Summit Strategies: Factors Shaping Success and Safety in Himalayan Expeditions
Introduction
In this data analysis, we explored the history and patterns of Himalayan mountaineering expeditions, using the Himalayan dataset. This dataset documents the climbs from 2020 to 2024 across the Nepal Himalayas. It was initially compiled by the journalist Elizabeth Hawley and is continued by the Himalayan Database organization. Our objective is to understand how factors such as seasonality, oxygen use, climbing routes, and team composition influence expedition outcome, which is measured by indicators such as success rates and death rates. Through this analysis, we aim to uncover insights into the different ways in which climbers adapt to extreme conditions and identify the choices that affect safety and success. This study not only contributes to mountaineering but also shows how people push through tough conditions, weigh risks, and make strategic choices in some of the harshest environments.
Dataset Description
The Himalayan Database contains detailed information on mountaineering expeditions in Nepal, which includes 2 tables:
exped_tidy: Expedition data from 2020 to 2024. Each row represents a single expedition.
peaks_tidy: Information about individual Himalayan peaks
Each row in exped_tidy represents a single expedition, described by a range of categorical and quantitative variables. Here are the key variables used in our analysis
Categorical Variables
SEASON_FACTOR – the season in which the expedition occurred (Spring, Summer, Autumn, Winter)
ROUTE1 – the primary climbing route used by the expedition
O2USED – whether supplemental oxygen was used during the expedition (TRUE or FALSE)
NOHIRED – whether no hired personnel were used above base camp (TRUE or FALSE)
TERMREASON – coded integer indicating why the expedition was terminated
To analyze success rates, we constructed a new binary variable success: A combined indicator derived from SUCCESS1, SUCCESS2, SUCCESS3, and SUCCESS4. It is TRUE if the expedition succeeded on at least one of the four recorded routes.
SMTDATE – Used to extract date and create derived time variables
Quantitative Variables
TOTMEMBERS: Total number of expedition members
TOTHIRED: Total number of hired personnel above base camp
MDEATHS: Number of member deaths during the expedition
HDEATHS: Number of hired personnel deaths during the expedition
HIGHPOINT: Highest altitude reached in meters
CAMPS: Number of high camps established above base camp
ROPE: The amount of fixed rope used (in meters)
The dataset, through a mix of quantitative and categorical variables, provides a detailed exploration into the outcomes of Himalayan expeditions. It allows us to investigate the relationship between expedition success and safety, and key factors of season, oxygen use, climbing routes, team composition, and environmental obstacles.
Research Question 1: How does logistical preparedness influence the primary reason an expedition is forced to terminate?
Before analyzing success factors, we first investigated the factors that impacted the reasons for termination among failed expeditions. This will help expedition teams understand exactly how certain preparations can help them avoid specific failure points. Since there are over 15 different reasons for termination in the dataset, I first combine these reasons into 4 main reasons for better visualization and analysis. The 4 main reasons are:
Termination due to bad weather such as storms and high winds
Termination due to accidents or health-related reasons such as illness and frostbite
Logistical and technical reasons such as lack of supplies or an overly difficult route
Other reasons as well as abandoning the expedition before reaching base camp (or not even attempting the climb)
To start, I wanted to investigate how logistical preparedness could impact reasons for termination, so I used variables such as the use of rope, high camps, oxygen, and hired personnel to produce a dendrogram, as shown below.
There are two clusterings of the leaves: one pink subset on the far left for expeditions that ended due to abandonment or other reasons, as well as a green subset for weather-related terminations. However, outside of those pockets, there is a lot of mixing and most failure reasons for scattered across the tree. Thus, while abandonment and weather terminations can sometimes group neatly, most expeditions can fail for very different reasons despite nearly identical logistical profiles.
To investigate this further, I examined two specific variables for logistical preparedness: the use of fixed rope and high camps. For each combination of the two, I created a stacked bar chart showing the proportion of expeditions for each termination reason. Note that I didn’t include a stacked bar chart for expeditions that used rope but not camps since only 1 expedition fit that category.
The main takeaway from this graph is that logistical preparedness reduces the abandonment rate. For expeditions that don’t use rope or high camps, the abandonment rate is nearly 50%. However, this statistic is cut by half for expeditions that use camps, and nearly quartered for expeditions that use both.
Isolating terminations due to abandonment, we have the “filtered” stacked bar charts on the right, which shows us reasons for termination among expeditions that actually got started. Here, the influence of ropes and camps is less obvious. Counter-intuitively, it seems like expeditions that don’t use ropes or camps have less terminations due to logistical and technical reasons. However, that may be due to those expeditions being abandoned at the start.
Research Question 2: How does the use of supplemental oxygen affect summit success rates across different seasons?
Next, we explored whether the use of supplemental oxygen influences summit success rates. We also categorized expeditions by season, as environmental conditions vary significantly throughout the year and may affect the need for oxygen. To investigate this question, we first defined what constitutes a successful summit. Each expedition includes up to four possible routes, with a logical variable indicating whether the summit was reached via each route. We aggregated these indicators to determine whether the team succeeded on any route.
First, we visualized the monthly success rates using a time series plot that aggregates expeditions across all years into a “typical year”. The dashed LOESS lines were added to smooth out month-to-month fluctuations and better illustrate general patterns over time. To provide more information, we included the number of expeditions as labels above each month, which is important to consider because there are months with very few expeditions. After showing the count, we found two key periods for climbing, which are the spring climbing season (March - May, highlighted using blue) and the Autumn climbing season (September - November, highlighted using orange).
From the above graph, we can observe several key patterns. Climbers using supplemental oxygen tend to have a consistently higher success rate than those who do not, especially in spring (the months of March to June) . Additionally, success rates for non-oxygen users fluctuate more, with lows in months like March and September, reaching less than 50%. This suggests greater sensitivity to harsh or unpredictable conditions without using supplemental oxygen. Interestingly, there is a sharp peak in success rates during summer months, with those who do not use oxygen being higher than those who do. This initially counterintuitive trend might be because of the small sample size, which inflates the apparent success rate. There might also be selection bias, because during this dangerous season, only the most prepared and experienced climbers come, contributing to the unusually high success rates.
Next, to take a closer look at how the group expedition success rate varies with oxygen by season, we grouped the climbs by season. When collecting the data, we noticed that some expeditions in June are not labelled as summer. Therefore, we manually re-categorized the data so that each season is within its defined months for consistency. Moreover, we found that the difference between oxygen users and non-oxygen users for autumn and winter is not clear in the time series. Therefore, to formally assess whether the differences observed in summit success between oxygen users and non-oxygen users are statistically significant, we used a Chi-square test for independence for each season. The null hypothesis is there is summit success rate is independent of oxygen use across different seasons. The alternative hypothesis is summit success rate depends on oxygen use across different seasons. Here is the bar plot that shows the relationship and the p-values from statistical testing:
From the above bar plot, we can observe several key patterns. In Spring, those who use supplemental oxygen have a dramatically higher success rate (94%) compared to those who do not (45%), which aligns with the time-series data generated. The p-value is 1.34e-25, which is below the usually used significance level of 0.05. This gives us sufficient evidence to reject the null hypothesis that oxygen use is not associated with success in this season. In Autumn, the difference is much smaller, and the p-value of 0.578 indicates that the difference is not statistically significant. For Summer and Winter, although there is an observable difference in the height of the bars, the p-value of 1 suggests no statistically meaningful difference in success rate. Again, these results are likely unreliable due to extremely small sample sizes (e.g., n=1 in winter without oxygen), and thus should be interpreted with caution.
Overall, the main takeaway from these two graphs is that supplemental oxygen is most strongly associated with increased summit success during Spring, which is the primary climbing season across all seasons. In other seasons, the effect of oxygen is less clear, and the statistical test results do not provide strong evidence of the presence of an association.
Research Question 3: How Do Season and Route Choice Influence Expedition Success?
As Himalayan expeditions continue to evolve with better gear and more diverse climbing teams, a fundamental question persists: how do seasonal conditions and chosen climbing routes affect the likelihood of climbing successfully? This question intrigued me because while many climbers prepare extensively for the climb, external factors such as weather (strongly tied to season) and route difficulty remain critical to the outcome.
To have an idea of when do most climbers go climbing Himalayan, we plotted a pie chart of the climbing records given the season.
This pie chart reveals that spring is by far the most popular climbing season, likely due to its favorable weather conditions and longer periods of stable temperatures. Autumn comes next, followed by relatively few expeditions in summer and winter. The uneven distribution across seasons strongly indicate the climbers’ preferences. Then, we are more curious how likely the climbers will success during the popular season and unpopular season, and are the two factors related?
To answer these questions, we first plotted a more detailed heat map.
This heat map visualizes the success rates of Himalayan expeditions by route and season. Each row represents a different climbing route, while the columns show the four seasons. The color intensity represents the proportion of successful expeditions, with darker green indicating higher success rates. As we expected, it has the same patter as the pie chart - concentration of climbs during the spring and autumn seasons. Despite the strong visuals, we still conducted chi-square test to test whether there is a relationship between the season and number of climbing.
Null Hypothesis: Number of expeditions is independent of season (equal probability across seasons)
Alternative Hypothesis:The number of expeditions depends on the season
Chi-squared test for given probabilities
data: table(exped_tidy$SEASON_FACTOR)
X-squared = 792.13, df = 3, p-value < 2.2e-16We choose to test at level of alpha = 0.05. The Chi-squared test produced a p-value < 0.001, indicating that the number of expeditions is significantly related to the climbing season. This confirms what we see visually — most expeditions occur in spring and autumn, while summer and winter see far fewer climbs, likely due to extreme weather conditions.
With a deeper research into the climate around Himalayan area, spring and autumn are the traditional windows for Himalayan expeditions due to more stable weather conditions, lower avalanche risk, and predictable jet streams. In contrast, very few teams attempt climbs in summer or winter. Summer coincides with the monsoon season, bringing heavy snowfall, rainfall, and poor visibility — all of which increase the danger significantly. Winter, on the other hand, poses challenges such as extreme cold, high winds, and icy terrain.
Yet what’s fascinating is that the few expeditions that did occur in these off-seasons — summer and winter — all succeeded. This strongly suggests that only highly experienced and well-supported teams are willing to take on the extreme risk. Their success is likely due to meticulous preparation, cutting-edge gear, and favorable weather windows.
Looking at the routes themselves, some stand out with consistently high success rates — for example, the N Col-NE Ridge route on Everest and the SW Ridge on Ama Dablam are not only popular but also have strong success records in spring and autumn. These are well-established routes with fixed lines, established camps, and known logistics, which make them more manageable even at high altitudes.
In summary, the data illustrates how seasonal timing and route selection play a critical role in determining the outcome of an expedition. Climbers are strategically optimizing for windows of opportunity, and while off-season climbs are rare, they can still succeed when executed with expertise.
Research Question 4: Is Expedition Member Death associated with the group scale or the reached highest altitude?
Finally, we focus on the relationship between member death and other factors. Although many factors can affect the member’s safety, I determine that the group scale and the altitude they arrived at play an important role in influencing the safety. Since a larger group scale determines a larger ability of facing the death risk to probably reducing the appearance of death by helping each other in the group; and the higher altitude determine lower temperature and stronger wind, that is the higher altitude implies to more trouble surroundings, which may increase the risk of member death.
To visualize this question, we create a statistical graph. Its x-axis represents the number of expedition people in the group, and its y-axis represents the number of hired people in the group. Since we only care about the relationship between member death and other factors, we draw these groups in different colors, according to the number of death members. We also use different shapes to represent the highest altitude that the group reaches.
Observing this plot, we first consider the relationship between the member’s death and the highest altitude the group reached. We can temporarily ignore the detailed color of these points and focus on the shape of these points. There are 26 colored points, and only 4 points are not triangles. That is, these groups with the appearance of death, 22 groups, reach the highest altitude greater than 8000 meters, accounting for about 85% of those groups with member death. For the rest colored points, the highest altitude reached by 1 group is 7000-8000 meters, 2 groups are 6000-7000 meters, 1 group is 5000-6000 meters, and no colored points with a highest altitude less than 5000 meters. There is an obvious relationship, so we conclude that the larger the highest altitude group reached causes to higher the death risk, especially over 8000 meters.
Secondly, we analyze whether group scale affects the occurrence of death. We can focus on the distribution of total points and the color points distribution in the contour plot. There is a mode on the bottom left, and half of more colored points gather there. For the rest of the colored points, it is difficult to conclude whether they form a mode in the middle of the plot. If the distribution of colored points and total points is similar, we can conclude that the group scale does not affect the death appearance. However, it is difficult to verify this; to do this, we use a t-test to verify our conclusion.
Null Hypothesis: The group scales of the group with member death and the groups without death are the same.
Alternative Hypothesis: The group scales of groups with member death and groups without death are different.
Welch Two Sample t-test
data: team_size by MDEATHS > 0
t = -3.1365, df = 26.804, p-value = 0.004123
alternative hypothesis: true difference in means between group FALSE and group TRUE is not equal to 0
95 percent confidence interval:
-22.236998 -4.645156
sample estimates:
mean in group FALSE mean in group TRUE
15.81818 29.25926 The p-value of the t-test is less than 0.05, so we reject the null hypothesis, and the group scales of the group with member death and the group without death are different. Comparing their mean values, we can conclude that a larger group scale has a higher probability of the occurrence of expedition members’ deaths. Since the number of groups with member death is too small, only 26 groups out of 882 total groups, the mode of groups with member death is not obvious in the graph.
Furtherly, we focus on the relationship between the number of death members and group scale. Back to the scatterplot, the colored points are on the diagonal, and most of them are purple. So, most of the groups with member deaths have only one expedition member death. However, we notice that there is a point with 3 dead members lying at the bottom and three points with other than 1 dead member lying on the right top, which affects our judgment. Due to the small size of the group with member death, our conclusion is easily influenced by the outliers, so we create a trend line with group scale and the number of death members to verify our conclusion.
The y-axis represents the number of death expedition members, and the x-axis represents the number of members and hired people in the group. We notice that the line is generally and roughly flat as the group scale is less than 50, even there is a point with 3 death members, while the line starts to be slope as the group scale is greater than 50, which may tell us 50 group scale is a turning point for the death member counts. Hence, we may conclude that group scale has a limited and weak impact on the number of member deaths. However, it is worth emphasizing that our conclusion may still have some bias due to the small number of deaths group and those outliers.
In total, the highest altitude group reached can affect member death; group scale can also affect the member death, but has a limited and weak impact on death number. Higher groups reached the highest altitude increases the risk of death; the increment of group scale increases the probability of occurrence of member death; but only the group scales larger than 50, the number of member deaths increases as group scale increases. Due to the small size of groups with member deaths, our exploration may involve some bias. We may consider other variables, such as the rate of death among members, or more developed statistical technology for small group sizes to reduce the bias. In the future, we may consider the development of technology as a factor to balance the weight of deaths across different years, for example, a death in the 2010s may be more significant and noteworthy than one in the 1940s, and give a better conclusion about climbing safety.
Conclusion
In this data analysis, we successfully conclude that expeditions that use rope and high camps have lower chance of abandonment; oxygen supplemental increase summit success across all seasons; climbing in spring and fall with wisdom route selection increase summit success rate; and higher altitude and more group scale increases the risk of death, while group scale is very weakly associated with the number of expeditions death. Our research can provide a summit strategy and for those people who want to summit, an analysis expedition death to reduce the risk of death to increase the safety of the group. In the further, base on our research, other researcher may add more complex factors to explore more detailed and specific question, for example, the development of climbing equipment whether increases summit success rate and reduce the risk of death, or the degree of wealth whether affect summit success and reduce the risk of death. Eventually, our research is successful in uncovering insights into the different ways in which climbers adapt to extreme conditions and identify the choices that affect safety and success.