← Back to Home

🌿 Pesticide Database Information

Comprehensive information about the pesticide database, definitions, API, and accuracy assessment

Database Background

Overview

This pesticide database is a comprehensive collection of EPA-registered pesticides, with crop and pest specific information availabile in a structured format. The purpose of this data is to improve the accessibility of detailed and structured pesticide information. The database contains detailed information extracted from official pesticide labels, including application rates, safety information, and crop-specific usage guidelines.

Data Sources

  • EPA APPRIL Registration Database: Primary source of pesticide registration information
  • EPA-approved Pesticide Labels: Detailed application information extracted from official PDF labels
  • FRAC, HRAC, and IRAC codes: Mode of action codes sourced from the FRAC, HRAC, and IRAC websites

Processing Methodology

The database was created using advanced natural language processing techniques to extract structured information from unstructured pesticide label text. This involves:

  • OCR processing of PDF labels for text extraction
  • AI-powered information extraction and categorization
  • Manual validation and quality control processes
  • Regular updates to maintain current information
Note: This database is automatically updated weekly with the latest actively registered pesticides in the US EPA database. A large language model was used to extract pesticide usage information from the PDF of each pesticide label along with crop-specific information for a selection of specialty crops.

Definitions & Terminology

Key Terms

Active Ingredient

The chemical component of a pesticide that provides the pesticidal activity. This is the primary active chemical that controls the target pest or disease.

Mode of Action (MOA)

The specific biochemical mechanism by which a pesticide affects the target organism. MOA is classified using the FRAC (Fungicide Resistance Action Committee), HRAC (Herbicide Resistance Action Committee), or IRAC (Insecticide Resistance Action Committee) system for fungicides and similar systems for other pesticide types.

REI (Restricted Entry Interval)

The time period after pesticide application during which entry into the treated area is restricted. This is a safety measure to protect workers from exposure to the pesticide.

PHI (Pre-Harvest Interval)

The minimum time that must elapse between the last pesticide application and harvest. This ensures that pesticide residues are below established tolerance levels.

Application Rate

The amount of pesticide product applied per unit area (e.g., pounds per acre, ounces per gallon). This includes both low and high rate ranges as specified on the label.

Signal Word

A standardized word (DANGER, WARNING, or CAUTION) that appears on pesticide labels to indicate the relative toxicity of the product.

PPE (Personal Protective Equipment)

Equipment worn to minimize exposure to hazards, including gloves, respirators, protective clothing, and eye protection as required by the pesticide label.

API Documentation

The pesticide database provides a RESTful API for programmatic access to the data. All endpoints return JSON responses.

Base URL

https://cosseboomlab.com/api/

Available Endpoints

Get All Pesticides

GET /pesticides?page=1&per_page=50

Retrieves a paginated list of all pesticides in the database.

  • page: Page number (default: 1)
  • per_page: Number of results per page (default: 50, max: 500)
Examples:
Get first 50 pesticides:
GET https://cosseboomlab.com/api/pesticides?page=1&per_page=50
Get second page with 100 pesticides:
GET https://cosseboomlab.com/api/pesticides?page=2&per_page=100

Search Pesticides

GET /search?q=search_term&type=both&page=1&per_page=200

Search pesticides by any detail including EPA number, trade name, ingredient, crop, pest, company, etc.

  • q: Search query (required)
  • page: Page number (default: 1)
  • per_page: Number of results per page (default: 200)
Examples:
Search for Captan:
GET https://cosseboomlab.com/api/search?q=Captan&type=both&page=1&per_page=50
Search for EPA number 100-1000:
GET https://cosseboomlab.com/api/search?q=100-1000&type=both&page=1&per_page=10

Get Specific Pesticide

GET /pesticide/{epa_reg_no}

Retrieves detailed information for a specific pesticide by EPA registration number.

Examples:
Get pesticide with EPA number 100-1000:
GET https://cosseboomlab.com/api/pesticide/100-1000
Get pesticide with EPA number 100-1014:
GET https://cosseboomlab.com/api/pesticide/100-1014

Filter by Crop and Pest

GET /filter?crop=Apple&pest=Bitter%20Rot&page=1&per_page=100

Filter pesticides by specific crop and pest combinations.

  • crop: Target crop name
  • pest: Target disease or pest
  • page: Page number (default: 1)
  • per_page: Number of results per page (default: 100)
Examples:
Filter for Apple and Bitter Rot:
GET https://cosseboomlab.com/api/filter?crop=Apple&pest=Bitter%20Rot&page=1&per_page=50
Filter for Grape and Powdery Mildew:
GET https://cosseboomlab.com/api/filter?crop=Grape&pest=Powdery%20Mildew&page=1&per_page=100

Get Available Crops

GET /enums/crops

Retrieves a list of all available crops in the database.

Examples:
Get all available crops:
GET https://cosseboomlab.com/api/enums/crops
Get all available crops (same endpoint):
GET https://cosseboomlab.com/api/enums/crops

Get Pests for Crop

GET /enums/pests?crop=Apple

Retrieves a list of pests/diseases available for a specific crop.

Examples:
Get pests for Apple:
GET https://cosseboomlab.com/api/enums/pests?crop=Apple
Get pests for Grape:
GET https://cosseboomlab.com/api/enums/pests?crop=Grape

Response Format

All endpoints return JSON responses with the following structure:

{
  "results": [...],
  "total": 15458,
  "pagination": {
    "page": 1,
    "per_page": 50,
    "has_next": true,
    "total_pages": 310
  }
}

Rate Limiting

API requests are limited to 1000 requests per hour per IP address. Exceeding this limit will result in a 429 Too Many Requests response.

Authentication

Currently, the API is publicly accessible. Future versions may require API keys for certain endpoints.

Accuracy Assessment

Data Quality Metrics

98%
Mode of Action Accuracy
99%
Pesticide Rate Accuracy (rate and units)
100%
Restricted-Entry Interval (REI) Accuracy
95%
Pre-Harvest Interval (PHI) Accuracy

More specifically, the AI pesticide database was compared to both the pesticide labels as well as an expert-entered database that is utilized in the educational application MyIPM. The accuracy of seven variables was evaluated including: mode of action, pest listed correctly, low rate, high rate, rate units, PHI, and REI. The data was rated as either correct, incorrect or missing.

Accuracy Assessment Results

Table. Summary of AI-extracted database data accuracy for the seven variables that were evaluated
Variable Incorrect Missing Accuracy (including missing) Accuracy (excluding missing)
Mode of Action 2% 4% 94% 98%
Pest Listed Correctly 0% 12% 88% 100%
Low Rate 2% 11% 87% 98%
High Rate 0% 11% 89% 100%
Rate Units 0% 9% 91% 100%
PHI 4% 13% 83% 95%
REI 0% 7% 93% 100%
Table. Summary of MyIPM app data accuracy for the seven variables that were evaluated
Variable Incorrect Missing Accuracy (including missing) Accuracy (excluding missing)
Mode of Action 0% 0% 100% 100%
Pest Listed Correctly 2% 0% 98% 98%
Low Rate 11% 3% 86% 89%
High Rate 5% 0% 95% 95%
Rate Units 5% 0% 95% 95%
PHI 5% 8% 87% 95%
REI 7% 0% 93% 93%

📊 Download Detailed Accuracy Report

Validation Process

The database accuracy for crop and pest-specific information was assessed:

  • Manual Review: Random sampling of 100 pesticide entries for manual verification against the official pesticide label. Occassional errors were found in the AI extracted data for crop and pest specific information. On average the accuracy of the AI extracted data was 99% accurate, while a separate, expert-entered database was 95% accurate.
  • Missing Data: The major errors found during the assessments was missing data (for example, a missing crop or target that was on the label, but was missed by the LLM). If missing data was included in accuracy calculations, the accuracy of the AI extracted database was 89% accurate, while the expert-entered database was 93% accurate.
  • Direct Cross-Reference: Multiple variables including EPA Registration numbers, active ingredients, and trade names were directly sourced from the EPA APPRIL database
  • Future: Expert Feedback: Continuous improvement based flagged inaccuracies, or corrected data from experts

⚠️ DISCLAIMER

This is a pesticide database built from AI interpretation of pesticide labels and may not be accurate. This information is not a substitute for pesticide labeling. Always read and understand the product label before using any pesticide.