Improving Growth Predictions in Aquaculture through an Improved Bioenergetics Model Incorporating Feed Composition and Nutrient Digestibility for Largemouth Bass (Micropterus salmoides)

This study presents a refined bioenergetics model for largemouth bass that incorporates feed composition and nutrient-specific digestibility coefficients, demonstrating significantly superior growth prediction accuracy compared to traditional gross energy-based models across both compiled and field datasets.

The Gist

Imagine you are trying to predict how fast a child will grow.

The Old Way (The "Calorie Counter" Model)
Traditionally, scientists used a simple calculator: "If the child eats 2,000 calories of anything, they will grow X amount." It didn't matter if those calories came from a candy bar, a steak, or a bowl of broccoli. The model just looked at the total number on the label.

For fish farmers, this was like using a generic rulebook. They would feed their fish a certain amount of food, measure the total energy in that food, and guess how big the fish would get. The problem? Fish are picky eaters. A fish can't turn a carrot into muscle the same way it turns a shrimp into muscle. The old model ignored what was in the food, only how much energy it had. As a result, the predictions were often way off—like guessing a child would grow 6 feet tall because they ate a lot of sugar, when in reality, they just got a cavity.

The New Way (The "Nutritionist" Model)
This paper introduces a smarter, more detailed approach for Largemouth Bass (a popular sport and food fish). Instead of just counting total calories, the researchers built a model that acts like a personal nutritionist.

Here is how the new model works, using simple analogies:

1. The "Ingredient Breakdown" (Digestibility)

Imagine you have two buckets of fuel.

• Bucket A is filled with premium, high-octane racing gas (high-quality protein).
• Bucket B is filled with muddy water mixed with some gas (low-quality filler).

The old model said, "Both buckets have 10 gallons of liquid, so the car will go the same distance."
The new model says, "Wait! The fish can only absorb 90% of the racing gas, but maybe only 40% of the muddy water. Let's calculate how much actual fuel gets into the engine."

The researchers added a step where they look at every single ingredient in the fish food (fish meal, soy, corn, etc.) and ask: "How much of this can the fish actually digest?" They call this the Digestibility Coefficient. It's like checking the "absorption rate" of every ingredient before calculating the energy.

2. The "Metabolic Tax" (Special Costs)

Different foods cost the fish different amounts of energy to process.

• Protein is like a heavy, complex package. The fish has to work hard to break it down, costing extra energy (like a high "tax" on the transaction).
• Fat is like a smooth, easy-to-digest gift. It costs less energy to process.

The new model accounts for this "tax." If the fish eats a diet high in protein, the model knows the fish will burn more energy just digesting it, leaving less energy for growing. The old model missed this entirely.

3. The "Growing Pains" (Body Size Matters)

The model also realized that a baby fish and a giant fish don't store energy the same way. A baby fish is mostly water and muscle; a big fish might have more fat. The new model adjusts the "energy density" of the fish as it grows, rather than assuming a fish is always the same "weight" of energy.

The Results: From Guessing to Knowing

The researchers tested this new "Nutritionist Model" against the old "Calorie Counter Model" using two groups of data:

1. A massive library of past studies (235 different experiments).
2. A real-life test in a fish farm in China with two different types of food.

The Outcome:

• The Old Model: It was like a weatherman who guessed "sunny" every day. It was okay for a general idea, but often wrong. In the real-world test, it failed miserably, predicting the fish would grow when they actually shrank or stayed the same.
• The New Model: It was like a meteorologist with a satellite. It predicted the fish's growth with 97-98% accuracy.

Why Does This Matter?

For fish farmers, this is a game-changer.

• Save Money: They can stop guessing which feed is best. They can use the model to design a diet that gives the most growth for the least cost.
• Save the Environment: If the fish can't digest the food, it poops it out, polluting the water. By predicting exactly what the fish can digest, farmers can reduce waste and keep the water clean.
• Better Fish: The fish grow healthier and faster because the food is perfectly matched to their biology.

In a nutshell: The researchers stopped treating fish food like a generic "energy block" and started treating it like a complex recipe. By understanding exactly how the fish digests every ingredient, they built a crystal ball that can predict fish growth with incredible precision.

Technical Summary

1. Problem Statement

Traditional bioenergetics models used to predict fish growth often rely on Gross Energy (GE) intake as a homogeneous input. This approach oversimplifies the biological reality by:

• Ignoring variations in macronutrient composition (protein, lipid, carbohydrate) across different feed formulations.
• Overlooking ingredient-specific digestibility coefficients (ADCs), treating all energy sources as equally available.
• Relying on parameters derived from low-energy, wild-forage diets (e.g., lake emerald shiners), which fail to accurately reflect modern intensive aquaculture conditions using high-energy, protein-dense commercial feeds.
• Assuming constant fish body energy density and neglecting the specific metabolic costs associated with digesting different nutrient types (e.g., the high Specific Dynamic Action (SDA) cost of protein).

Consequently, these models often exhibit significant prediction errors, systematic biases, and poor generalizability when applied to diverse feed formulations in controlled aquaculture systems.

2. Methodology

Data Collection

• Compiled Dataset: A comprehensive dataset of 235 records was aggregated from peer-reviewed literature and theses. It included growth performance, detailed feed compositions (protein, lipid, energy density >12 kJ·g⁻¹), and environmental conditions for Micropterus salmoides (Largemouth Bass).
• Field Experiment: A validation study was conducted in a Recirculating Aquaculture System (RAS) in Hebei, China, using two distinct dietary treatments (Diet A: 53% protein/8% lipid; Diet B: 43% protein/12% lipid) over 50 days.

Model Development

The study followed a two-stage modeling approach:

A. Baseline Model Optimization

• Framework: A classical energy budget equation ($E_{growth} = E_{gross} - RA - SDA - F - U$).
• Optimization: Parameters ($a_2, b_2, m, k_{SDA}, f, u$) were re-estimated using constrained nonlinear least-squares optimization (MATLAB fmincon with multi-start global search and L2 regularization) on the compiled dataset.
• Goal: To calibrate the model to modern high-energy diets before structural changes.

B. Refined Bioenergetics Model
The authors proposed a structural refinement to address the limitations of the GE-based approach:

1. Macronutrient-Resolved Digestible Energy: Instead of using total GE, the model calculates Digestible Energy ($E_{digestible}$) by decomposing feed into protein, lipid, and carbohydrate fractions, applying specific Apparent Digestibility Coefficients (ADCs) to each, and summing their energy contributions ($23.6, 39.5, 17.2$ kJ·g⁻¹ respectively).
2. Dynamic Body Energy Density: Replaced the constant assumption ($4.184$ kJ·g⁻¹) with a power function dependent on body weight ($E_{fish} = \alpha W^\beta$), better reflecting ontogenetic changes in body composition.
3. Revised Energy Losses:
   • Egestion ($F$): Removed from the equation as digestible energy is calculated directly.
   • Excretion ($U$) & SDA: Linked specifically to the digestible portion of nutrients rather than gross intake, acknowledging that metabolic costs arise from absorbed nutrients.
   • Carbohydrate Exclusion: Given the carnivorous nature of largemouth bass, carbohydrate contribution to energy was minimized/excluded in the final formulation.

Evaluation Strategy

• Calibration: 165 records used for parameter optimization; 70 records for independent testing.
• Validation: The field experiment data served as a robust external validation.
• Metrics: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Coefficient of Determination ($R^2$), and residual analysis for Specific Growth Rate (SGR) and Feed Conversion Ratio (FCR).
• Sensitivity Analysis: Local (One-At-a-Time) and Global (Monte Carlo) sensitivity analyses were performed to identify influential parameters.

3. Key Contributions

1. First Composition-Aware Framework for Largemouth Bass: This is the first bioenergetics model for M. salmoides that explicitly integrates feed formulation breakdown and nutrient-specific ADCs to compute macronutrient-resolved digestible energy.
2. Parameter Re-calibration for Modern Diets: The study successfully recalibrated respiration and metabolic parameters ($k_{SDA}$, $u$, etc.) for high-energy, protein-dense commercial feeds, revealing that metabolic costs shift from purely respiratory constraints to postprandial (SDA) and excretory costs in intensive systems.
3. Structural Refinement: By moving from a "Gross Energy" to a "Digestible Energy" framework and adopting a mass-dependent energy density, the model captures the physiological reality that different ingredients yield different usable energy, even at the same gross energy level.
4. Dual-Data Validation: The model was validated against both a large-scale literature dataset and a controlled field experiment, ensuring both generalizability and practical applicability.

4. Results

Baseline Model Performance:

• Initial (Literature Parameters): Poor accuracy ($R^2 = 0.62$, RMSE = 60.91 g), with significant overestimation of weight.
• Optimized (Calibrated Parameters): Significant improvement ($R^2 = 0.96$, RMSE = 20.68 g). However, it still failed to capture growth dynamics in the field experiment ($R^2 \approx 0$), indicating structural limitations remained.

Refined Model Performance:

• Compiled Dataset: Achieved $R^2 = 0.97$, with RMSE = 19.86 g and MAE = 10.31 g. This represented a 4.13% reduction in RMSE and 19.98% reduction in MAE compared to the optimized baseline model.
• Field Experiment (Critical Validation):
  • Diet A (53% Protein): $R^2$ improved from 0.33 (baseline) to 0.98.
  • Diet B (43% Protein): $R^2$ improved from 0.06 (baseline) to 0.97.
  • The refined model reduced RMSE by over 80% in the field setting, accurately capturing growth differences between the two diets which the baseline model missed entirely.

Sensitivity Analysis:

• The allometric coefficient for metabolic rate ($b_2$) remained the most influential parameter.
• Post-optimization, the sensitivity of SDA ($k_{SDA}$) and excretion ($u$) increased, reflecting the model's new ability to distinguish between metabolic costs driven by diet composition rather than just body size and temperature.

5. Significance and Implications

• Precision Aquaculture: The model enables formulation-aware growth prediction, allowing farmers and researchers to simulate how changing ingredient ratios (e.g., substituting fish meal with plant proteins) will impact growth and waste without costly trial-and-error.
• Nutrient Management: By accurately modeling digestible energy and nitrogenous waste (via excretion coefficients), the model supports Total Ammonia Nitrogen (TAN) estimation, crucial for water quality management in Recirculating Aquaculture Systems (RAS).
• Sustainability: The framework facilitates the optimization of feed efficiency and the reduction of environmental impacts (nitrogen/phosphorus discharge) by identifying the most efficient nutrient profiles for specific growth stages.
• Scalability: While currently specific to Largemouth Bass, the methodology provides a template for refining bioenergetics models for other species, moving the field from empirical curve-fitting toward mechanistic, nutritionally explicit modeling.

In conclusion, this study demonstrates that decoupling feed composition from energy intake and incorporating nutrient-specific digestibility is essential for accurate growth prediction in modern, high-intensity aquaculture systems.