Application of classification-tree methods to identify nitrate sources in ground water.

Area of Science:

• Environmental Science
• Water Quality Analysis
• Statistical Modeling

Background:

• Nitrate contamination in groundwater poses risks to ecosystems and human health.
• Identifying specific nitrate sources is crucial for effective remediation strategies.
• Common sources include agricultural and golf course fertilizers, livestock waste, and septic systems.

Purpose of the Study:

• To develop and evaluate statistical classification-tree models for identifying nitrate sources in groundwater.
• To determine if these models can classify sources with at least 80% accuracy.
• To identify key predictor variables for distinguishing between different nitrate sources.

Main Methods:

• Collected 48 groundwater samples from five distinct nitrate source categories.
• Developed two classification-tree models using 32 and 4 variables respectively.
• Model 1 utilized delta 15N, nitrate to ammonia ratio, sodium to potassium ratio, and zinc.
• Model 2 used sodium plus potassium, nitrate to ammonia ratio, calcium to magnesium ratio, and sodium to potassium ratio.

Main Results:

• Both models successfully classified all five nitrate source categories with over 80% overall accuracy.
• Individual category success rates ranged from 71% to 100% using learning samples.
• An independent test set of 17 samples was 100% correctly classified using Model 2 for three categories.

Conclusions:

• Classification-tree models are highly effective for identifying groundwater nitrate contamination sources.
• The models successfully pinpointed key chemical variables indicative of specific nitrate origins.
• This approach demonstrates significant potential for environmental monitoring and source attribution.