Quantitative Rigor

Being quantitative with environmental data can be challenging at times, especially when the assumptions implicit in many common statistical test are violated. Although not trained as a statistician, I recognize the importance of seeking the best tools available in order to be quantitatively rigorous. This includes reporting uncertainty (error) whenever possible and often seeking advice from those more knowledgeable.

Ecosystem Modeling

Linear Inverse Models

LIMs solve the inverse problem by constraining an otherwise free solution space based on mass balance, theoretical bounds and observed values. Compared to forward models, LIMs take an agnostic perspective on parameterizing transfer functions--and therefore avoids some of the issues related to tuning even simple models.

Net Community Production

NCP can be estimated independently of direct biological measurements through concentrations of dissolved oxygen and argon. These "mechanistic" models of air-sea gas exchange can provide valuable constraints of the ecosystem and export dynamics.

Trace Metals

Sea Surface Microlayer

Finite difference schema were used to solve 3 coupled PDEs dictating the transformation and sinking of trace elements through the water column. Kinetic rate coefficients were tuned from laboratory data and particulate settling velocities were estimated from Stokes' Law. Model included 46 size classes and three phases for each TE (dissolved, reactive particulate and refractory particulate; 12 TEs).


Glacial Accretion/Retreat

Implementing the NASA Ice Sheet System Model, we calculated glacial accretion and retreat over two seasons using field-site data on the Greenland Ice Sheet. The model included snowpack, ice, sublimation and refreezing processes, which were parameterized against Positive Degree Days (PDD) per the documentation.



  • R
  • Matlab
  • Julia
  • Python


  • Non-parametric Bootstrapping
  • Monte Carlo Markov Chain
  • Finite Difference
  • Remote Sensing

Client Testimonials