An evidence-gated audit algorithm for low-altitude-economy readiness in mountainous Yunnan, China

In the rugged, high-altitude landscapes of southwestern China, a new kind of economic activity is taking flight. The "low-altitude economy" refers to the use of aircraft that fly below 3,000 meters, including drones and small electric planes, for tasks like delivering medical supplies, inspecting power lines, or transporting tourists. While cities with flat terrain and established data have clear blueprints for where these flights should go, mountainous regions face a different challenge. Here, the roads are winding, the weather changes rapidly, and there is no historical record of flight demand because the industry has not yet started. Planners in these areas must decide where to send their first teams to investigate potential routes and sites, but they lack the usual evidence of market success or safety records. They are forced to guess which towns are ready for this future, often relying on general economic numbers or government policy lists that might not tell the whole story.

This is the precise problem a team of researchers from Yunnan Communications Vocational and Technical College set out to solve. They developed a new method to help local governments decide where to begin their field investigations without falling into the trap of circular reasoning. In many planning exercises, a computer model might simply learn to repeat the government's existing labels, telling planners to invest in places that are already famous for having a "low-altitude economy" plan, rather than identifying places that are actually structurally ready for it. The researchers wanted to strip away these self-fulfilling prophecies to find the genuine structural signals hidden in the data. Their goal was not to predict the future of flight traffic or to design specific air routes, but to create a disciplined, auditable checklist that separates administrative ambition from actual operational readiness.

The researchers focused on 129 counties across Yunnan Province, a region known for its dramatic elevation changes and uneven economic development. They started by examining a massive list of data points for each county, including GDP, digital infrastructure, and tourism numbers. However, they quickly discovered a critical flaw in how such data is often used: the "policy support score," a number given by government departments to indicate how much a county is encouraged to develop this sector, was almost identical to the target they were trying to predict. If a model used this score, it would be like a student reading the answer key on a test; it would perform perfectly but learn nothing new. To fix this, the researchers built a "gate" that strictly removed any data that was just a copy of the government's own goals. They also removed variables that were direct results of the policy, ensuring the model only looked at the raw, structural conditions of the land and economy.

With these circular clues removed, the team ran a series of computer simulations to see which counties looked most ready based on their physical and economic foundations. They did not just trust a single computer program; instead, they used a "committee" approach where multiple different models voted on the rankings. Crucially, they tested these models in a way that mimics real-world expansion: they trained the models on some parts of the province and asked them to predict the readiness of entirely different regions they had never seen before. This ensured the results would hold up when applied to new places, rather than just memorizing the specific details of the training data. The system also added a "penalty" for uncertainty. If the different computer models disagreed on a county's score, or if the county was in a unique geographic spot that the other data didn't cover well, the system lowered its confidence in that ranking. This created a conservative, cautious list that prioritized safety and reliability over optimistic guesses.

The results of this careful audit were revealing. When the researchers looked at the top 20 counties recommended by their new system, they found that these places captured nearly 87% of the counties that were already considered "established" in the industry, and they included almost all the specific application scenarios that companies had already submitted for review. This proved that the method could successfully identify where the industry was naturally clustering. However, the study also showed that a simple ranking based purely on economic size (GDP) performed almost as well at finding these spots. This finding is significant because it suggests that the complex new algorithm does not need to be more powerful than simple economic metrics to find the right places. Instead, its true value lies in what it does not do: it refuses to be tricked by policy labels, it explicitly measures how uncertain the prediction is, and it clearly states what the list can and cannot be used for.

The researchers organized their final list into specific action tiers to guide decision-makers. The top tier included counties that were both high-ranking in the structural audit and had confirmed applications from companies, marking them as the highest priority for immediate field investigation. The next tier included counties that looked structurally ready but lacked confirmed company interest, suggesting they needed further local verification. Lower tiers were reserved for areas with insufficient evidence or those that simply did not meet the structural criteria. This tiered approach allows planners to see not just a list of names, but a map of confidence. It tells them that while a county might have a high score, if the computer models disagree about it or if the local data is sparse, the score should be treated with caution.

Ultimately, the paper argues that in regions where data is scarce and the terrain is difficult, the most important tool is not a crystal ball that predicts the future, but a rigorous filter that prevents bad decisions. The study demonstrates that while economic strength is the dominant factor in determining where these industries will start, the process of finding them must be transparent and free from the bias of government wish lists. By removing the circular evidence and penalizing uncertainty, the researchers provided a framework that can be used in other mountainous or unevenly developed regions around the world. The takeaway is not that a specific algorithm has solved the problem of low-altitude flight, but that a disciplined, evidence-gated approach can help planners start their investigations in the right places, avoiding the trap of investing in areas that only look good on paper.