the trend of "dry-wet", eight rivers showed the trend of "dry-wet-dry", and one river showed the trend of "wet-dry-wet". The two rivers showed a trend of "wet-dry-wet-dry", were the Pearl and Rhine rivers, but the mutation time of the Pearl River relatively lagged behind the Rhine River. Among them, the North American rivers mainly showed the trend of "wet-dry" with a mutation time around the 1980s; the five rivers in Africa all showed the trend of "wet-dry" with a mutation time around the 1970s; the mutation time of the European and Asian rivers was mainly in the 1980s and 1990s. The mutation time varies among rivers in South America. Table 3. Mutation and segmentation trends of the river discharge changes.

Spatial Distribution
From the statistical results, in the seven rivers with significant upward trends (Figure 3), the Orinoco River Basin is located in the tropics between 0 • and 10 • N, and the main climate type is a tropical savanna climate. The main climate in the other six river basins between 30 • N-75 • N, which is the mid-high latitude zone, is a temperate continental climate. Except for the Mississippi and Orinoco, the other five rivers belong to the Arctic river and are directed into the Arctic Ocean.    The twelve rivers that showed significant downward trends ( Figure 4) are all located in the mid-low latitude zone. The Congo, Nile and Sao Francisco rivers are located in tropical climates, and the other nine rivers are located in temperate climates. Regarding the main climate types, the Sao Francisco River Basin is located in a tropical savanna climate, and the rest have diverse climate types.

Influencing Factors
In this study, the factors affecting climate change mainly include precipitation, evapotranspiration and glacial meltwater. The comprehensive contribution rate of precipitation and evapotranspiration is recorded as δ z , and the total contribution rate of climatic factors is recorded as δ c . The influencing factors of human activities mainly include underlying surface conditions and Among the 10 rivers with an insignificant downward trend, from the statistical results ( Figure 6), except for the Niger, Amazon, Mekong and Pearl River Basins located at low latitudes, the other six basins are located at the mid-high latitudes. The Niger and the Amazon River Basin are located in the tropics, the Mekong River Basin spans the tropics and the temperate zone, and the remaining 7 basins are located in the temperate zone. There are many climate types in the 10 basins.

Influencing Factors
In this study, the factors affecting climate change mainly include precipitation, evapotranspiration and glacial meltwater. The comprehensive contribution rate of precipitation and evapotranspiration is recorded as δ z , and the total contribution rate of climatic factors is recorded as δ c . The influencing factors of human activities mainly include underlying surface conditions and

Influencing Factors
In this study, the factors affecting climate change mainly include precipitation, evapotranspiration and glacial meltwater. The comprehensive contribution rate of precipitation and evapotranspiration is recorded as δ z , and the total contribution rate of climatic factors is recorded as δ c . The influencing factors of human activities mainly include underlying surface conditions and artificial water withdrawal and transfer, and the contribution rate is recorded as δ l . The underlying surface coefficient n, the elastic coefficient ε, and the contribution rate δ are shown in Table 4.

Contribution Rate of Influencing Factors
From the above analysis results, we found that the contribution rate of precipitation in all basins is above 29%, and the precipitation elastic coefficient in 27 basins is larger than other factors. We can conclude that precipitation factors dominate the long-term discharge changes in these basins. The contribution rate of E 0 is below 24%, and the impact is relatively small. The contribution rate of the R r is less than 20%, except for the Yana River and the Lena River, indicating that the influence of this factor is relatively small.
In the seven basins where the discharge has a significant upward trend (Table 4), the δ z values in the Lena River and the Yana River are less than 50%. The top three basins have precipitation contribution rates of 55%, 48%, and 42%, and the precipitation in these three basins has an upward trend. The Northern Dvina River has passed the 90% significance test; therefore, the discharge changes in these three basins are mainly attributed to the increase in precipitation. The precipitation trends of the Olenyok, Yenisei, and Yana rivers are not obvious, but the R r is positive and the contribution rate is large. These three rivers are all Arctic rivers and are considered to be affected by glacial meltwater; that is, the δ c in the basin is 71%, 72%, and 74%, respectively, so the discharge is mainly affected by climatic factors. The North Dvina River is also an Arctic river with few human activities and a positive value of R r in the basin; therefore, R r mainly indicates an increase in glacial meltwater and δ c is 77%. In the Lena River Basin, although the contribution rate of precipitation is only 33%, the precipitation has a clear upward trend and passes the 99% significance test. It is also an Arctic river, the R r is positive and the contribution rate is large, and the influence of glacial meltwater is also considered to be large, so the δ c is 74%. The R r of the Orinoco River is negative, indicating artificial water withdrawal. The Mississippi River Basin is less affected by glacial meltwater, and it is considered that R r mainly means artificial water transfer, that is, the δ l values are 26% and 38%. According to the analysis, seven river discharges with significant upward trends are mainly affected by climatic factors.
In the 12 basins where the discharge has a significant downward trend, the δ z values in the four basins are greater than 50%. Among them, the precipitation in the Congo River Basin showed a significant downward trend, and the significance test confidence level reached 99.9%; thus, the decrease in discharge was mainly due to a significant decrease in precipitation. In the Sao Francisco River Basin, the precipitation did not change significantly, but the δ E 0 and the δ n values is relatively high (19% and 37%). So we considered that the significant decrease in discharge is both affected by climatic factors and human activities. In the four basins, only the R r value in the Fraser River Basin is positive due to the mountainous areas and few human activities in the basin. It is considered that R r mainly indicates an increase in glacial meltwater, that is, the δ c is 76%. The R r values in the Congo River and the Sao Francisco River are negative, which suggest artificial water withdrawal, that is, the δ l values are 35% and 42%, respectively. In the Yellow River Basin, the δ z is 50%, and the R r is negative, indicating human water withdrawal; thus, the δ l is 50%, which shows that both climatic and human activity has an impact, but the δ n is 41%; therefore, it is believed that the influence of human activities in this basin is dominant. The δ z in the seven basins is less than 50%, of which the δ z in Columbia River is 49% and the R r is positive. Due to large amounts of water conservancy and hydropower projects in the basin, discharge is greatly disturbed by human activities, that is, R r mainly means artificial water transfer. Additionally, the R r values in the other six rivers are negative, which means artificial water withdrawal, that is, the δ c of the seven rivers is consistent with the δ z . Based on the above analysis, in the 12 basins with a significant downward trend, the Fraser, Ganges, Congo and Sao Francisco River Basins are mainly affected by climatic factors, and the remaining 8 rivers are mainly affected by human activities.
In the 11 rivers with an insignificant upward trend, only the R r in the five watersheds, such as the Magdalena River, is negative, which indicates artificial water withdrawal. The R r in the Neva River is positive with a contribution rate of 3%. Because the basin mainly includes three large lakes and the human activities are less affected, the δ c is considered to be 74%. Precipitation in the Magdalena and Neva rivers showed a significant upward trend, which passing 90% and 95% significant tests, respectively. The main factors affecting discharge changes in these two basins were climate change. So the discharge shows an insignificant upward trend that can be attributed to increased precipitation. The R r values in the Ob, Pechora and Kolyma rivers are positive, and the three rivers are all Arctic rivers. The R r is considered to be mainly affected by glacial meltwater; that is, the δ c values in the three basins are 64%, 79%, and 71%. The R r of the Yangtze and Nelson rivers are positive, and the contribution rate is 4%. Because of the large number of water conservancy projects in these two basins, the δ l in the two basins were 31% and 39%. According to the above analysis, the discharge of the Zambezi and the Colorado River Basins is mainly affected by human activities, and the remaining 9 rivers are mainly affected by climatic factors.
In the 10 watersheds with an insignificant downward trend, the R r values in the Rhine, Pearl, Mekong and Niger River Basins are negative and indicate artificial water withdrawal and the values in the four basins are 27%, 34%, 39%, and 55%, respectively. The precipitation in the Niger River Basin shows a significant downward trend, passing a 90% significance test. It is believed that the cause of the discharge changes includes both precipitation and human activities, but human activities have a greater impact. The R r values in the Yukon and the Mackenzie River Basin are positive, and the two rivers are all Arctic rivers. The human activities in these basins are less, and the R r mainly indicates an increase in glacial meltwater and the δ c values in the two basins are 70% and 69%, respectively. The R r in the Amazon, St. Lawrence, Danube and Amur River Basins are positive with contribution rates of 6%, 3%, 9%, and 7%, respectively, indicating that the impact of R r on discharge changes is relatively small and the four basins are less affected by glacial meltwater, so R r mainly indicates artificial water transfer and the δ l in the four basins are 26%, 30%, 35%, and 42%. According to the above analysis, only the Niger River discharge is greatly affected by human activities, and the remaining nine rivers are mainly affected by climatic factors.
Based on the above analysis, 29 river basin discharge changes are mainly affected by climatic factors, and 11 river basins are mainly affected by human activities, including the Yellow, Columbia, Nile, Indus, Murray-Darling, Rio Grande, Orange, Zambezi, Colorado, Niger, and the Two River Basin. The distribution of factors affecting the discharge variation in each basin is shown in Figure 7.  Based on the above analysis, 29 river basin discharge changes are mainly affected by climatic factors, and 11 river basins are mainly affected by human activities, including the Yellow, Columbia, Nile, Indus, Murray-Darling, Rio Grande, Orange, Zambezi, Colorado, Niger, and the Two River Basin. The distribution of factors affecting the discharge variation in each basin is shown in Figure 7.
These results are generally consistent with previous studies. For the Yellow, Colorado, Murray-Darling, Orange and Columbia River, the contribution rate of human activities in this study is between 50% and 60%. It is considered that both climatic factors and human activities have effects, which is similar to the result of some studies [17][18][20][21][22][23]. But the result of the Murray-Darling River is different from that of the study by Potter and Chiew [20]. This may be due to different research periods. For the Arctic rivers (except the Yana River), the sum of the contribution rates of precipitation and evapotranspiration is greater than 50%, indicating that climatic factors are still the main influencing factors. This is generally similar to the results studied by Brabets et al. [12] and Adam et al. [15]. The influencing factors include snowmelt, permafrost degradation, and increased fire frequency in their research. For the Yenisei and Lena River, the contribution rate of human water transfer and glacial melting in this study is relatively large (19% and 25%), which is generally similar to the results studied by Berezovskaya et al. [9,11] and Stuefer et al. [10]. Since the discharge directed into the ocean is the link between the terrestrial water cycle and the ocean water cycle, Su et al. believed that discharge changes are also related to ocean signals such as El Niño-Southern Oscillation (ENSO), Arctic Oscillation (AO), North Atlantic Oscillation (NAO) and Pacific Decadal Oscillation (PDO) [52]. These results are generally consistent with previous studies. For the Yellow, Colorado, Murray-Darling, Orange and Columbia River, the contribution rate of human activities in this study is between 50% and 60%. It is considered that both climatic factors and human activities have effects, which is similar to the result of some studies [17,18,[20][21][22][23]. But the result of the Murray-Darling River is different from that of the study by Potter and Chiew [20]. This may be due to different research periods. For the Arctic rivers (except the Yana River), the sum of the contribution rates of precipitation and evapotranspiration is greater than 50%, indicating that climatic factors are still the main influencing factors. This is generally similar to the results studied by Brabets et al. [12] and Adam et al. [15]. The influencing factors include snowmelt, permafrost degradation, and increased fire frequency in their research. For the Yenisei and Lena River, the contribution rate of human water transfer and glacial melting in this study is relatively large (19% and 25%), which is generally similar to the results studied by Berezovskaya et al. [9,11] and Stuefer et al. [10]. Since the discharge directed into the ocean is the link between the terrestrial water cycle and the ocean water cycle, Su et al. believed that discharge changes are also related to ocean signals such as El Niño-Southern Oscillation (ENSO), Arctic Oscillation (AO), North Atlantic Oscillation (NAO) and Pacific Decadal Oscillation (PDO) [52].

Impact of Climate Zone on River Discharge
In the Mississippi, Yangtze, Pearl and the Parana River Basins, the main climate type is a subtropical monsoon humid climate, and the discharge changes are dominated by climatic factors. All basins are located between 20 • -40 • in the north and south latitudes, except for the insignificant downward trend of the Pearl River discharge, the other three basins all have an upward trend, therefore, the discharge in the basin in the subtropical monsoon humid climate zone mainly has an upward trend. The Congo and the Amazon River Basin include the tropical rainforest climate, and the discharge changes are dominated by climatic factors. The discharge is declining, and the confidence level in the Congo River Basin is 99.9%. It is considered that the discharge in the tropical rainforest climate zone mainly has a downward trend. The Magdalena, Orinoco, Sao Francisco and the Zambezi River Basins all contain a large proportion of the tropical savanna climate, except for the downward trend of the discharge in the Sao Francisco River Basin, the other basins showed an upward trend, and in the Orinoco River, the confidence level reached 99%. As mentioned above, the conditions of the underlying surface in the Sao Francisco River Basin are relatively large, so the discharge in the tropical savanna climate zone mainly has an upward trend. The Alpine climate accounts for a large proportion in the Columbia and Fraser River Basins, and the discharge changes in both basins are significantly declining. The discharge changes in the Fraser River Basin are mainly affected by climatic factors. Both climate factors and human activities impact the Columbia River Basin, so the change in discharge in the Alpine climatic zone mainly shows a downward trend, but the confidence is low.
The tropical monsoon climate is the main climate type in the Ganges and the Mekong River Basin, and the discharge is mainly affected by climatic factors and is decreasing, therefore, the discharge in the tropical monsoon climate zone mainly shows a downward trend. The temperate monsoon climate in the Amur River Basin is the main climate type and the δ c in the basin is 58%, but the δ n is 35%. The discharge in this basin is decreasing but is not significant, so the discharge changes in the temperate monsoon climate are not obvious. In the 11 river basins, namely, the Yukon, Mackenzie, St. Lawrence, Nelson, Northern Dvina, Danube, Don, Pechora, Neva, Ob and Kolyma River Basin, most of the climate types are the temperate continental climate, and the discharge changes are mainly affected by climatic factors; however, only the discharge in the Northern Dvina River Basin shows a significant trend. Due to the wide distribution of temperate continental climatic zones, the regional differences are large, so the trend is insignificant in the temperate continental climate zone. In the Rhine River Basin, the temperate marine climate is the main climate type, the change in discharge is dominated by climatic factors, and the discharge trend is not obvious. Thus, the discharge in the temperate maritime climate zone is considered basically stable.
Although few studies have focused on the impact of climatic zones on runoff, some studies have been summarized by geographic location and can also support the findings of this paper. For example, runoff in the subtropical humid monsoon and savanna climate zones predominantly showed upward trends in this study, and eight rivers (the Mississippi, Yangtze, Pearl, Parana, Magdalena, Orinoco, Sao Francisco, and Zambezi River) mainly contain these two climatic zones. The studies (Dai et al. [25] and Milly et al. [53]) concluded that the runoff near the Gulf of Mexico, southern South America, and southeastern Africa showed upward trends. In this study, runoff in the tropical rainforest and the tropical monsoon climate zones predominantly showed downward trends, and four rivers (the Congo, Amazon, Ganges, and Mekong River) mainly contain these two climatic zones. In the conclusion studied by Milly et al. [53] and Gerten et al. [54], the runoff in Sub-Saharan Africa and Central /South Asia showed downward trends, which is generally similar to the results in this study.

Uncertainty
During the data selection, we first considered the data abundance at most of the sites and the lack of data at other sites. Then we considered the impact of climate change and human activities and chose 1960 as the base year. The data series are mainly measured data, and the missing data is obtained by data extension and interpolation from other data sources. This process may cause data errors, but from the result analysis, the general trend of change is basically consistent with the previous study [15,16,18,20,22,24,25], so the error caused by the data is negligible.
In the analysis of the attribution of the changes in river discharge, we developed the analysis method of the discharge variation attribution based on the Budyko hypothesis to adapt for large-scale basins. The influence factors of human activities and glacial meltwater were added into the formula represented by R r . In the analysis of the effects of R r , qualitative identification is currently only based on the geographical location of the basin and the development of human society. For example, the Arctic basin was less affected by human activities, and the R r in these basins represents the influence of glacial meltwater. If there are many water conservancy and hydropower projects in the basin and the glacier has less impact, then the R r represents artificial water transfer. Due to the large scale of the basin, the error is relatively small. In future study, more data will be considered for in-depth analysis.

Conclusions
By analyzing the trends and distribution characteristics and the attribution of the discharges directed into the ocean in the 40 major global rivers, we conclude the following: (1) At present, the discharges of the major global rivers directed toward the oceans more preferentially showed downward trends and fewer upward trends. The downward trends accounted for 57.5% and the upward trends accounted for 42.5% of the total. In North America and Africa, approximately two-thirds of the rivers showed downward trends. In Europe, 80% of the rivers showed upward trends. In Asia and South America, the number of rivers that showed upward trends and downward trends was approximately equal. In Oceania, the Murray-Darling River is the major river, and it showed a significant downward trend. Regarding the segmentation trend, the North American rivers mainly show the trend of "wet-dry" with mutation points around the 1980s. The five rivers in Africa all showed the trend of "wet-dry" with a mutation time in approximately the 1970s. The mutations in different rivers in South America were different. The mutation time in the European and Asian rivers was mainly in the 1980s and 1990s.
(2) Climatic factors were the main influencing factors for the change in the discharge of major global rivers directed into the ocean, and only approximately 25% of river discharge was greatly affected by human activities. From the perspective of the climate zone, the river discharge in the subtropical monsoon humid climate and tropical savanna climate zone mainly exhibited an upward trend. In the tropical rainforest climate and the tropical monsoon climate zone, river discharge mainly demonstrated a downward trend. The change in the river discharge in the temperate monsoon climate, the temperate continental climate, and the temperate marine climate were relatively stable. From the perspective of latitude, the discharge in the mid-high latitudes mainly showed an upward trend, and the discharge in the mid-low latitudes with the influence of human activities mainly showed a downward trend. The runoff coefficient also decreased.