Collection of discharge data for the Okanogan River at Oroville, Wash., U.S. Geological Survey (USGS) gaging station (USGS 12439500) (USGS, 2024a) began in 1942 , and the station is herein referred to as ORO–WA. Approved discharge records have been available since 1942; additional paper records from before 1942 are archived at the USGS Upper Columbia Field Office. The gaging station is located 2.8 km downstream from the outlet of Osoyoos Lake and 0.3-km downstream from Zosel Dam (fig. 1). Discharge data from the USGS gaging station at Similkameen River near Nighthawk, Wash. (USGS 12442500) (USGS, 2024c), which flows into the Okanogan River approximately 4-km downstream of the Osoyoos Lake outlet, were used to examine potential effects on recorded discharge at the ORO–WA gaging station caused by high flows on the Similkameen River that can produce backwater effects at ORO–WA.

The Osoyoos Lake near Oroville, Wash., gaging station (USGS 12439000) (USGS, 2024b) is located approximately 1.1-km northwest (upstream) of the lake outlet. Data collection, approval, and publication at USGS 12439000, hereafter referred to as OLO–WA, began in 1965. However, the USGS began collecting Osoyoos Lake level (gage height) data at the site in 1928, but these early data were not approved or published. The Government of Canada published the 1928–1965 USGS-collected Osoyoos Lake level data, which are publicly accessible (Government of Canada, 2025). Those data were quality assured by comparison with the USGS archived working record. Researchers resolved discrepancies by removing all data prior to the start of the 1930 water year (October 1, 1929).

Air temperature data were compiled from four meteorological stations (table 1; fig. 1), extending the period of available data to 1942. Wind speed and direction data were not available until 1994, after Environment and Climate Change Canada (ECCC) started operating the Osoyoos CS meteorological station (Government of Canada, 2024a).¹ 1

Data collected by Environment and Climate Change Canada are cited as sourced by the Government of Canada, as these data are not attributed at the organizational level.

All satellite imagery from Landsat 8–9 (USGS, 2024d) and Sentinel–2 (Sinergise Solutions d.o.o., 2024) for February 11, 2013–March 31, 2024, was examined for visual evidence of ice-jam occurrences and to view ice at various stages of formation and breakup. The nominal revisit times over Washington State for Landsat 8–9 and Sentinel–2 are approximately 8 days and 2–3 days, respectively, but winter cloud cover over Osoyoos Lake severely reduces the number of usable images. A collection of cloud-free available ice images (appendix 1) was compiled for February 11, 2013–March 31, 2024, after a thorough review of the online databases (Sinergise Solutions d.o.o., 2024; USGS, 2024d). An increase in the resolution of satellite imagery began in 2013 with the launch of Landsat 8–9, which enabled visual verification of the occurrence of ice jams and observation of the characteristics of ice. When possible, satellite imagery was used to examine ice morphology on the lake and verify potential ice jams delineated using changes in the discharge rate on the Okanogan River, the lake level of Osoyoos Lake, and the air temperature in the area.

Because the most prominent adverse effect of ice jams on Osoyoos Lake is a decreased discharge downstream, these data, along with air temperature and lake level, were used to identify days when ice jams occurred. We examined changes in the daily mean and in moving windows of 2-, 3-, and 4-day moving averages to identify periods indicative of persistent ice jams that could cause substantial changes to the lake level and streamflow. Ice-jam occurrences were identified through discharge data from ORO–WA, lake level data at OLO–WA, and air temperature data from four weather stations near Osoyoos Lake (table 1; fig. 2).

Additional factors to consider when identifying ice-jam occurrences on Osoyoos Lake include (1) operational changes in discharge and (2) the impact of a backwater effect that sometimes results in negative discharge at the ORO–WA gaging station caused by high flows on the Similkameen River. Air temperature data were used as a first check to filter out decreases in discharges associated with water operational changes because effects on discharges associated with ice jams would not occur when temperatures were above freezing. Because high flows on the Similkameen River are usually caused by rapid snow melt or heavy rainfall, backwater that may affect flow depth and discharge at ORO–WA is unlikely to occur when it is cold enough that precipitation falls as snow and ice jams form on Osoyoos Lake. However, discharge data from the Similkameen River near Nighthawk gaging station (USGS 12442500) were used to verify that the timing of high-flow conditions on the Similkameen River did not coincide with the timing of potential ice jams.

Although the USGS hydrologic stations ORO–WA and OLO–WA currently collect water temperature data, temperature data collection at ORO–WA began in 2005 and at OLO–WA in 2012. Because this limited period of record restricts the timeframe for analysis, the correlation between water and air temperature as a potential proxy for relative daily mean variations in water temperature was examined. Daily mean air temperature at the Osoyoos CS meteorological station shows a Spearman correlation (r_s) and root mean square error (RMSE) with (1) the daily mean water temperature at OLO–WA of r_s=0.93 (probability value [p] p < 0.0001) and RMSE=0.86 and (2) the daily mean water temperature at ORO–WA of r_s=0.93 (p< 0.0001) and RMSE=0.92. Thus, observed relative changes in air temperature were used as a proxy for relative changes in water temperature at this study site.

Ice jams were identified for the period from October 1, 1942, to March 31, 2024, based on available discharge, lake level, and air temperature data. Daily minimum and daily mean air temperature data from the Osoyoos CS meteorological station, beginning August 1, 1990, when the record began, were used. Daily air temperature data from before August 1, 1990, were taken from Osoyoos West (Government of Canada, 2024b), Oroville (National Centers for Environmental Information [NCEI], 2024a), and Oroville 3 NW (NCEI, 2024b) stations, where only maximum and minimum daily air temperature is continuously provided (table 1; fig. 1). A continuous air temperature record for the full period of the discharge data record was assembled using the closest available stations to Oroville (table 1). At stations where daily mean values were unavailable—before August 1, 1990—the midrange daily value (mean of the daily maximum and minimum) was used.

Short-lived disruptions in discharge, those on the order of hours, do not tend to cause large changes in the lake level. Lake level changes are also difficult to detect on time spans shorter than a day because hourly variability can be greatly affected by wind and waves. High winds in the region result in high uncertainty in instantaneous 15-minute lake-level data that exceed the 0.01-ft (0.3-cm) accuracy of measurements. The 2-day averages smooth out daily variability and allow comparisons within the accuracy of the measurements. These criteria exclude ice accumulations that are short-lived or cause smaller discharge decreases, but they also minimize false positives and emphasize the ice jams that are most consequential. Finally, because dewatering for periods less than a day does not pose as large a risk to all fish life stages, unlike multiday periods (Becker and others, 1983), the use of daily values is adequate for the detection of consequential ice jams.

Absolute change and percent change in daily, 2-, 3-, and 4-day averages of streamflow and lake level data were examined as indicators of ice-jam occurrences using observed ice-jam events verified with satellite imagery. In addition, various thresholds in maximum, mean, and minimum daily air temperature were examined, with the condition that 0 degrees Celsius (°C) or lower is a minimum requirement for ice-jam formation. The 2-day averages for (1) the absolute change in the average lake level at OLO–WA, (2) the percentage of change in streamflow data at ORO–WA, and (3) the minimum and mean air temperature threshold at Osoyoos CS or nearby weather stations from October to April each year were identified as the best indicators of verified ice-jam occurrences since 2019. The examination of daily, 2-, 3-, and 4-day moving averages indicated that averages longer than 2 days diluted the signal of changes in conditions associated with ice jams, whereas daily means produced false positives for ice-jam detection. Ice jams were identified on days when:

Using these criteria, historical ice-jam days were classified for October 1, 1942–March 31, 2024. The determination of potential ice-jam occurrences was restricted to periods when air temperature, river discharge, and lake level data were all available. Notable gaps in the records for when ice-jam occurrences were not assessed include—

In addition to the above-mentioned criteria, once an ice jam was identified, the following day was considered to have an ice-jam occurrence if (1) the 2-day average percentage change in discharge was less than or equal to 0, and (2) the 2-day average percentage change in lake level was greater than or equal to 0. These thresholds were selected by identifying known ice jams since 2019 and eliminating the false identification of non-ice-jam days. This conservative approach was appropriate based on the limitations of these data and the effects of Osoyoos Lake ice jams.

To examine environmental factors associated with ice-jam occurrences on Osoyoos Lake, we relied primarily on data from the ECCC climate station, Osoyoos CS. Historical records for wind speed and direction were available for 1994 through 2024. Because reliable wind data were unavailable before 1994, the analysis was limited to March 1, 1994, through March 31, 2024, covering October–April of each water year (table 1). Wind data for January 1, 2012, through July 17, 2012, were considered erroneous because of missing values, and all other wind speed and direction values from that period were listed as 0. For this reason, and because no days during that period were classified as ice-jam days, those dates were removed from the analysis.

Daily averages and 2-, 3-, and 4-day moving averages for wind direction, wind speed, minimum air temperature, and mean (midrange was used when daily mean was unavailable) air temperature were calculated and examined for trends associated with ice jams on Osoyoos Lake to assess the potential for predicting ice-jam occurrence. Averages were not calculated for days with missing data or nonconsecutive dates, and observations for those days were removed prior to additional analyses.

Various statistical methods were used to examine differences in air temperature and wind conditions associated with ice-jam and non-ice-jam days. A permutational multivariate analysis of variance (PERMANOVA) with 999 permutations was used to test for significant differences among multivariate predictors between ice-jam and non-ice-jam days using the “adonis” function of the “vegan: Community Ecology Package” in R statistical software (Oksanen and others, 2022; R Core Team, 2024). PCA was used to visualize the variability in 2-day averages of meteorological conditions associated with identified ice-jam days using the “PCA” function in the base package of R statistical software (R Core Team, 2024). The “lda” function in R was used for LDA (R Core Team, 2024) to examine the potential for predicting ice jams. The assumption of homoscedasticity for LDA was met for all predictor variables after taking the square root of wind speed. The highest Spearman rank correlation between all variables, including transformations, was r_s=0.58 (p-value< 0.0001) for wind speed and wind direction (fig. 3). The assumption of normality was relaxed and not formally tested because the dataset exceeded 5,000 observations (beyond the practical limit for some tests), and the central limit theorem suggests the data approach normality with large samples (Shapiro and Wilk, 1965). Histograms for each variable illustrate some skewness but otherwise approximate normality upon visual inspection (fig. 3). Thus, results from the PCA and LDA should be interpreted with caution regarding predictions, but these analyses provided usable information about environmental conditions linked to ice jam occurrence on Osoyoos Lake.

Available satellite images of ice on Osoyoos Lake (appendix 1) revealed apparent differences in lake ice conditions and showed potential differences in the mechanisms of ice movement that facilitate ice jams (fig. 4). Several images showed ice across most of the surface of the lake, with snow accumulated in potential cracks in the ice (figs. 4A, 4D, 4E). However, in these images, the outlet of Osoyoos Lake clearly showed a lack of ice, which supported the analysis because ice jams were not identified on those dates. Conversely, images that depicted ice at the constriction point of the lake outlet, extending a short distance down the channel from the southern banks of the lake, coincided with dates classified as ice-jam events (figs. 4C, 4F), except the image from February 12, 2018, in figure 4B.

Although ice was present at the outlet in figure 4B (February 12, 2018), this day was not identified as an ice-jam event because the 2-day average decrease in discharge of 10.9 percent did not exceed the threshold of 15 percent determined by calibration of known ice-jam events after 2019, as described in the “Methods” section. Mean daily discharge decreased to only 20.16  m³/s, with a minimum daily value of 9.43 m³/s. Although the 2-day average rise in the lake level of 0.6 cm met the threshold of 0.6 cm, the level started to decline, as did the effects on discharge the following day (February 13, 2018). This may have been a minor ice jam, but it did not persist long enough to affect the discharge or lake level substantially enough to meet the criteria of an ice-jam occurrence. In the calibration of the algorithm used to detect ice jams in the historical record, a threshold of a 10-percent decrease in 2-day average discharge falsely classified non-ice-jam days as ice-jam days. This instance provides an example of the justification used during the calibration of the algorithm.

Decadal discharge plots in figure 5 highlight the year-to-year variability of winter discharges at ORO–WA downstream of Osoyoos Lake. Ice jams occur sporadically throughout the entire study period (1942–2024), sometimes appearing in clusters. The largest cluster occurred during the 2019–2024 period and was preceded by a long period of no ice jams that began in 1990. Another cluster occurred during the 1983–1985 period.

In total, 31 days were identified by the algorithm as being affected by a total of 16 ice jams (durations ranging from 1 to 4 days). Identified ice jams are listed in table 2 with a minimum 2-day mean discharge, a maximum percent change in 2-day average discharge, and a maximum 2-day mean Osoyoos Lake level rise during the period of each identified ice jam. The 15-minute interval instantaneous discharge data, which began at ORO–WA in 1987, were used to determine the minimum discharge also presented in figure 2 for each identified ice jam since 1987. The maximum lake level is a conservative estimate because, in some cases, the lake level could have risen for some duration (fig. 6). However, the algorithm identified ice-jam days after the initial formation of an ice jam based on a continuous effect on both the discharge and the lake level. If the 2-day average discharge stopped declining by the defined threshold of 15 percent on a particular day, it is possible that a lag existed before the 2-day average lake level stopped rising by the defined threshold of 0.6 cm. This conservative approach emphasizes the effect of the downstream dewatering of the Okanogan River and reduces errors associated with lake-level rise before the formation of an ice jam.

An analysis of wind conditions showed that daily mean wind speed increased by a factor of two to three over a 3–5 day period before ice jams occurred (fig. 7). Wind speed was highest on the day of the ice jam, reaching an average of 22 km/h, and then decreased to typical speeds of 5–10 km/h within 2–3 days after the ice jam. The increase in wind speed was accompanied by a change in wind direction from southeasterly (blowing from the southeast) at approximately 150 degrees azimuth to northwesterly at greater than 300 degrees on the day of the ice jam (fig. 7). Daily mean wind direction values that occurred on the day of all ice jams identified by the algorithm were always above 300 degrees azimuth (fig 7).

Prior to the ice jam, changing wind conditions were accompanied by decreasing air temperature from near freezing to below freezing (≤ 3 degrees Celsius) the day before an ice jam (fig. 8). Water temperature at ORO–WA (fig. 9) and OLO–WA (fig. 9) also decreased substantially 2–3 days before an ice jam along with air temperature, potentially accelerated by wind-driven mixing. After an ice jam, air temperature (fig. 8) tended to increase after 2–3 days, and water temperature (fig. 9) remained near freezing for an extended period.

Statistical analyses of environmental conditions associated with ice-jam events on Osoyoos Lake indicated significant differences in environmental conditions associated with and without days affected by ice jams. Results from the PERMANOVA test indicated that the multivariate combination of minimum air temperature, wind direction, and wind speed was significantly different for ice-jam and non-ice-jam days at the 99 percent confidence level (p=0.001, pseudo-F-statistic=106.82). Pseudo-F is the test statistic derived from permutations in PERMANOVA. Figures 7, 8, and 9 illustrate a pattern of departure in air temperature, wind speed, and wind direction before the initial day of an ice jam. While this pattern appeared to start 1–3 days before an event, variability 2–3 days prior remained high for all three variables (figs. 7, 8, 9). Results from the nonparametric Wilcoxon rank-sum test indicated significant univariate differences between ice-jam and non-ice-jam days for wind speed (p<0.0001, W=37,622,545) and wind direction (p<0.0001, W=37,622,805; fig. 10). Differences in minimum air temperature, however, were not significantly different between ice-jam and non-ice-jam days (p=0.87, W=18,841,924). The results indicate that a shift toward westerly or northwesterly winds above approximately 200 degrees azimuth with a speed near or above 10 km/h during freezing conditions, particularly with daily minimum temperatures below –9.4 °C, are conditions under which ice jams are more likely to form (fig. 10).

PCA illustrated the separation of ice-jam days from non-ice-jam days in multidimensional space with 86 percent of the variability explained by the first two principal components (fig. 11; table 3). Vector arrows showing the variability in the data associated with wind speed and wind direction illustrated the relatively high correlation between the two variables. The 95-percent confidence-level ellipses for ice-jam and non-ice-jam days showed a separation based on air temperature and wind, where ice jams tended to form during low minimum air temperatures, high wind speeds, and northerly wind directions. In many cases, non-ice-jam events that fell within the 95-percent confidence-level ellipse for ice jams represented days immediately after ice-jam events, which was partially a result of using the rate of change in 2-day averages.

LDA using the same three variables used in the PCA suggested a high potential for predicting ice-jam events based on 2-day averages of wind direction, wind speed, and minimum air temperature. LDA results indicated a prediction accuracy of 99 percent, but because the number (n) of non-ice-jam days (n=6,121) was much more abundant than ice-jam days (n=13), the confusion matrix and success rate for ice-jam and non-ice-jam days must be examined. Because the number of verified ice-jam days was only 13 out of 6,134 observation days, it was not statistically robust to split observations into a training dataset and a validation dataset. This means that only the success rate of the training dataset was reported; a verification on a separate dataset was not conducted. Of the 13 ice-jam days, 12 days were successfully identified by LDA using 2-day averages of the three variables, with the remaining event misclassified as a non-jam day, resulting in a 92-percent success rate for ice-jam events. Non-ice-jam days were correctly identified at a 99-percent success rate, with 6,087 non-ice-jam days correctly identified and 34 non-ice-jam days misclassified as potential ice-jam days. An examination of the PCA plot (fig. 11) illustrates the non-ice-jam days that overlapped the 95-percent confidence ellipse for ice-jam day for the first (dimension 1) and second (dimension 2) principal components. As noted above, these days were commonly those preceding or following ice-jam days when meteorological conditions favored ice jams, but one had not started to develop. The 2-day averages, however, could skew the signal of post-ice-jam conditions after an ice jam started to break up.

While the statistical analysis presented here should be interpreted with caution regarding the prediction of future events, it provides insight into the environmental conditions associated with the occurrence of ice jams on Osoyoos Lake. The sample size of occurrences of ice-jam days was relatively small (n=13), and prevented assessment of the performance of the LDA model on both a training and a validation dataset, but the combination of results from other statistical analyses reinforced conclusions about factors that affect ice-jam formation on Osoyoos Lake. Significant differences between the wind conditions during ice-jam and non-ice-jam days (p<0.0001) are visible in the boxplots (fig. 10). The PCA plot provides a visualization of differences between non-ice-jam and ice-jam days in multidimensional space (fig. 11). The PCA plot also illustrates the differences between groups associated with results from the LDA, which indicates that air temperature, wind speed, and wind direction explain differences in ice-jam and non-ice-jam days.

The results indicate that cold conditions are not the only meteorological factor responsible for the formation of ice jams on Osoyoos Lake. Satellite imagery that shows Osoyoos Lake covered entirely in ice does not coincide with related ice-jam events during that period, whereas imagery with ice-over conditions at the outlet—but a lake surface mostly clear of ice—corresponded to ice-jam events (fig. 4). It is possible that conditions that facilitate a complete freeze-over on Osoyoos Lake are not the same conditions that cause ice jams. Strong winds can inhibit continuous ice formation and promote breakup and accumulation on the leeward side of the lake.

A working hypothesis for the formation of ice jams provides an explanation for the patterns observed under the meteorological conditions examined in this study. Frazil ice is composed of small (<1mm–10mm), often elongate ice crystals suspended in super-cooled and turbulent water, which can accumulate or coalesce in a variety of ways to form ice cover on the water’s surface (Barrette, 2021). Westerly and northwesterly winds on Osoyoos Lake appear to cause turbulent flow and frazil ice formation that accumulates as a slow-moving mass at the southern lake outlet. This mass can then cause a partial blockage that reduces flow through the river channel, promoting the further consolidation and fusion of ice particles into the continuous ice masses observed in figures 4C and 4F.

While the results suggest that ice-jam breakups on Osoyoos Lake are likely to occur once the conditions that facilitate their formation are no longer present, the results presented here provide a limited ability to infer breakup conditions. While higher lake levels may help facilitate a mechanical breakup of ice jams, ice jams seem to persist until some combination of warmer air temperatures, lower wind speed, and an easterly or southerly shift in wind direction occurs. The only day on which an ice jam was identified that had a 2-day average wind direction less than 240 azimuth degrees (applicable only for the period of record with wind data starting in 1994), for example, was the last day of the longest ice jam detected in the analysis. During the 4-day ice jam that initiated on December 27, 2021, wind direction shifted from 340 degrees the day before the start of the ice jam to 117 degrees and a 2-day average wind direction of 190 azimuth degrees on the last day of the ice-jam event. The lack of ice-jam events with dominantly southerly and easterly components to average wind direction indicate that wind direction is a major factor for ice-jam formation on Osoyoos Lake.

Potential increases in ice jams on Osoyoos Lake in the last decade are of concern, but the ice jams identified in this study were not significant enough to substantially affect lake levels. Interpretations of the effects of changes in discharge that could adversely affect fish, however, are more complicated. Experiments that examined the effects of dewatering on the survival of the four developmental phases of Chinook salmon suggested that the timing of ice jams may be critical for the survival of specific fish species. Early stages of Chinook salmon development, including cleavage eggs and embryos, had only a 2-percent mortality rate during 12 consecutive dewatering days. In contrast, later stages experienced high mortality rates under much shorter dewatering periods (Becker and others, 1983).

Nearly 100-percent mortality of Chinook salmon eleutheroembryos and preemergent alevins occurred after 48 hours and 6 consecutive hours of dewatering, respectively (Becker and others, 1983). This means that earlier developmental stages of salmon can endure the typical length of ice jams on Osoyoos Lake (1–4 days), but ice jams with a duration of less than 1 day (which were not examined in this study) could be detrimental to the later stages of salmon development if salmon spawning habitat is dewatered. In the broader context, ice jams that substantially reduce flow during the earlier part of fall-run spawning periods (October–November) may not be detrimental, but those that occur after November may be more adverse to salmon in a more sensitive life stage; however, the dewatering of salmon redds is only one aspect of mortality, because ice formation in dewatered rivers can be lethal to fish at any life stage.

The results presented here can inform land managers and dam operators about potential conditions under which ice-jam formation on Osoyoos Lake is more likely and prioritize future efforts to further understand ice jams and develop potential mitigation strategies to prevent them. Research that examines the physical characteristics of ice formation, breakup, and transport on Osoyoos Lake could provide additional insight into the mechanisms for ice-jam occurrence. Further research could involve deterministic or numerical modeling of ice formation, water temperature monitoring at various depths, examination of how warming winter temperatures might affect ice formation and ice-jam occurrences, or monitoring ice and ice-jam formation and movement using cameras.

Although Zosel Dam on the Okanogan River regulates lake levels on Osoyoos Lake, lake ice jams seem to share characteristics similar to those on rivers. Thus, additional research about the mechanistic causes of ice jams and the potential for predicting ice jams on Osoyoos Lake could incorporate methods used to study river ice jams. While the work presented here provides an upper temperature threshold above which ice jams are unlikely to occur on Osoyoos Lake, the LDA results presented also point to wind conditions as an important factor. White (2003) emphasizes the importance of considering additional environmental conditions and discusses the use of LDA, like the results presented here. In the summary of the numerous approaches to predictive modeling for the occurrence of river ice jams, White (2003) also discusses the potential for artificial intelligence and machine learning models to develop more sophisticated models that can provide various outputs for nuanced scenarios and conditions. Field methods such as additional temperature gaging at various depths and the measurement of ice formation may better inform such physically based and predictive models (Beltaos, 2008).

Local and Indigenous knowledge of ice formation and ice jams on Osoyoos Lake indicate that ice jams occurred less frequently and that it was typical for the lake to freeze over as solid ice in the past (Arnie Marchand, Okanagan Indian and member of the Confederated Tribes of the Colville Reservation, oral commun., 2023). This report’s analysis indicates a record of likely ice-jam occurrences on Osoyoos Lake since the 1950s and indicates a possible increase in ice-jam occurrences in the last decade of the study period (2014–2024). Still, these results cannot be used to infer anything about ice behavior before the simultaneous monitoring of discharge and lake levels began in 1942. The recent cluster of ice jams (2019–2024) might give the impression of increased frequency, but this conclusion depends on the temporal range considered and the time interval used for binning data. Ice jams occurred more frequently over the past 10 winters than during any previously observed interval (fig. 12). However, the results summarized in figure 12 could be biased, because the data analysis for recent decades is informed by data of higher quantity and resolution.

The meteorological data compiled (Government of Canada, 2024a, b; NCEI 2024a, b) provided an opportunity to assess long-term trends in air temperatures that may affect the formation and breakup of ice jams. Figure 13 shows the trend components of time series decompositions of the air temperature records for daily maximum, mean or midrange, and minimum temperature from 1945 to 2024. The analysis excludes missing or erratic data in the first 2–3 years of the temperature records. The results show a warming rate of 0.015 °C per year since 1945 in mean or midrange temperature (or 0.15 °C per decade) (Government of Canada, 2024a, b; NCEI 2024a, b). The daily maximum air temperature has increased at a rate of 0.0092 °C per year, while the daily minimum air temperature has increased at a rate of 0.021 °C per year (Government of Canada, 2024a, b; NCEI 2024a, b). In contrast to air temperature, analysis of water temperature at both ORO–WA and OLO–WA hydrologic stations (USGS 2024a, b) showed no significant trends, but their records are much shorter.

A warming trend was present in the daily mean or midrange air temperature, demonstrating the shifting climatic conditions of the Osoyoos Lake region. A daily mean warming rate of 0.15 °C per decade and a faster warming rate for minimum air temperature are consistent with observations for the broader inland Pacific Northwest region (Berkeley Earth, 2024; EPA, 2025). The effect of increasing air temperatures, particularly minimum air temperature, on ice-jam processes is uncertain, but one plausible hypothesis is that warmer air temperatures delay the formation of a rigid ice layer before or during storms, and this promotes or prolongs the formation of the mobile ice particles that can jam the narrow outlet of Osoyoos Lake. A potential relationship between ice-jam occurrences and increases in daily mean temperatures is likely more complicated than overall average warming winter conditions. The algorithm indicates that the mean winter air temperature was lower on ice-jam days, although the difference was not statistically significant (p-value=0.1). This suggests that a higher variability in temperature conditions or winter storm characteristics, in combination with wind behavior and other environmental factors, may contribute to increased ice-jam occurrences.