Replacing Labor with Logic: Smart Models Reveal Beet Armyworms 10 Days Early for I-lan Scallion Farmers.

Scallion fields in Sanxing, Yilan, have faced immense challenges in recent years. On one hand, farmers must cope with heavy rainfalls, high temperatures, and droughts brought about by climate change. On the other hand, they have to battle an elusive yet highly destructive pest—the beet armyworm.

Beet armyworms hide during the day and come out at night. Adults enter the scallion fields after dark to lay eggs. Once the larvae hatch, they burrow into the scallion tubes to feed, hollow out the leaves from the inside. This damage is not always obvious from the exterior, meaning that by the time it is discovered, it is often too late. Furthermore, due to the long-term, intensive cultivation of Sanxing Scallions, the population density of these pests remains high, making pest control even more difficult.

Since 2019, green scallion production in the Sanxing region has experienced a steady decline. At its peak, annual production reached 13,000 metric tons. During the drought disaster of 2020, an outbreak of beet armyworms forced many farmers to abandon their harvests entirely. By 2023, production dropped to approximately 8,000 tons, with the estimated loss in output value exceeding NT$300 million.

In summary, the key challenges faced by scallion farmers include:

• High Cryptic Nature of Beet Armyworms Leading to Late Detection: Once the larvae burrow into the scallion tubes, the exterior looks almost perfectly normal. Traditional field scouting makes it extremely difficult to detect them in time.
• High Pest Density, Reduced Chemical Efficacy, and Soaring Control Costs: The high population density combined with developed pesticide resistance means that even frequent spraying yields limited results, leading to a continuous increase in long-term expenses.
• High Operational Barriers for Control Methods: Even though farmers know that nighttime water spraying is effective, implementing it manually is incredibly difficult because it requires specific grouping, precise timing, strict time limits, and avoiding rainy periods.
• Scattered Management Information Leading to Heavy Decision-Making Burdens: Farmers have to check weather forecasts, calculate pest numbers, and schedule irrigation intervals every single day—a process that is both time-consuming and labor-intensive.



・・・



The Hualien District Agricultural Research and Extension Station once proposed a pesticide-free solution:

timed nighttime water spraying during peak beet armyworm activity periods.

By conducting short, staged irrigations on the hour from 7:00 PM to 1:00 AM, the damage rate can be effectively reduced from 23.1% down to 13.5%.

However, this method only works if all the following conditions are met simultaneously:

• Staged, short-duration irrigation must be conducted every hour on the hour from 7:00 PM to 1:00 AM.
• The total duration of water spraying must not exceed 20 minutes.
• If it rains or if pesticides have just been applied, the spraying schedule must be postponed for two days.

With such complex conditions, it is not that farmers are unwilling to do it, but rather that relying purely on manual labor makes it practically impossible to implement.



・・・



This time, we are attempting to transform HDARES's pest control recommendations into a system that can "run automatically" right inside the Sanxing scallion fields.

Co-developed by AgriWeather and the Hualien District Agricultural Research and Extension Station (HDARES), this system combines research-grade beet armyworm prediction models with our years of field experience in automatic control and sensing. The ultimate goal is to ensure that the model does not just predict, but actively "drives on-site actions."



・・・



1. The Prediction Model Calculates Risk: Each night, the system computes data from the Central Weather Administration (CWA) and on-site field sensors (such as temperature, humidity, and rainfall) to predict the risk of beet armyworm outbreaks over the next 10 days.
2. Condition Comparison Determines Activation: When the model identifies a "high-risk period," the system simultaneously cross-checks the following conditions:
   • Has it rained recently? (To avoid overwatering)
   • Have pesticides just been applied? (To avoid washing away pesticides)
3. Automatic Irrigation Logic Is Triggered: If all conditions are met, the automatic irrigation system activates between 7:00 PM and 1:00 AM. It waters according to preset zoned and staged logics, with each zone lasting no more than 5 minutes and the total irrigation duration not exceeding 20 minutes.
4. Post-Irrigation Logic Assessment: Once the sprinkler irrigation is complete, the system logs the irrigation session and automatically pauses nighttime irrigation for the next 2 days, unless a new risk threshold is reached again.
5. Fully Traceable Early Warnings and Execution: All prediction and execution records are pushed to farmers via the Agri-PestMaster App interface. This allows users to review the entire operational process and leverages the data for subsequent adjustments and analysis.

This design truly integrates "prediction" with "control"—not only reminding farmers to "pay attention," but automatically carrying out the actual pest control actions.

AgriWeather Early Warning System Video: The Ultimate Weapon for Scallion Farming—Predicting Pest Outbreaks 10 Days in Advance, Delivering Even Greater Impact When Paired with Automated Sprinklers!

👉 Free Trial: https://account.agriweather.com.tw/login
You can check out AgriWeather's tutorial video, "The Complete Guide to AI Early Warning System Operations." We will guide you through the process step-by-step, from registration and logging in to full system utilization.



・・・



We want you to know that AgriWeather does more than just manufacture sensors and irrigation controls; we also possess the independent capability to develop "climate-driven prediction models."

This collaboration on the beet armyworm began with historical pest data provided by HDARES, paired with meteorological data from the Central Weather Administration (CWA). After building the model, we integrated it into our automatic control platform, enabling the system to deliver automated early warnings and execute actions autonomously.

We can assist you with:

• Data Cleaning and Model Construction: Processing and analyzing your pest and disease survey data.
• Applying Predictions to On-Site Management Decisions: For instance, automatically triggering irrigation, pushing pest control notifications, or further linking predictions with field operational workflows.
• Establishing Demonstration Cases: Setting up cases in experimental fields or extension areas to rapidly validate technical feasibility and scaling potential.

If you have the data, we can turn it into a prediction model; if you have the model, we can turn it into an affordable tool for farmers.

If you are interested in collaborating or learning more, welcome to contact us!



・・・



Q1: What equipment needs to be installed on-site for this system?
A1: The basic package includes meteorological sensors (temperature, humidity, and rainfall) and an automatic irrigation system. This setup can be flexibly adjusted depending on the size and specific needs of the venue.

Q2: Does the prediction model require on-site pest population data?
A2: Yes. To ensure the model can be effectively trained and its accuracy validated, this prediction model must be built upon at least two or more years of historical pest population or disease outbreak records.

Q3: I do not grow scallions. Can I still use this system?
A3: Currently, the model is specifically validated for green scallions and beet armyworms; however, the system architecture is highly scalable. If you have a specific crop + pest control model, we can assist in integrating it into our sensing and control system.

Q4: Can we build the prediction model ourselves, or must we collaborate with AgriWeather?
A4: If you are a research institution, university, or agricultural research and extension station and already have your own self-built model, we can help embed your model into our automatic control system. You are not restricted to using our existing models, and we highly welcome joint model development.



・・・



Learn more interesting stories about scallions and the beet armyworm?

👉 May 2026 — Beet Armyworm Control Guide for Scallions: 8 Methods at a Glance