Dataset for the article: "Assessing climate change impacts for small-scale fisheries in the Gulf of California using Deep Learning"

CC-DLM-MFisheries Project Directory Structure

Data Directories

/activation_graphs/

Neural network activation analysis files

activation_values_*.csv: Time series of neuron activations for each model

Purpose: Contains activation values used for analyzing model behavior

Format: CSV files with neuron IDs, layer information, and activation values

/data/

Core data storage

data.csv: Historical fisheries catch data and environmental variables

forecast_data/:

Contains transformed temperature data by cluster

Files format: transformed_cluster_temp_yrXXXX_clustY_rZiWpVfU.csv

Years range: 2025-2084

Clusters: 0-7

future_data.csv: Aggregated future temperature projections

/Graficas/

Visualization outputs

Boxplots of predictions and historical data

Species-specific visualization files

Cluster analysis graphs

Naming convention: boxplot_[species]_cluster_[number].png

/modelos_moe/

Trained models and associated files

[species]_cluster_[number]_moe_model.h5: Neural network model files

[species]_cluster_[number]_moe_scaler.pkl: Data scalers for each model

Model types: LSTM, RNN, CNN implementations

Purpose: Production-ready models for forecasting

/resultados_moe/

Model outputs and predictions

[species]_cluster_[number]_predictions.csv: Forecast results

SHAP analysis results

Model performance metrics

Validation results

/scripts/

Utility scripts and functions

prepare_future_data.py: Data preparation for forecasting

prepare_data.py: General data preprocessing

shap_analysis.py: Model interpretation tools

Purpose: Reusable code components for data processing

/visualizacion_resultados/

Analysis visualization outputs

heatmap_visualization.png/svg: Neural activation patterns

detailed_heatmap_visualization.png/svg: Detailed analysis views

activation_forecast_relationship.png/svg: Activation-forecast correlations

Purpose: High-quality visualizations for analysis and presentation

Analysis Files (Root Directory)

Notebooks

model_opener_2.ipynb: Neural network activation analysis

Prepare_data.ipynb: Data preparation pipeline

SHAP2.ipynb: SHAP value analysis for model interpretation

Forecast_one_species.ipynb: Single species forecasting

INTEGRATE.ipynb: System integration

general_analysis1.ipynb: Statistical analysis

SHAP_Analysis.ipynb: Feature importance analysis

Scripts

plots_future.py: Forecast visualization utilities

Purpose: Main analysis entry points and experimental code

Data Structures

Temperature Measurements

Standard depths (meters):

Copiar

6.00, 7.93, 9.57, 11.40, 13.47,
15.81, 18.50, 21.60, 25.21, 29.44

Aggregated Metrics

Mean temperature (10m)

Mean temperature (30m)

Cluster-specific metrics

Model Types

LSTM (Model 12): Temporal pattern analysis

RNN (Model 13): Sequential data processing

CNN (Model 14): Spatial pattern recognition

File Naming Conventions

Model Files

Format: [species]_cluster_[number]_moe_[type].[extension] Example: BANDERA_cluster_1_moe_model.h5

Temperature Data

Format: transformed_cluster_temp_yr[year]_clust[number]_[scenario].csv Example: transformed_cluster_temp_yr2058_clust0_r1i1p1f1.csv

Results

Format: [species]_cluster_[number]_[type].[extension] Example: CAMARON_cluster_1_predictions.csv

Project Dependencies

Python 3.x

TensorFlow for deep learning

Pandas & NumPy for data processing

Matplotlib & Seaborn for visualization

SHAP for model interpretation

Scikit-learn for preprocessing