Monitored Natural Attenuation (MNA) is gaining prominence as an effective method for managing soil and groundwater contamination due to its cost-efficiency and minimal environmental disruption. Despite its benefits, MNA necessitates extensive groundwater monitoring to ensure that contaminant levels decrease to meet safety standards. This study expands the capabilities of PyLEnM, a Python package designed for long-term environmental monitoring, by incorporating new algorithms to enhance its predictive and analytical functionalities. We introduce methods to estimate the timeframe required for contaminants like Sr-90 and I-129 to reach regulatory safety standards using linear regression and to forecast future contaminant levels with the Bidirectional Long Short-Term Memory (Bi-LSTM) networks. Additionally, Random Forest regression is employed to identify factors influencing the time to reach safety standards. Our methods are illustrated using data from the Savannah River Site (SRS) F-Area, where preliminary findings reveal a notable downward trend in contaminant levels, with variability linked to initial concentrations and groundwater flow dynamics. The Bi-LSTM model effectively predicts contaminant concentrations for the next four years, demonstrating the potential of advanced time series analysis to improve MNA strategies and reduce reliance on manual groundwater sampling. The code, along with its usage instructions, validation, and requirements, is available at: https://github.com/csplevuanh/pylenm_extension.

翻译：监测自然衰减（MNA）因其成本效益高且对环境影响最小，正日益成为管理土壤与地下水污染的有效方法。尽管具有诸多优势，MNA仍需进行广泛的地下水监测，以确保污染物浓度降至安全标准。本研究拓展了PyLEnM（一个专为长期环境监测设计的Python软件包）的功能，通过集成新算法以增强其预测与分析能力。我们引入了利用线性回归估算Sr-90和I-129等污染物达到监管安全标准所需时间的方法，并采用双向长短期记忆（Bi-LSTM）网络预测未来污染物浓度。此外，通过随机森林回归识别影响达标时间的因素。我们使用萨凡纳河场址（SRS）F区的数据对方法进行了说明，初步结果表明污染物浓度呈显著下降趋势，其变化与初始浓度及地下水流动力学相关。Bi-LSTM模型有效预测了未来四年的污染物浓度，证明了先进时间序列分析在优化MNA策略、减少人工地下水采样依赖方面的潜力。相关代码、使用说明、验证及要求详见：https://github.com/csplevuanh/pylenm_extension。