A question lingering in the back of my mind

When I started my PhD, my goal was clear: I wanted to explore how passive design strategies could reduce indoor overheating in tropical climates. My research included several case studies, ranging from modern buildings in high-income Singapore to low-income housing in Honduras. But it was the latter that remained closest to my heart—low-income housing where people often suffer the most from extreme heat. In countries like Honduras, where air conditioning is a luxury far beyond the reach of most families, passive design isn’t just a sustainable option—it’s often the only viable path to achieving thermal comfort.

What I kept asking myself was this: How much daylight are we sacrificing when we try to reduce overheating? Or the opposite—how much hotter do we make homes by allowing more daylight in?

Natural daylight is not only a matter of visual comfort. Natural daylight can reduce electricity bills and improve wellbeing. Yet many strategies that reduce overheating—like small windows, deep overhangs, or shaded patios—can also make interiors dark and gloomy.

I didn’t know it then, but this seemingly simple question would later lead me to evaluate a mathematical method almost never used in building performance research.

Four Years Later: a New Direction

It's been almost four years since I completed my PhD. Today, I work at the Institute of Data Science and Artificial Intelligence (DATAI) at the University of Navarra, where I focus on AI- and data-driven approaches to building performance and sustainable design. Surrounded by new tools and new ways of thinking, I saw an opportunity to revisit my old question through a fresh lens.

My colleague and supervisor at DATAI, Jesús López Fidalgo, a leading expert in the optimal design of experiments, suggested I look into a technique called Response Surface Methodology (RSM) ^{1, 2}. At first, it sounded like something from another field—and in many ways, it was. RSM has been widely used in statistics, manufacturing, and even pharmaceutical research, but as I dug deeper, I was surprised to find something striking:

RSM had barely been used in the field of building performance simulation.

That gap was too big to ignore.

The Balancing Act: Heat vs. Light

In building science, the conflict between thermal comfort and daylight isn’t new. What is new is how we’re beginning to solve it. Many existing optimization methods—like genetic algorithms—are highly effective but come at a cost: they require running thousands of simulations.

Instead of using brute-force optimization, I turned to RSM—all developed using a basic office laptop. You can think of RSM like a smart shortcut. Rather than testing every possible design, it builds a statistical model—a “response surface”—based on a small number of carefully selected simulations. That model then guides the search for the best design combinations, saving time and computational effort while still delivering powerful insights.

But optimizing for two outcomes—overheating and daylight—requires more than just a map. It requires a compass.

That compass, in this study, was desirability functions ³.

Desirability Score

Desirability functions take multiple objectives—like “as little overheating as possible” and “as much daylight as possible”—and turn them into a single, composite score. Each outcome is transformed into a scale from 0 to 1 (undesirable to ideal), and the algorithm searches for designs where both are maximized simultaneously.

This way, instead of treating daylight and overheating as rivals, we treat them as co-players in a game of compromise.

The result?

A design process that required only 138 simulations—not thousands—to find solutions that reduce indoor overheating without sacrificing daylight. That’s not just an academic achievement. It’s something that could genuinely be used in low-resource settings.

And the results weren’t just approximate guesses—they came with statistical confidence. By using 1,000 bootstrap replications, the study was able to validate the reliability of the optimal values. Here's what that looked like:

• West-facing window-to-wall ratio (WWR): 3.76% ± 2%

• South-facing WWR: 29.3% ± 2.75%

• Roof overhang depth (west and south facades): 3.78 m ± 0.25 m

These values strike a carefully balanced compromise between reducing indoor overheating and maximizing daylight availability—tailored specifically for low-income tropical housing. It’s a reminder that a good experiment doesn’t have to be expensive or time-consuming— it just has to be smart.

Want to explore the details?

If you're curious about how it all works under the hood—from the simulations to the RSM models and desirability functions—you can dive into the full code and data behind this study. The entire repository is openly available on GitHub:

👉 github.com/juan-gamero-salinas/rsm-thermal-daylight-optimization

And if you're looking for the full peer-reviewed paper published in Scientific Reports, you can read it here:

📄 https://doi.org/10.1038/s41598-025-96376-x

References

1. Myers, Raymond H., Douglas C. Montgomery, and Christine M. Anderson-Cook (2016). Response surface methodology: process and product optimization using designed experiments. John Wiley & Sons.

2. Montgomery, D. C. (2017). Design and analysis of experiments. John Wiley & Sons.

3. Derringer, G., & Suich, R. (1980). Simultaneous optimization of several response variables. Journal of quality technology, 12(4), 214-219.