Projecting Agricultural Shifts using Gaussian Processes and K-Means Clustering

My goal with this project was to better understand how our shifting climate might affect one of society’s most foundational industries: agriculture. Using remote sensing data from the World Soil Information Service (WoSIS) in conjunction with other public-access datasets, I implemented a stochastic mathematical model called a Gaussian process, employing Cholesky decomposition for efficient computation. It’s capable of predicting values of nitrogen, phosphorus, and potassium in soil over decades-long timescales. The WoSIS data was quite messy, and any fellow data science folks reading this know that combining multiple datasets always requires some tinkering.

You can read the paper and check out the associated code here!

• Skills: Mathematical modeling, time series data, EDA/ETL, data processing/visualization
• Tools: scikit-learn, Pandas, NumPy