Comparative Evaluation of AMMI, GGE, GYT, and Bayesian Models for Risk-Based Selection of Forage Sorghum in multi-environmental trials

Imagine you are a chef trying to create the perfect recipe for a new kind of soup. You have a pantry full of different ingredients (the "genotypes"), but you also know that the soup will taste different depending on where you cook it. Maybe the water in your kitchen is hard, or the stove runs hot, or the humidity in the air changes how the vegetables cook. In the world of farming, this is called Genotype × Environment Interaction. It's the fancy way of saying that a plant that grows like a champion in one field might struggle in another, just because the weather, soil, or season is different.

Farmers and breeders want to find the "super-ingredients"—plants that are not only high-yielding (giving lots of food) but also stable. A stable plant is like a reliable friend who shows up and does a great job whether it's sunny, rainy, or windy. For a long time, scientists have used math to figure out which plants are the most reliable. They use tools that look at the average performance, kind of like calculating a student's grade point average. But averages can be tricky. A student who gets an A one week and an F the next has the same average as a student who gets a B every single day. The second student is more stable, but the math of the "average" doesn't always tell you that.

Recently, a new kind of math called Bayesian statistics has entered the kitchen. Instead of just giving you a single number (like a grade), it gives you a probability. It's like asking, "What are the odds this plant will be a star performer in any random future season?" This approach doesn't just tell you who is the best on average; it tells you who is the safest bet to avoid a total flop. This is crucial for crops like forage sorghum, which is a grass grown to feed cows and other livestock. If a farmer bets on the wrong plant and it fails, the animals go hungry. So, finding the plant that is both high-yielding and reliable is a high-stakes game.

---

The Great Plant Showdown: Who Wins the "Most Reliable" Trophy?

In this study, a team of scientists decided to throw a massive tournament to see which forage sorghum plants could handle the heat. They gathered 30 different types of sorghum (think of them as 30 different teams of athletes) and put them through the wringer. They didn't just test them in one place; they tested them across five different environments in India, spanning from the humid, rainy seasons in Assam to the drier, hotter conditions in Hyderabad. These tests happened over different seasons (winter, summer, and the monsoon) during 2020 and 2021.

The goal was simple: measure how much green fodder (the leafy food for animals) each plant produced. But the real challenge was figuring out which plants were the "champions" that could handle any of these five environments without crashing.

The Old School vs. The New School

To solve this puzzle, the researchers used two different teams of detectives.

Team 1: The Classic Detectives (AMMI, GGE, and GYT)
These are the traditional methods scientists have used for decades. They look at the data and draw colorful maps called biplots. Imagine these maps as a radar screen where plants are dots and environments are arrows.

• AMMI looks at how much a plant's performance wobbles when the weather changes.
• GGE tries to find the "ideal" environment and see which plants hug that ideal the closest.
• GYT is a bit more complex; it doesn't just look at how much food the plant makes, but also checks if the plant has other good traits, like thick stems or lots of leaves, to see if those traits help the yield.

Team 2: The Probability Wizards (Bayesian Models)
This team uses a different approach. Instead of just saying "Plant A is better than Plant B," they ask, "What is the probability that Plant A will beat Plant B?" They use a computer to run thousands of simulations, creating a "cloud" of possible outcomes for every plant. This gives them a risk score. It tells them not just who is the best, but who is the safest to bet on.

The Results: Who Are the Superstars?

After crunching the numbers, both teams of detectives pointed their fingers at the same few plants, which is a very good sign that the results are solid.

The clear winner, the plant that stood out as the most stable and high-yielding, was G24 (348B).

• The Classic Detectives said: "G24 is the most stable and high-yielding. It performed great in almost every environment."
• The Probability Wizards said: "We are very confident (about 96% probability) that G24 is a top performer, and it also has a high chance of being stable."

Other plants that made the "Elite List" included G30 (SSG-59-3), G26 (314B), and G19 (327B). These plants were consistently good across the board.

However, the study found something very interesting about a plant named G21 (CSV32F).

• If you only looked at the average yield, G21 looked like a superstar, right up there with G24.
• But the Probability Wizards revealed a secret: G21 was a "high-risk" player. It had a high chance of being great, but it also had a high chance of being terrible depending on the season. Its "stability probability" was very low (only about 0.08).
• The Lesson: Just because a plant has a high average score doesn't mean it's a safe bet. G21 is like a student who gets an A on a test they studied hard for but fails the one they didn't. G24, on the other hand, is the student who gets a solid A- every single time.

The "Risk Map"

The researchers created a special map (a four-quadrant risk map) to visualize this.

• Top Right (The Elite): Plants that are both high-yielding AND stable. G24 sat right here.
• Top Left (The Risky Gamble): Plants that might be huge winners but could also be total losers. G21 was here.
• Bottom Right (The Steady but Small): Plants that are safe but don't produce much food.
• Bottom Left (The Losers): Plants that are neither safe nor productive.

What About the Environments?

The study also figured out which test locations were the best judges.

• E1, E2, and E3 were found to be the "ideal" test environments. They were good at spotting the best plants and were representative of the real world.
• The study also discovered that the region could be split into two "mega-environments." This means that while some plants are great everywhere, others are specialists. For example, G3 was the best in two specific environments but not the others. This tells farmers that if they are in a specific type of climate, they might want to pick a specialist plant rather than a generalist one.

The "Secret Sauce" Traits

The researchers also looked at what made the winning plants so good. They found that plants with thicker stems (SGT), wider leaves (LFW), and a higher Leaf Area Index (LAI) (which basically means more leaf surface to catch sunlight) tended to produce more food. It turns out that if you want a lot of green fodder, you need a plant that builds a big, robust "solar panel" system.

The Bottom Line

This study didn't just find a few good plants; it proved that using both the old-school math and the new-school probability math gives you the best results. The old methods gave a clear picture of who was good, and the new methods gave a clear picture of how risky it was to pick them.

The final verdict? G24 (348B) is the plant to watch. It's the reliable, high-yielding champion that farmers can trust in different climates, including the tricky, non-traditional growing areas like Assam. G30, G26, and G19 are also strong contenders. But if you pick G21, you'd better be ready for a gamble.

By combining these two ways of looking at the data, the scientists have given breeders a powerful new tool to pick the right plants for a future where the weather is becoming more unpredictable. It's not just about finding the biggest plant; it's about finding the one that won't let you down when the storm hits.