A mathematical model for Nordic skiing

This paper presents a mathematical model for Nordic skiing that uses a three-dimensional space curve and a system of nonlinear ordinary differential equations to simulate skier dynamics, employing a specialized algorithm combining Hermite spline interpolation, numerical quadrature, and high-order ODE solvers to validate the approach against real-world data while demonstrating the practical application of undergraduate calculus and scientific computing concepts.

The Gist

Imagine you are trying to predict exactly how fast a cross-country skier will finish a race. It's not just about how strong their legs are; it's a complex dance between their muscles, the shape of the snow-covered path, gravity, wind, and even how they turn corners.

This paper is like a mathematical recipe book for simulating that race. The authors, who are mathematicians and scientists, built a computer program that acts as a "virtual skier" to see how different factors change the outcome. Here is how they did it, explained in simple terms:

1. Drawing the Map (The Course)

Real ski courses aren't perfect straight lines; they are winding, bumpy, 3D paths. Usually, we only have a few scattered GPS points (like dots on a map) to describe the course.

• The Problem: If you just connect those dots with straight lines (like a child connecting dots on a page), the path looks jagged and unrealistic. If you try to smooth it out with standard math curves, it sometimes creates "ghost hills" or dips that don't exist in reality (like a wobbly drawing).
• The Solution: The authors used a special type of mathematical smoothing called a Hermite spline. Think of this as a flexible ruler that bends perfectly through the GPS dots without creating fake bumps. It creates a smooth, realistic road for their virtual skier to travel on.

2. The Virtual Skier's Physics (The Engine)

Once the road is drawn, they put a "virtual skier" on it. This skier is governed by the laws of physics (Newton's laws), which the authors turned into a set of equations.

• The Forces: The skier is being pushed and pulled by four main things:
  1. Muscle Power: The skier pushes forward. The model assumes they push hardest when going uphill (slow) and coast more when going downhill (fast).
  2. Gravity: Gravity pulls them down hills (speeding them up) and holds them back going up hills (slowing them down).
  3. Friction: The snow rubs against the skis, slowing them down.
  4. Wind Resistance: The air pushes back against them, especially when they are going fast.
• The Math: They solved these equations using a high-tech calculator (a computer solver) that adjusts its speed to get the answer exactly right, even when the terrain gets tricky.

3. The 3D Twist (Turning and Braking)

Most previous models only looked at the race from the side (2D), like watching a movie on a flat screen. But real skiing happens in 3D.

• The New Feature: The authors added the ability for the skier to turn left and right. When a skier turns sharply on a downhill, they have to brake to avoid flying off the track.
• The Analogy: Imagine driving a car around a sharp curve. If you go too fast, you slide. The skier has to "skid" or "step" to slow down. The model calculates this "braking force." They found that how a skier turns can add or subtract several seconds from their total time—a huge deal in a race where winners are often separated by fractions of a second.

4. Testing the Model

The team tested their virtual skier against real-world data:

• The "Baseline" Test: They ran a simulation on a 4.2 km course and compared it to real race times. Their model was incredibly accurate, matching real results within a few seconds.
• The "Elite" Test: They simulated a 15 km race with 36 different real athletes. By tweaking the "muscle power" setting in their computer, they could perfectly match the finish times of slow skiers, fast skiers, and even the race winner.
• The Fatigue Factor: They noticed that real skiers slow down at the end of a long race because they get tired. Their basic model didn't account for this, so they showed how to add a "fatigue switch" to make the virtual skier get slower as the race goes on.

Why This Matters

The authors say this isn't just for sports fans. They designed this paper to show that math you learn in college (like calculus and computer coding) can solve real, messy problems.

• It proves that using a smoother, more accurate map (the spline) gives better results than using a jagged, simple one.
• It shows that 3D effects (like turning and braking) are crucial for understanding how elite athletes win.
• It provides a free, open-source computer code that coaches, scientists, and students can use to experiment with different race strategies.

In short, the paper builds a digital twin of a cross-country skier. It takes a rough map, applies the laws of physics, and simulates a race so accurately that it helps us understand the tiny details—like how a skier turns a corner—that can mean the difference between gold and silver.

Technical Summary

Technical Summary: A Mathematical Model for Nordic Skiing

Problem Statement
Nordic (cross-country) skiing presents a complex dynamic system where an athlete navigates a three-dimensional space curve characterized by frequent changes in elevation, direction, and curvature. Unlike alpine skiing, where gravity primarily drives motion, Nordic skiers must generate propulsive forces to maintain forward motion on flats and uphills while managing resistance from snow friction and aerodynamic drag. Existing models, such as the one proposed by Moxnes, Sandbakk, and Hausken (MSH), have utilized ordinary differential equations (ODEs) to describe skier dynamics on two-dimensional (2D) courses. However, these models often rely on piecewise linear approximations of course geometry, which fail to capture the smooth curvature of real tracks and introduce numerical artifacts. Furthermore, standard 2D models neglect the critical 3D effects of turning, specifically the braking forces required to negotiate tight downhill curves safely. This paper addresses the need for a more realistic, high-fidelity mathematical model that integrates smooth course interpolation, rigorous force balancing, and 3D turning dynamics, while remaining accessible to undergraduate students in calculus and scientific computing.

Methodology
The authors develop a comprehensive modeling framework grounded in Newtonian mechanics, formulated as a system of nonlinear ordinary differential equations (ODEs). The methodology proceeds through several key stages:

1. Course Geometry and Interpolation:

   • The paper moves beyond the piecewise linear representations used in prior work. Instead, it employs Cubic Hermite spline interpolation (specifically the makima variant in MATLAB) to construct smooth, shape-preserving curves from sparse GPS data or 2D elevation profiles.
   • This approach avoids the "Runge's phenomenon" (spurious oscillations) associated with standard cubic splines when fitting irregularly spaced data.
   • The course is parameterized by projected arc length ($\xi$) to accurately represent geometric quantities like slope and curvature.
   • The model distinguishes between 2D simulations (elevation profile only) and 3D simulations (full spatial coordinates $x, y, z$), calculating inclination angles ($\theta$) and azimuth angles ($\phi$) from spline derivatives.

2. Governing Equations (Dynamics):

   • 2D Model: The skier is treated as a point mass. The system solves for speed ($v$) and projected distance ($\xi$) using two coupled ODEs:
     $$v' = \frac{P(v)}{mv} - g \sin \theta - \mu g \cos \theta - \beta v^2$$
     $$\xi' = v \cos \theta$$
     Here, forces include propulsion ($F_p$), gravity ($F_g$), snow friction ($F_s$), and aerodynamic drag ($F_d$). The propulsion force is derived from a speed-dependent power function $P(v)$, and drag is modeled as quadratic in speed with a coefficient that switches between "upright" and "tuck" positions based on velocity thresholds.
   • 3D Extension: The model incorporates a braking force ($F_b$) to account for centripetal acceleration in tight turns. This force is proportional to the centripetal acceleration ($\kappa v^2$) and is only activated when the skier is on a downhill section ($\theta < 0$) and the acceleration exceeds a specific threshold. The braking coefficient ($\gamma$) varies based on the technique used (skid turn vs. step turn).

3. Numerical Implementation:

   • The ODE systems are solved using MATLAB's ode113, a high-order, variable-step, variable-order solver (Adams-Bashforth-Moulton) with adaptive time stepping and error control.
   • Arc length integrals for distance and curvature are approximated using Simpson's rule (fourth-order accuracy).
   • The code ensures the skier remains constrained to the interpolated spline path, eliminating "drift" errors common in augmented ODE systems that solve for vertical position independently.

Key Contributions

• Smooth Geometry Representation: The paper demonstrates that replacing piecewise linear course approximations with Hermite splines yields more realistic simulations. While MSH found cubic splines to be less accurate due to oscillations, this work shows that shape-preserving Hermite splines eliminate these artifacts while providing a smoother, more accurate representation of the track.
• 3D Braking Dynamics: The authors introduce a novel 3D component to Nordic skiing models: a braking force model dependent on track curvature and skier speed. This allows for the simulation of strategic braking techniques (skid vs. step turns) on steep, tight downhill sections, a factor previously omitted in 2D models.
• Validation and Calibration: The model is validated against:
  • The MSH baseline data (4.2 km course), showing close agreement in finish times and speed profiles.
  • Experimental data from Welde et al. (15 km race), successfully differentiating between "slow," "fast," and "winner" skiers by adjusting the maximum power parameter ($P_{max}$).
  • Real FIS-certified courses (Ole and Stephanie courses in Toblach, Italy), verifying that the model adheres to homologation criteria (e.g., gradient limits, course length).
• Educational Framework: The paper explicitly structures the model to utilize elementary concepts from undergraduate calculus (vector calculus, derivatives, integrals) and scientific computing (ODE solvers, interpolation), serving as a bridge between theoretical mathematics and applied sports science.

Results

• 2D Simulations: On the MSH course, the Hermite spline model produced a finish time of 823 seconds, falling between the linear spline result (813 s) and the oscillatory cubic spline result (estimated 835 s). The model successfully replicated the speed profiles and force balances observed in the baseline study.
• Parameter Sensitivity: The model confirmed that skier mass and power output are critical differentiators. By adjusting $P_{max}$, the model could match the performance of elite skiers (Winner: $P_{max} = 434$ W) versus slower competitors ($P_{max} = 346$ W).
• 3D Braking Impact: Simulations on the 4.2 km Ole course showed that applying braking forces increased finish times by 10–30 seconds depending on the technique and threshold. The "skid turn" technique (higher braking coefficient) resulted in slower times than the "step turn." On the more curvy "Stephanie" course, the time penalty for braking was even more pronounced (up to +63 seconds), highlighting the strategic importance of turn management.
• Fatigue Modeling: The paper notes that the constant-power model cannot fully capture the "positive pacing" (slowing down over the race) observed in real races. A simple modification allowing $P_{max}$ to decay over time was proposed to mimic muscle fatigue, though a full physiological model was not developed.

Significance and Claims
The authors position this work as a demonstration of how elementary mathematical concepts can be applied to solve real-world problems in sports science. They claim that their model provides:

1. Novel Insights: Specifically, the quantification of how braking techniques on tight downhill curves significantly impact race times, a factor often overlooked in 2D analyses.
2. Educational Value: The model serves as a concrete example for undergraduate courses in differential equations and numerical methods, showing students how to move from physical laws to computational solutions.
3. Practical Utility: The open-source MATLAB code allows sports scientists and coaches to test race strategies and analyze course geometries without needing to develop complex biomechanical models from scratch.

The paper remains modest in its scope, acknowledging that it is a "preliminary step" toward a truly realistic model. It explicitly excludes complex metabolic processes, detailed muscle biomechanics, and advanced fatigue models, focusing instead on the balance of forces and the geometric constraints of the course. The authors suggest that future work could integrate "look-ahead" braking strategies and more sophisticated pacing models based on experimental data.