1. Introduction
Dams are critical regulating infrastructures that play an important role in managing water resources. These infrastructures are constructed to achieve various purposes, including flood control by regulating streamflow during periods of heavy rainfall and reducing the risk of downstream flooding (Hecht et al. 2019; Nguyen-Tien et al. 2018; Yao et al. 2019). Dams also impound reservoirs to provide water supply for drinking, irrigation, and industrial purposes (Altinbilek 2002; Khalkheili and Zamani 2009; Tatlhego et al. 2022). Many dams were designed and constructed with hydrologic information decades ago (Ho et al. 2017; Chen and Hossain 2019; Perera et al. 2021). Over time, hydrology of the upstream watersheds of these dams and subsequently the dam inflows have changed due to the changes in climate and land cover and subsequently hydrologic regime; these have affected the functionality of the dams (Getachew and Melesse 2012; López-Moreno et al. 2014; Babur et al. 2016; Naz et al. 2018; Muhammad et al. 2020).
Given the observed changes in climatic variables, land use, hydrologic regime, and vegetation, it is important to evaluate temporal changes in the dam inflows in order to inform the reservoir managers about the needs for future adaptation plans and revisiting the operation rules. Asarian and Walker (2016) applied nonparametric Mann–Kendall tests on precipitation and streamflow from 1953 to 2012 in Northern California and south Oregon and identified a decreasing trend in streamflow at most stream gauges. Kalra et al. (2017) found that, from 1906 to 2010, in the Colorado River basin, streamflow increased in winter–spring and decreased in summer–autumn, with temperatures rising and streamflow and precipitation decreasing after the 1930s. McCabe et al. (2017) noted that since the late 1980s, increases in temperature in the Upper Colorado River basin have caused a substantial reduction in river flows. Naz et al. (2018) investigated the impact of climate change on dam inflows and found significant decreasing trends in dam inflows in the western United States. Xiao et al. (2018) reported that the naturalized flow of the Colorado River has decreased by about 15% from 1916 to 2014, with approximately half of this long-term trend attributable to rising temperatures and the remainder to changes in precipitation patterns.
However, large-scale analyses that evaluate changes in several dams have been limited. Such studies are needed to guide coordinated efforts that are required at the watershed and regional scale to achieve integrated goals of water supply, flood control, and irrigation, among others. Past research has primarily focused on changes at an annual time scale with limited attention to seasonality (timing) of low/high dam inflows and changes at the seasonal time scale (e.g., Fleming and Weber 2012). A further limitation in the body of literature is that only a few studies explored the role of multiple factors that can contribute to certain trends in dam inflows.
The objective of this paper was to explore temporal changes in magnitude and seasonality of low, median, and high inflows of 51 dams across the western and southwestern United States. Temporal changes in precipitation, air temperature (as an indicator of snowpack and evaporation), soil moisture, and normalized difference vegetation index (NDVI; represented vegetation) were also examined to identify potential reasons for significant changes in the dam inflows. Using monotonic and nonmonotonic trend tests as well as changepoint detection, we analyzed temporal changes in inflows and four driving factors (precipitation, air temperature, soil moisture, and vegetation) at annual and seasonal time scales. In addition, we explored the changes in the timing of low and high inflows. All analyses were conducted on the dam inflows and the four driving factors across the upstream watersheds over the 30-yr period of 1993–2022. The unique aspect of this study is investigating multiple factors that explain the temporal trend of reservoir inflows at the annual and seasonal time scales.
2. Study area
We studied 51 dams in the western and southwestern United States owned and operated by the U.S. Bureau of Reclamation (USBR). These dams were selected based on data availability; only those with at least 30 years of dam inflows (1993–2022 period) were chosen. Among these, the dams with large data gaps (e.g., about a continuous year) and dams under significant influence of upstream dam were discarded from our analyses. Additionally, we analyzed the connectivity of each dam using the National Hydrography Dataset (NHD; USGS 2004). This dataset provided detailed information on river networks, allowing us to identify dams that were significantly influenced by upstream dams. We excluded dams directly connected to upstream dams that shared a single river Geographic Names Information System (GNIS) identifier as defined in the NHD. This led to the final list of 51 dams located in six states—California, Colorado, Montana, New Mexico, Utah, and Wyoming (Fig. 1). These dams were located across four hydrologic regions [USGS’s Hydrologic Unit Code 2 (HUC2)] of California, Great Basin, Missouri, and Upper Colorado (Fig. 1).
Location of 51 dams in the western and southwestern United States investigated in this study.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
The dam inflow data were acquired from the USBR’s Reclamation Information Sharing Environment (USBR 2020). The selected dams have a variety of purposes, including irrigation, water supply, flood control, fish and wildlife conservation, hydroelectric generation, and recreation according to National Inventory of Dams (NID) of the United States Army Corps of Engineers (2021). Given the variety of purposes, analyses of both low and high inflows were relevant on these dams. All of these dams are classified as high-hazard according to NID.
We also incorporated monthly precipitation, air temperature, and soil moisture data from the fifth major global reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) (ERA5)-Land dataset (Muñoz Sabater 2019). This global dataset combines modeled data with observations using the physical laws to produce a complete and consistent record from 1981 onward. The native spatial resolution of ERA5-Land is 9 km on a reduced Gaussian grid and the data have been regridded to a regular latitude–longitude grid of 0.1° × 0.1° (Muñoz Sabater 2019).
In addition to the climatic variables, we incorporated monthly NDVI data from the NOAA climate data record (CDR) program (Vermote 2019, 2022). The NDVI dataset provides a long-term, consistent record of vegetation cover, derived from satellite observations, spanning from 1981 to the present time. This dataset represents vegetation dynamics across the western and southwestern United States. The spatial resolution of the NDVI data is 0.05° × 0.05°, comparable with the datasets used for the three other driving factors.
3. Methodology
We analyzed the temporal trend of inflows for the selected dams over the period of 1993–2022 at annual and seasonal time scales. The analyses were conducted on three categories of inflow: low (Q5 or fifth percentile), median (Q50 or 50th percentile), and high (Q95 or 95th percentile). For the annual time scale, we also explored shifts in the timing of minimum and maximum inflows. Both monotonic and nonmonotonic trends were studied. The same trend analysis method was repeated for the four driving factors—precipitation, air temperature, soil moisture, and NDVI—at annual and seasonal time scales. Precipitation and soil moisture contribute to dam inflows through affecting runoff generation and drought periods. Air temperature was used as an indicator of (or proxy for) snowpack and evaporation, while vegetation affects the dam inflow by affecting surface roughness, evapotranspiration, and runoff production. A schematic of our methodology is presented in Fig. 2.
Schematic view of the methodology.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
To prepare basin-scale upstream averages of precipitation, air temperature, soil moisture, and NDVI, we utilized a digital elevation model (DEM). First, we delineated the boundary of the upstream watersheds for each of the 51 dams using the 30-m DEM from USGS’s National Elevation Dataset (USGS 2022). We then derived time series of the driving factors—precipitation, air temperature, soil moisture, and NDVI—across these watersheds. We then calculated the spatial average values of these variables for each watershed by aggregating the data from the corresponding grid cells. This process was done for both annual and seasonal time scales.
Using the time series of dam inflow and the four driving factors—precipitation, air temperature, soil moisture, and NDVI—temporal trends were investigated over our period of interest (1993–2022). Figure 3 shows examples of five types of temporal trends conducted on the dam inflow data.
Examples of five types of temporal trends in this study: (a) upward trend, (b) downward trend, (c) consistent nonmonotonic upward trend, (d) consistent nonmonotonic downward trend, and (e) inconsistent trends.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
a. Dam inflow
The low and high inflows provide information on hydrologic extremes—droughts and floods—that are important for water supply, irrigation, and flood risk reduction (Loukas et al. 2007; Kumambala and Ervine 2009; Whitehead et al. 2018). High inflows are also important in terms of overtopping failure risk of the dams (Yu et al. 2014; Ahmadisharaf and Kalyanapu 2015; Monteiro-Alves et al. 2022). The analyses of median inflows also provide information on central tendency of the dam inflows. We conducted these analyses at annual and seasonal time scales; the latter time scale included spring (March–May), summer (June–August), fall (September–November), and winter (December–February). For the annual time scale, we also explored shifts in the timing of low and high inflows. This helped us investigate whether there is a shift in the timing of minimum and maximum dam inflows over time.
b. Driving factors
We analyzed ERA5-Land precipitation, air temperature data, and soil moisture corresponding to the dam inflow over the period of interest (1993–2022). The monthly precipitation data were aggregated into annual and seasonal time scales; the latter time scale included spring, summer, fall, and winter, which matches with the time scales of dam inflow data. For air temperature and soil moisture, we calculated the mean values for both annual and seasonal time scales, ensuring consistency with the precipitation data and the dam inflow analyses. All these variables were processed across the upstream watersheds.
c. Vegetation
We analyzed NDVI data from the Advanced Very High Resolution Radiometer (AVHRR) and the Visible Infrared Imaging Radiometer Suite (VIIRS) provided by NOAA for the period of 1993–2022 (Vermote 2019, 2022). These daily data were aggregated into monthly averages to match the time step of the other variables. The NDVI values were then processed to create annual and seasonal averages for each watershed, consistent with our methodology for the other driving factors—precipitation, air temperature, and soil moisture.
d. Monotonic trend analyses
Monotonic trend analyses were performed using the original Mann–Kendall test (Mann 1945; Kendall 1975). The Mann–Kendall is a nonparametric test, which is known for its ability to identify monotonic trend of hydrologic data (Burn and Elnur 2002; Yue et al. 2002; Hamed 2008; Fathian et al. 2016; Tosunoglu and Kisi 2017; Wang et al. 2020). We evaluated the autocorrelation of the time series to ensure that the test can be performed. If the data are autocorrelated, other tests like modified Mann–Kendall (Hamed and Rao 1998; Yue and Wang 2004) were used. The dam inflow data and the precipitation data were not statistically autocorrelated, so we were able to use the original Mann–Kendall test. We used the pyMannKendall python package (Hussain and Mahmud 2019) to perform the monotonic trend analyses.
e. Nonmonotonic trend analyses
Nonmonotonic trend analyses were performed using multiple changepoint detection methods based on the pruned exact linear time (PELT) algorithm (Killick et al. 2012). In the PELT algorithm, multiple changepoints were determined based on log-likelihood and penalty function (Zhang and Siegmund 2007). We selected the modified Bayesian information criterion (mBIC) to derive the log-likelihood and penalty functions. The PELT algorithms and mBIC have been used in previous studies for identifying changepoints (e.g., Touzani et al. 2019; Gosiewska et al. 2021; Xie et al. 2021; Jegede and Szajowski 2022). Here, a changepoint was defined as the point at which the regression slope significantly changes. Using the PELT algorithm, it was possible to identify the slopes and changepoints that minimize the penalty function and maximize log-likelihood.
f. Time shift in dam inflows
For the annual low and high inflows, we explored temporal changes in the occurrence (time shift). We evaluated the frequency of occurrence of low and high inflows for each year during the period of analyses (1993–2022) for all 51 dams. To evaluate the frequencies, dam inflows higher than the high inflow (>Q95) and those lower than low inflow (<Q5) were analyzed. The period was grouped into three decadal time windows: 1993–2002, 2003–12, and 2013–22. We then explored the frequency of occurrence of these flow thresholds in different seasons. This enabled us to investigate the shift in the seasonality of low and high inflows.
4. Results
a. Temporal trends of driving factors
Figure 4 shows the results of monotonic and nonmonotonic trend analyses for the four driving factors—precipitation, air temperature, soil moisture, and NDVI—at the annual time scale. These analyses identified statistically significant trends across the upstream watersheds of each dam.
Monotonic trends in annual, spring, summer, fall, and winter precipitation, air temperature, soil moisture, and NDVI for the period of 1993–2022. Blue triangles represent decreasing trends, while red triangles denote increasing trends. Black and white squares indicate inconsistent trend and no trend, respectively.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
Overall, precipitation trends indicated a downward trend, especially in the California and Upper Colorado regions. For the annual time scale, more than 50% of the dams in the Upper Colorado region exhibited a significant downward trend. Additionally, three dams in the California region also showed a downward trend. However, the spring and summer seasons exhibited far fewer significant trends in precipitation. Only two dams in the Upper Colorado region showed a downward trend in precipitation during the spring season. In the fall, the significant decrease was identified in the Upper Colorado region, while four dams in California showed a significant decrease in the winter.
The temporal trend of air temperature was consistently upward. At the annual time scale, significant upward trends in air temperature were observed in most dams within the Upper Colorado region and all dams within the California region. Specifically, during the summer, over 90% of the dams experienced increasing trends, while in the fall, more than 70% of the dams exhibited the same trends. In contrast, there were no increasing trends in the spring and winter temperatures, except for three dams showing an increase trend during winter.
Soil moisture indicated a consistent downward trend across all seasons for most dams. Unlike the precipitation patterns, significant declines in soil moisture were observed during both spring and summer. Eleven dams showed a downward trend in soil moisture during the spring and nine dams exhibited the same trends in the summer.
The NDVI trends were predominantly upward across most regions, except for the spring season. Annually, and during the summer and fall seasons, over half of the dams showed an increasing trend in NDVI, indicating overall vegetation growth. In contrast, the number of dams displaying upward trends in NDVI during the spring and winter seasons was comparatively smaller than other seasons. During the spring, three dams in the Missouri region and four dams in the Upper Colorado region exhibited a downward trend in the average values of NDVI across their upstream watersheds. This decline can be attributed to wildfires that occur in the winter and spring seasons (Potter 2019; Notaro et al. 2019).
b. Temporal trends of dam inflows
Figure 5 and Table S1 in the online supplemental material show the results of trend analyses (monotonic and nonmonotonic) on low, median, and high inflows at the 51 dams. The predominant trend of dam inflows at all time scales was downward.
Monotonic trends in annual, spring, summer, fall, and winter inflow for the period of 1993–2022 (significance level of 0.05). Blue triangles represent decreasing trends, while red triangles denote increasing trends. Black and white squares indicate inconsistent trend and no trend, respectively.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
The statistically significant downward trend of the low inflows was particularly widespread in the Upper Colorado and California regions. For the annual time scale, 11 dams exhibited a significant downward trend; nine were in Upper Colorado and California regions. Seasonally, the downward trend was most prominent in the spring; of the 11 dams with this trend, 10 were in the Upper Colorado and California regions. Low inflows in the summer also had similar trends, with 10 dams showing decreased inflows, seven of which were in the Upper Colorado and California regions. In contrast, during the fall, seven dams indicated a downward trend in inflows, although none were in the California region. In the winter, six dams displayed downward trends. Figure S1a shows the decreasing trend of annual low inflows with its time series observed at Bradbury Dam.
Median inflows also demonstrated a predominance of downward trends, particularly in the Upper Colorado and California regions. At the annual time scale, 18 dams showed a significant decrease in inflows, with 14 located in the Upper Colorado and California regions. Seasonally, this pattern persisted. In the spring, nine dams exhibited a downward trend, with eight being in the Upper Colorado and California regions. Similarly, during the summer, 11 dams experienced a downward trend in the median inflows. The fall season showed a more pronounced impact, with 15 dams showing a downward trend, 12 of which were in the Upper Colorado and California regions. In the winter, 13 dams showed decreasing trends and 11 of these were in the Upper Colorado and California regions. This consistent pattern underscores a significant and widespread decreasing trend in median inflows across the Upper Colorado and California regions. Figure S1b illustrates the downward trend of summer median inflows with its time series observed at Whiskeytown Dam.
High inflows also showed a significant downward trend, consistent with the patterns identified in median and low inflows. At the annual time scale, 12 dams exhibited downward trends, with 8 of these located in the Upper Colorado and California regions. Of the four seasons, spring had the greatest number of dams (15) with decreasing trends and 11 of these were within the Upper Colorado and California regions. In the summer, 7 dams showed downward trends, while during the fall, 12 dams indicated decreasing high inflows. In the winter, 10 of the 12 dams showed downward trends were in the Upper Colorado and California regions. Figure S1c represents the decreasing trend of fall high inflows with its time series observed at Lost Creek Dam.
The downward trends identified in the Upper Colorado and California regions were also reported by previous studies on the hydrology of these regions. Asarian and Walker (2016) evaluated the long-term trends annual streamflow in Northern California and Southern Oregon using data from 67 stream gauges and found that 24% of the gauges had a downward trend. Monthly streamflow also showed consistent downward trends at over 50% of the gauges. Miller and Piechota (2011) evaluated the trend of daily streamflow at 43 stream gauges across the Colorado River basin. A downward trend in the annual streamflow was found at 67% of the gauges; at 79% of the gauges, a similar downward trend was also detected for streamflow between April and July. Wright et al. (2023) found a downward trend in the annual streamflow of 65 stream gauges in Utah, which encompasses a large portion of the Upper Colorado region, during the period 1980–2021.
c. Seasonality in dam inflows
We found changes in the timing of annual low inflows, which primarily occurred in the spring and summer. From 1993 to 2022, there was an increased proportion of low inflows during summer and fall along with a decreased occurrence in winter (Figs. 6a–c). Among these dams, 23 showed at least one shift in the timing of their annual low inflows, with four dams experiencing two shifts each. For example, at the Currant Creek Dam (Upper Colorado region), 60% of the low inflows occurred in the fall during the period of 1993–2002, declining to 40% in the winter during the period of 2003–12 and increasing to 80% in the fall during the last decade (2013–22). Joes Valley Dam (also in the Upper Colorado region) exhibited similar shifts, with 50% of low inflows in the fall in the first decade (1993–2002), increasing to 90% in the winter in the next decade (2003–12) and returning to 50% in the fall over the last decade. Upper Stillwater Dam showed a shift from 100% in the spring in the first decade (1993–2002) to 40% in the winter in the next decade (2003–12) and then to 70% in the spring over the last decade. Deer Creek Dam’s low inflows changed from 40% in the winter (1993–2002) to 30% in the next decade, with fall inflows increasing to 40% and then winter inflows rising again to 40% in the last decade.
Decadal proportion of seasons with (a)–(c) annual low and (d)–(f) annual high dam inflows during the period of 1993–2022.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
For high inflows, which primarily occurred in fall and winter, there was a noticeable timing shift; the proportion of high inflows in the fall decreased, while the proportion in the spring slightly increased (Figs. 6d–f). Among these dams, 15 experienced at least one shift in the season of high inflow occurrence, with five dams undergoing two shifts. For instance, at the Stateline Dam (Upper Colorado region), 70% of high inflows occurred in the summer during the first decade (1993–2002), shifting to 70% in the spring in the second decade, and back to 60% in the summer in the last decade (2013–22). Similar patterns were observed at Friant (California region) and Whiskeytown (California region) dams, with shifts between summer and spring. Willow Creek Dam (Missouri region) shifted from 60% of high inflows in the summer over the first decade to 80% in the spring in the second decade and then to 50% in the fall over the last decade. At Huntington North Dam (Upper Colorado region), high inflows changed from 50% in spring in the first decade to 20% over the next decade, with winter inflows increasing to 50%, and then a complete shift to 100% in the summer over the last decade. These shifts highlight the dynamic nature of inflow patterns across different hydrologic regions and seasons.
5. Discussion
a. Relationships between the trends of dam inflow and the driving factors
We examined the dams displaying identical or opposing trends in dam inflows and the four driving factors. Table 1 presents the proportion and number of dams with these aligned or contrasting trends. The number of dams with statistically significant upward trends of dam inflows was limited. Due to this limitation, it is challenging to establish a relationship between increasing dam inflows and increasing trends in any of the driving factors (precipitation or soil moisture). Furthermore, we found no clear relationship between the dam inflow and air temperature. Similarly, our analyses did not reveal a clear connection between dam inflow patterns and the NDVI.
Proportion and number of dams with the same trend of dam inflow and the driving factor—precipitation, air temperature, soil moisture, and NDVI. The inconsistency in the numbers of dams in different rows (different time scales and trend direction) is because the number in each row indicates only the dams with statistically significant trends. The bold text shows the proportions exceeding 30%.
Unlike the upward trends that were found in only a few dams and could not be explained by the four driving factors, downward trends were observed in more dams and more consistent trends with the driving factors were evident. The downward trend in precipitation was consistent with the downward trend in dam inflows. At the annual time scale, 50% of the dams with a decreasing trend in inflow also experienced a downward trend in precipitation. Similarly, in the fall season, 32% of the dams showed this consistency. However, this proportion was relatively lower for the other seasons (0%–26%).
Air temperature was found as an important factor for the trend of dam inflows. The matching proportion between dam inflow and air temperature was high at the annual time scale (73%) as well as summer and fall seasons (100% and 88%). Figures 7a–c highlight the locations of dams where air temperature and dam inflow trends are the opposite during these time scales. All dams during the summer season exhibited a significant upward trend in the air temperature, as illustrated in Fig. 7b. In this case, the increase in air temperature likely led to an increased evaporation, leading to a decrease in dam inflows. Specifically, within the California region, all dams experienced both an increasing trend in air temperature and a decreasing trend in dam inflow on an annual time scale (Fig. 7a). In contrast, during the fall season, dams with opposing trends were primarily found in the Upper Colorado region (Fig. 7c).
Locations of dams where air temperature trends are opposite to dam inflow trends and where precipitation and soil moisture trends matched with dam inflow trends (green circles). The white circles represent all other cases.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
However, in the winter and spring seasons, there were no clear relationships between the trends of air temperature and dam inflow. This finding can be attributed to the impact of mountain snowpack in the western United States. Huning and AghaKouchak (2018) showed that, in the western United States, warmer temperatures lead to a decrease of snow water equivalent volume in April. The authors showed that a 1.0°–2.0°C increase in the average temperature results in a 20%–40% higher likelihood of below-average snow water equivalent. Huning and AghaKouchak (2020) found that snow droughts have become more prevalent, intensified, and prolonged in the western United States since 1980. Consequently, reduced mountain snowpack due to warmer temperatures and snow droughts can prevent streamflow discharges from being stored as snow, leading to higher runoffs during winter and spring and low flows later. In this context, warmer temperatures in winter and spring might have a complex relationship and opposing impacts on dam inflows. On the one hand, higher temperatures can increase flows due to faster snowmelt (assuming availability of snow). On the other hand, higher atmospheric water demand and water loss can result in decreasing flows.
Interestingly, the trend of soil moisture exhibited a clearer relationship with dam inflows in the spring and winter compared to the two climatic variables. Soil moisture trends matched dam inflow trends in 51% of the dams at the annual time scale, 29% in spring, and 45% in winter. These suggested that soil moisture is important for understanding downward trends of dam inflows during the winter and spring where precipitation and air temperature do not provide any consistent relationships. In Figs. 7d and 7e, we see that the number of dams where winter air temperature and precipitation match dam inflow trends is very limited and confined to the California region. On the other hand, Fig. 7f underscores the importance of soil moisture in understanding dam inflow trends. It shows that soil moisture trends are very useful for interpreting dam inflow patterns in the Upper Colorado region and the Great Basin region, especially since there are no matching cases between precipitation or air temperature trends and dam inflow in these areas.
In the case of NDVI, the proportion of matching dam inflow trends was high across all seasons, with an annual time scale match as high as 90%. Despite the high correlation between NDVI and dam inflow trends across all seasons, the overall increase in NDVI complicates its use as a reliable indicator of dam inflow trends. Even dams without significant inflow trends show an increase in NDVI, making it challenging to attribute the trends of inflows with vegetation.
b. Patterns of changepoints
We found that 22 dam inflows exhibited nonmonotonic trends, inconsistent to their statistically significant monotonic trends. Among them, 10 showed nonmonotonic trends, in which the trends changed from downward to upward and 12 in the opposite direction. Figure 8 shows the year of changepoints of those 22 dam inflows.
Strip plot depicting the changepoints in trends for 22 dam inflows, categorized by type of inflow.
Citation: Journal of Hydrometeorology 25, 12; 10.1175/JHM-D-23-0217.1
In Fig. 8, the changepoints for the nonmonotonic trends were mostly identified in the second study decade (2002, 2003, 2006, 2007, 2008, 2009, 2010, and 2011). Notably, more than half of these trends were identified in 2011 across seven dams. Of these, three dams—Joes Valley, Moon, and Currant Creek—are in the Upper Colorado region, three—Spring Creek, Trinity, and Bradbury—are in the California region and one—Wanship—is in the Great Basin region. The three dams in the Upper Colorado region exhibited a consistent change in trend direction, shifting from an upward to a downward. Similarly, in the California region, two dams—Bradbury and Huntington North—displayed a uniform change in trend direction from a downward to an upward.
To explain the substantial number of changepoints in 2011, we explored the extreme events that occurred in the Upper Colorado and California regions then. According to the U.S. Drought Monitor’s report, 90% of western region, including Colorado, Wyoming, and Utah states, were under drought conditions in 2011 (David and Ahira 2021). Additionally, approximately 54% of the western United States experienced extreme or exceptional classes of drought (David and Ahira 2021). These droughts may have reduced inflows to the Upper Colorado region in the 2020s, resulting in a downward trend after 2011.
The drought in California over the period of 2012–16 was one of the most severe in the state history [California Natural Resources Agency (CNRA) 2021]. California also experienced its driest four consecutive years ever recorded from 2012 to 2015, with 2014 being the third driest year of the entire history (CNRA 2021). Additionally, California Department of Water Resources (2018) reported that 2017 was a year of devastating floods for parts of the state of California, especially in the northern part, which experienced one of its wettest winters in the past century. It is reasonable to assume that those severe droughts from 2012 to 2016 and floods in 2017 resulted in upward trends for dams in the California region after 2011.
Furthermore, we analyzed the nonmonotonic trends of the dam inflow and the four driving factors. Table 2 lists the dams with the detected changepoints, along with the corresponding trends in precipitation, air temperature, soil moisture, and NDVI.
Dams with statistically significant changepoints and the trend of hydrometeorological variables and NDVI.
In Table 2, our analyses revealed that only two changepoints occurred during the fall season, with the majority being in the winter. Additionally, a significant relationship was observed between the trends soil moisture and dam inflow. At the annual time scale as well as during spring and summer seasons, 80% of the dams exhibited a downward trend in soil moisture. During the summer season, all dams with detected changepoints also showed an increasing trend in air temperature. At the annual time scale, Bradbury, Joes Valley, and Huntington North Dams demonstrated significant trends across the four driving factors analyzed. Comparatively, spring exhibited fewer consistent trends across these factors. This suggested the need for future research focused on investigating these seasonal changes in spring, potentially utilizing additional factors to explain the dam inflow trends.
c. Limitations and directions for future research
We used daily data for the trend analyses of dam inflows. For low and high inflows, the analyses should be done using subdaily data to capture the intradaily variability where minimum and maximum inflows occur (instantaneous time scale). However, such data for dams are not typically available to the public due to security reasons and may hinder detailed analyses.
To explain the reasons of dam inflow trends, we studied four factors of precipitation, air temperature, soil moisture, and NDVI. Other factors, such as groundwater, snow water equivalent, and surface imperviousness, can provide additional insights about the causes of the dam inflow trends and changepoints.
Another limitation of our study is that we did not assess the potential interdependencies among the dams or analyze their interactions within a connected river. Dams located on the same river might exhibit correlated behaviors due to shared catchment characteristics and weather patterns. This is especially true for rivers with substantial streamflow, where multiple upstream dams can significantly influence the operation of the downstream dam. To address these potential interdependencies, we excluded dams directly connected to upstream dams sharing a single river GNID as defined in the NHD. While this approach helped filter out some complexities, it underscores the need for further analyses to better understand how interconnected dams may influence each other.
We also focused solely on historical trends of dam inflows. It is anticipated that the inflows are affected by nonstationary stressors such as climate change. Our methods can be applied to study the trends of projected future inflows. Comparative analyses of future inflows against historical data can reveal the extent of future changes and the need for climate adaptation strategies. In addition to nonstationary climate changes, stationary changes can be analyzed using stochastic simulations such as those developed by Brunner and Gilleland (2020).
For future research, a detailed analysis of how extreme weather events, such as heavy precipitation and prolonged droughts, influence nonmonotonic dam inflow trends is essential, particularly in regions like California that exhibited this pattern. By systematically removing or analyzing separately the inflow data during extreme events from the overall 30-yr record, the underlying monotonic trends can potentially be isolated. This approach can provide a clearer assessment of whether the observed nonmonotonic trend is mainly due to these extreme events imposed on longer-term monotonic patterns, or if the inflows are inherently exhibiting more complex, multidirectional shifts.
6. Summary and conclusions
This paper explored temporal changes in magnitude and seasonality of low, median, and high inflows of 51 dams in the western and southwestern United States. Low, median, and high inflows were calculated by selecting the 5th, 50th, and 95th of the daily inflow data. The temporal changes of precipitation, air temperature, soil moisture, and NDVI were also examined for evaluating dam inflow trends. We applied monotonic trend analysis using the Mann–Kendall approach, nonmonotonic trend tests using the PELT algorithm, and changepoint detection analyses. We provide a framework for understanding how dam inflow trends respond to various climatic and hydrometeorological factors.
The major findings of this study can be summarized as follows:
-
From 1993 to 2022, our trend analyses revealed a widespread downward trend in dam inflows, particularly across the Upper Colorado and California regions. More than 30% of dams showed a significant decrease in their annual median inflows. This trend was particularly pronounced during the spring and fall seasons.
-
At the annual time scale, the overall downward trend in dam inflows was consistent with precipitation. Increasing air temperatures contributed to declining annual, summer, and fall inflows, likely due to higher evaporation rates. Soil moisture demonstrated an even stronger connection to the downward trends of inflows, especially in the winter and spring. Although there was a strong connection between NDVI and dam inflow trends across all seasons, the general increase in NDVI complicated any attribution to the dam inflow trends.
-
The first two major findings suggest that the overall downward trend in dam inflows can be explained by a combination of precipitation, soil moisture, and air temperature. Each variable matched differently with seasonal inflow trends, so using a combination of these factors was beneficial. Specifically, soil moisture is crucial for understanding the downward trends in dam inflows during the winter and spring seasons, where precipitation and air temperature do not provide consistent explanations.
-
Our results revealed notable shifts in seasonality patterns for both low and high inflows. Specifically, we observed an increase in the proportion of summer and fall inflows and a decrease in the proportion of winter inflows for low inflow events. Conversely, for high inflows, there was an increase in the proportion of spring inflows and a decrease in the proportion of fall inflows. These results highlighted notable changes in the timing and distribution of inflows over the study period.
-
Our changepoint analyses detected nonmonotonic trends between 2002 and 2012 in about 13% of the dams, in which the majority were located in the Upper Colorado and California regions. More than half of these changepoints were detected in year 2011, likely due to widespread droughts then across Upper Colorado and California regions.
Our findings emphasized the importance of understanding the relationships between the dam inflow and four driving factors to enable informed decision-making for dam operations in long term. Previous studies found that considering the temporal trends can improve the reservoir inflow forecasting, particularly for long timeframes (Ndione et al. 2020; Jones and Hammond 2020; Pishgah Hadiyan et al. 2022; Talukdar et al. 2023). That said, our study can help improve the performance of inflow forecast models. By incorporating the observed relationships between hydrometeorological variables, vegetation, and dam inflows into the forecast models, managers can achieve more accurate predictions of future inflow patterns. Unfavorable long-term trends of dam inflows, which can be attributed to the four factors investigated in this paper, can be potentially handled by managing anthropogenic factors like vegetation loss due to land use change.
Furthermore, the detected relationships between the trend of dam inflow and the four driving factors can be used to help dam operators plan for upcoming weather conditions and plausible changes in soil moisture and vegetation. As an example, for the dams with the consistent trend between the inflow and air temperature, if seasonal forecasts upstream of a given dam show abnormally warm weathers over the coming summer, adjustments can be made in the reservoir operation to address the water needs related to the applicable dam operation purposes (e.g., water supply). This can be critical as abnormal soil moisture deficit and more frequent warmer weathers are expected under climate change (Gergel et al. 2017; Hallema et al. 2017; Ban et al. 2020; McEvoy and Hatchett 2023) and air temperature and soil moisture were found to be key for dam inflow trends in this study. Our findings can, thus, support dam operators build climate resilience into the dams across the study area, while our methods present a framework for supporting similar actions on other regulating infrastructure and across other geographic regions.
Data availability statement.
Dam inflow data provided by USBR’s Reclamation Information Sharing Environment are available at https://data.usbr.gov/catalog. Precipitation, air temperature, and soil moisture data provided by ERA5-Land are available at https://cds.climate.copernicus.eu/datasets/reanalysis-era5-land-monthly-means?tab=download. NDVI data from the AVHRR and the VIIRS provided by NOAA are available at https://www.ncei.noaa.gov/data/land-normalized-difference-vegetation-index/access/.
REFERENCES
Ahmadisharaf, E., and A. J. Kalyanapu, 2015: Investigation of the impact of streamflow temporal variation on dam overtopping risk: Case study of a high-hazard dam. World Environmental and Water Resources Congress 2015, Austin, TX, 1050–1057, https://doi.org/10.1061/9780784479162.103.
Altinbilek, D., 2002: The role of dams in development. Water Sci. Technol., 45, 169–180, https://doi.org/10.2166/wst.2002.0172.
Asarian, J. E., and J. D. Walker, 2016: Long-term trends in streamflow and precipitation in Northwest California and Southwest Oregon, 1953-2012. J. Amer. Water Resour. Assoc., 52, 241–261, https://doi.org/10.1111/1752-1688.12381.
Babur, M., M. S. Babel, S. Shrestha, A. Kawasaki, and N. K. Tripathi, 2016: Assessment of climate change impact on reservoir inflows using multi climate-models under RCPs—The case of Mangla dam in Pakistan. Water, 8, 389, https://doi.org/10.3390/w8090389.
Ban, Z., T. Das, D. Cayan, M. Xiao, and D. P. Lettenmaier, 2020: Understanding the asymmetry of annual streamflow responses to seasonal warming in the Western United States. Water Resour. Res., 56, e2020WR027158, https://doi.org/10.1029/2020WR027158.
Brunner, M. I., and E. Gilleland, 2020: Stochastic simulation of streamflow and spatial extremes: A continuous, wavelet-based approach. Hydrol. Earth Syst. Sci., 24, 3967–3982, https://doi.org/10.5194/hess-24-3967-2020.
Burn, D. H., and M. A. H. Elnur, 2002: Detection of hydrologic trends and variability. J. Hydrol., 255, 107–122, https://doi.org/10.1016/S0022-1694(01)00514-5.
California Department of Water Resources, 2018: HYDROCLIMATE REPORT Water Year 2017. California Department of Water Resources Rep., 32 pp., https://water.ca.gov/-/media/DWR-Website/Web-Pages/Programs/Flood-Management/Flood-Data/Climate-summaries/Hydroclimate_Report_2017-ADA-Final.pdf.
Chen, X., and F. Hossain, 2019: Understanding future safety of dams in a changing climate. Bull. Amer. Meteor. Soc., 100, 1395–1404, https://doi.org/10.1175/BAMS-D-17-0150.1.
CNRA, 2021: Report to the legislature on the 2012–2016 drought. California Natural Resources Agency Rep., 74 pp., https://water.ca.gov/-/media/DWR-Website/Web-Pages/Water-Basics/Drought/Files/Publications-And-Reports/CNRA-Drought-Report-final-March-2021.pdf.
David, S., and S. L. Ahira, 2021: National drought summary for August 31, 2021. National Drought Mitigation Center Rep., https://goldrushcam.com/sierrasuntimes/index.php/news/local-news/32982-california-and-national-drought-summary-for-august-31-2021-10-day-weather-outlook-and-california-drought-statistics-88-in-extreme-drought.
Fathian, F., Z. Dehghan, M. H. Bazrkar, and S. Eslamian, 2016: Trends in hydrological and climatic variables affected by four variations of the Mann-Kendall approach in Urmia Lake basin, Iran. Hydrol. Sci. J., 61, 892–904, https://doi.org/10.1080/02626667.2014.932911.
Fleming, S. W., and F. A. Weber, 2012: Detection of long-term change in hydroelectric reservoir inflows: Bridging theory and practise. J. Hydrol., 470–471, 36–54, https://doi.org/10.1016/j.jhydrol.2012.08.008.
Gergel, D. R., B. Nijssen, J. T. Abatzoglou, D. P. Lettenmaier, and M. R. Stumbaugh, 2017: Effects of climate change on snowpack and fire potential in the western USA. Climatic Change, 141, 287–299, https://doi.org/10.1007/s10584-017-1899-y.
Getachew, H. E., and A. M. Melesse, 2012: The impact of land use change on the hydrology of the Angereb watershed, Ethiopia. Int. J. Water Sci., 1, 4.
Gosiewska, A., A. Kozak, and P. Biecek, 2021: Simpler is better: Lifting interpretability-performance trade-off via automated feature engineering. Decision Support Syst., 150, 113 556, https://doi.org/10.1016/j.dss.2021.113556.
Hallema, D. W., G. Sun, K. D. Bladon, S. P. Norman, P. V. Caldwell, Y. Liu, and S. G. McNulty, 2017: Regional patterns of postwildfire streamflow response in the Western United States: The importance of scale-specific connectivity. Hydrol. Processes, 31, 2582–2598, https://doi.org/10.1002/hyp.11208.
Hamed, K. H., 2008: Trend detection in hydrologic data: The Mann–Kendall trend test under the scaling hypothesis. J. Hydrol., 349, 350–363, https://doi.org/10.1016/j.jhydrol.2007.11.009.
Hamed, K. H., and A. R. Rao, 1998: A modified Mann-Kendall trend test for autocorrelated data. J. Hydrol., 204, 182–196, https://doi.org/10.1016/S0022-1694(97)00125-X.
Hecht, J. S., G. Lacombe, M. E. Arias, T. D. Dang, and T. Piman, 2019: Hydropower dams of the Mekong River basin: A review of their hydrological impacts. J. Hydrol., 568, 285–300, https://doi.org/10.1016/j.jhydrol.2018.10.045.
Ho, M., U. Lall, M. Allaire, N. Devineni, H. H. Kwon, I. Pal, D. Raff, and D. Wegner, 2017: The future role of dams in the United States of America. Water Resour. Res., 53, 982–998, https://doi.org/10.1002/2016WR019905.
Huning, L. S., and A. AghaKouchak, 2018: Mountain snowpack response to different levels of warming. Proc. Natl. Acad. Sci. USA, 115, 10 932–10 937, https://doi.org/10.1073/pnas.1805953115.
Huning, L. S., and A. AghaKouchak, 2020: Global snow drought hot spots and characteristics. Proc. Natl. Acad. Sci. USA, 117, 19 753–19 759, https://doi.org/10.1073/pnas.1915921117.
Hussain, M., and I. Mahmud, 2019: pyMannKendall: A python package for non parametric Mann Kendall family of trend tests. J. Open Source Software, 4, 1556, https://doi.org/10.21105/joss.01556.
Jegede, S. L., and K. J. Szajowski, 2022: Change-point detection in homogeneous segments of COVID-19 daily infection. Axioms, 11, 213, https://doi.org/10.3390/axioms11050213.
Jones, J. A., and J. C. Hammond, 2020: River management response to multi-decade changes in timing of reservoir inflows, Columbia River basin, USA. Hydrol. Processes, 34, 4814–4830, https://doi.org/10.1002/hyp.13910.
Kalra, A., S. Sagarika, P. Pathak, and S. Ahmad, 2017: Hydro-climatological changes in the Colorado River basin over a century. Hydrol. Sci. J., 62, 2280–2296, https://doi.org/10.1080/02626667.2017.1372855.
Kendall, M. G., 1975: Rank Correlation Methods. 4th ed. Charles Griffin, 202 pp.
Khalkheili, T. A., and G. H. Zamani, 2009: Farmer participation in irrigation management: The case of Doroodzan Dam Irrigation Network, Iran. Agric. Water Manage., 96, 859–865, https://doi.org/10.1016/j.agwat.2008.11.008.
Killick, R., P. Fearnhead, and I. A. Eckley, 2012: Optimal detection of changepoints with a linear computational cost. J. Amer. Stat. Assoc., 107, 1590–1598, https://doi.org/10.1080/01621459.2012.737745.
Kumambala, P. G., and A. Ervine, 2009: Site selection for combine hydro, irrigation and water supply in Malawi: Assessment of water resource availability. Desalination, 248, 537–545, https://doi.org/10.1016/j.desal.2008.05.099.
López-Moreno, J. I., and Coauthors, 2014: Impact of climate and land use change on water availability and reservoir management: Scenarios in the Upper Aragón River, Spanish Pyrenees. Sci. Total Environ., 493, 1222–1231, https://doi.org/10.1016/j.scitotenv.2013.09.031.
Loukas, A., N. Mylopoulos, and L. Vasiliades, 2007: A modeling system for the evaluation of water resources management strategies in Thessaly, Greece. Water Resour. Manage., 21, 1673–1702, https://doi.org/10.1007/s11269-006-9120-5.
Mann, H. B., 1945: Nonparametric tests against trend. Econometrica, 13, 245–259, https://doi.org/10.2307/1907187.
McCabe, G. J., D. M. Wolock, G. T. Pederson, C. A. Woodhouse, and S. McAfee, 2017: Evidence that recent warming is reducing upper Colorado River flows. Earth Interact., 21, https://doi.org/10.1175/EI-D-17-0007.1.
McEvoy, D. J., and B. J. Hatchett, 2023: Spring heat waves drive record western United States snow melt in 2021. Environ. Res. Lett., 18, 014007, https://doi.org/10.1088/1748-9326/aca8bd.
Miller, W. P., and T. C. Piechota, 2011: Trends in Western U.S. snowpack and related upper Colorado River basin streamflow. J. Amer. Water Resour. Assoc., 47, 1197–1210, https://doi.org/10.1111/j.1752-1688.2011.00565.x.
Monteiro-Alves, R., M. Á. Toledo, R. Moran, and L. Balairón, 2022: Failure of the downstream shoulder of Rockfill dams due to overtopping or throughflow. Water, 14, 1624, https://doi.org/10.3390/w14101624.
Muhammad, A., G. R. Evenson, F. Unduche, and T. A. Stadnyk, 2020: Climate change impacts on reservoir inflow in the Prairie pothole region: A watershed model analysis. Water, 12, 271, https://doi.org/10.3390/w12010271.
Muñoz Sabater, J., 2019: ERA5-Land monthly averaged data from 1950 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS), 5 July 2023, https://doi.org/10.24381/cds.68d2bb30.
Naz, B. S., S.-C. Kao, M. Ashfaq, H. Gao, D. Rastogi, and S. Gangrade, 2018: Effects of climate change on streamflow extremes and implications for reservoir inflow in the United States. J. Hydrol., 556, 359–370, https://doi.org/10.1016/j.jhydrol.2017.11.027.
Ndione, D. M., S. Sambou, S. Kane, S. Diatta, M. L. Sane, and I. Leye, 2020: Ensemble forecasting system for the management of the Senegal River discharge: Application upstream the Manantali dam. Appl. Water Sci., 10, 126, https://doi.org/10.1007/s13201-020-01199-y.
Nguyen-Tien, V., R. J. R. Elliott, and E. A. Strobl, 2018: Hydropower generation, flood control and dam cascades: A national assessment for Vietnam. J. Hydrol., 560, 109–126, https://doi.org/10.1016/j.jhydrol.2018.02.063.
Notaro, M., K. Emmett, and D. O’Leary, 2019: Spatio-temporal variability in remotely sensed vegetation greenness across Yellowstone National Park. Remote Sens., 11, 798, https://doi.org/10.3390/rs11070798.
Perera, D., V. Smakhtin, S. Williams, T. North, and A. Curry, 2021: Ageing water storage infrastructure: An emerging global risk. UNU-INWEH Rep. Series, Vol. 1, 31 pp., http://collections.unu.edu/view/UNU:7930.
Pishgah Hadiyan, P., R. Moeini, E. Ehsanzadeh, and M. Karvanpour, 2022: Trend analysis of water inflow into the dam reservoirs under future conditions predicted by dynamic NAR and NARX models. Water Resour. Manage., 36, 2703–2723, https://doi.org/10.1007/s11269-022-03170-9.
Potter, C., 2019: Changes in vegetation cover of Yellowstone National Park estimated from MODIS greenness trends, 2000 to 2018. Remote Sens. Earth Syst. Sci., 2, 147–160, https://doi.org/10.1007/s41976-019-00019-5.
Talukdar, S., S. Pal, Shahfahad, M. W. Naikoo, A. Parvez, and A. Rahman, 2023: Trend analysis and forecasting of streamflow using random forest in the Punarbhaba River basin. Environ. Monit. Assess., 195, 153, https://doi.org/10.1007/s10661-022-10696-3.
Tatlhego, M., D. D. Chiarelli, M. C. Rulli, and P. D’Odorico, 2022: The value generated by irrigation in the command areas of new agricultural dams in Africa. Agric. Water Manage., 264, 107517, https://doi.org/10.1016/j.agwat.2022.107517.
Tosunoglu, F., and O. Kisi, 2017: Trend analysis of maximum hydrologic drought variables using Mann–Kendall and Şen’s innovative trend method. River Res. Appl., 33, 597–610, https://doi.org/10.1002/rra.3106.
Touzani, S., B. Ravache, E. Crowe, and J. Granderson, 2019: Statistical change detection of building energy consumption: Applications to savings estimation. Energy Build., 185, 123–136, https://doi.org/10.1016/j.enbuild.2018.12.020.
United States Army Corps of Engineers, 2021: National inventory of dams. US Army Corps of Engineers, https://nid.sec.usace.army.mil/#/.
USBR, 2020: Reclamation Information Sharing Environment (RISE). US Department of the Interior, accessed 7 May 2023, https://data.cnra.ca.gov/dataset/reclamation-information-sharing-environment-rise.
USGS, 2004: National hydrography dataset. United States Geological Survey, accessed 29 March 2023, https://www.usgs.gov/national-hydrography/national-hydrography-dataset.
USGS, 2022: 3D Elevation Program 1 arc-second Digital Elevation Models. United States Geological Survey, accessed 4 April 2024, https://www.usgs.gov/3d-elevation-program.
Vermote, E., 2019: NOAA CDR Program NOAA Climate Data Record (CDR) of AVHRR Normalized Difference Vegetation Index (NDVI), version 5. NOAA National Centers for Environmental Information, accessed 2 June 2024, https://catalog.data.gov/dataset/noaa-climate-data-record-cdr-of-avhrr-normalized-difference-vegetation-index-ndvi-version-52.
Vermote, E., 2022: NOAA CDR Program NOAA Climate Data Record (CDR) of VIIRS Normalized Difference Vegetation Index (NDVI), version 1. NOAA National Centers for Environmental Information, accessed 2 June 2024, https://www.geoplatform.gov/metadata/a33d7b15-0c45-4af6-8d86-4ef68bf0b88e.
Wang, F., W. Shao, H. Yu, G. Kan, X. He, D. Zhang, M. Ren, and G. Wang, 2020: Re-evaluation of the power of the Mann-Kendall test for detecting monotonic trends in hydrometeorological time series. Front. Earth Sci., 8, 14, https://doi.org/10.3389/feart.2020.00014.
Whitehead, P. G., L. Jin, I. Macadam, T. Janes, S. Sarkar, H. J. E. Rodda, R. Sinha, and R. J. Nicholls, 2018: Modelling impacts of climate change and socio-economic change on the Ganga, Brahmaputra, Meghna, Hooghly and Mahanadi River systems in India and Bangladesh. Sci. Total Environ., 636, 1362–1372, https://doi.org/10.1016/j.scitotenv.2018.04.362.
Wright, T. E., J. Stuivenvolt-Allen, G. Affram, N. A. Hasan, C. Ratterman, and W. Zhang, 2023: Responses of extreme discharge to changes in surface-air and dewpoint temperatures in Utah: Seasonality and mechanisms. Water, 15, 688, https://doi.org/10.3390/w15040688.
Xiao, M., B. Udall, and D. P. Lettenmaier, 2018: On the causes of declining Colorado River streamflows. Water Resour. Res., 54, 6739–6756, https://doi.org/10.1029/2018WR023153.
Xie, Y., D. Kulpanowski, J. Ong, E. Nikolova, and N. M. Tran, 2021: Predicting Covid‐19 emergency medical service incidents from daily hospitalisation trends. Int. J. Clin. Pract., 75, e14920, https://doi.org/10.1111/ijcp.14920.
Yao, H., Z. Dong, W. Jia, X. Ni, M. Chen, C. Zhu, and D. Li, 2019: Competitive relationship between flood control and power generation with flood season division: A case study in downstream Jinsha River cascade reservoirs. Water, 11, 2401, https://doi.org/10.3390/w11112401.
Yu, M. M.-J., J.-H. Lee, M.-C. Lee, C.-P. Lee, and J. R.-C. Hsu, 2014: Effect of wave dissipators on overtopping over a bermed seawall in deepwater. Coastal Eng. Proc., 1, 8, https://doi.org/10.9753/icce.v34.structures.8.
Yue, S., and C. Wang, 2004: The Mann-Kendall test modified by effective sample size to detect trend in serially correlated hydrological series. Water Resour. Manage., 18, 201–218, https://doi.org/10.1023/B:WARM.0000043140.61082.60.
Yue, S., P. Pilon, and G. Cavadias, 2002: Power of the Mann–Kendall and Spearman’s rho tests for detecting monotonic trends in hydrological series. J. Hydrol., 259, 254–271, https://doi.org/10.1016/S0022-1694(01)00594-7.
Zhang, N. R., and D. O. Siegmund, 2007: A modified Bayes information criterion with applications to the analysis of comparative genomic hybridization data. Biometrics, 63, 22–32, https://doi.org/10.1111/j.1541-0420.2006.00662.x.