WHAT ARE THE NEXT STEPS?

• Compute $\sqrt{k_r}$ in physical space and look at it. How best to do this? How best to package the software?
• Invent a synthetic data test.
• Think about how it might impact Sean (or Sergey). Which example of Sean's would be worth redoing?
• Extend this idea to 3-D (lettuce versus noodles).
• Matt Schwab and I were frequently disappointed in the performance of local PEFs for the task of visualizing data. This might explain it. Which example of his might be worth redoing?