Give credit where credit is due! (erf-model#1760)

Co-authored-by: Aaron Lattanzi <[email protected]>
AMLattanzi · Aug 28, 2024 · a646f6f · a646f6f
1 parent e027c26
commit a646f6f
Show file tree

Hide file tree

Showing 2 changed files with 21 additions and 1 deletion.
diff --git a/Docs/sphinx_doc/Inputs.rst b/Docs/sphinx_doc/Inputs.rst
@@ -741,6 +741,10 @@ List of Parameters
 |                                  | advection scheme   |                     |              |
 |                                  | for scalars        |                     |              |
 +----------------------------------+--------------------+---------------------+--------------+
+| **erf.use_mono_adv**             | Use order reduction| true/false          | false        |
+|                                  | for scalar         |                     |              |
+|                                  | boundedness        |                     |              |
++----------------------------------+--------------------+---------------------+--------------+
 
 The allowed advection types for the dycore variables are
 "Centered_2nd", "Upwind_3rd", "Blended_3rd4th", "Centered_4th", "Upwind_5th", "Blended_5th6th",
@@ -762,6 +766,12 @@ schemes are as follows when using efficient advection option: roughly 30% for Ce
 and Centered_6th, 35% for Upwind_5th, roughly 45% for WENO5 and WENOZ5, and roughly 60% for
 Upwind_3rd, WENO3, WENOZ3, and WENOMZQ3.
 
+The monotonic advection option is an order reduction technique adapted from the PINACLES
+software developed at PNNL by K. Pressel et al.; see `pnnl/pinacles github <https://github.com/pnnl/pinacles>`_.
+When this flag is enabled, ERF will compute global mins and maxes for the scalar variables
+and then test whether the selected advection operator (e.g., centered, upwind, or WENO)
+will break these bounds. If boundedness is broken, the fluxes are recomputed with a
+0-th order upwind approach.
 
 
 Diffusive Physics

diff --git a/Source/Advection/AdvectionSrcForState.cpp b/Source/Advection/AdvectionSrcForState.cpp
@@ -255,7 +255,17 @@ AdvectionSrcForScalars (const Real& dt,
         }
     }
 
-    // Monotonicity preserving order reduction for SLOW SCALARS (0-th upwind)
+    /* =======================================================================
+       Monotonicity preserving order reduction for scalars (0-th upwind).
+
+       NOTE:
+       The following order reduction operations for scalar advection have been
+       adapted from the PINACLES code developed at PPNL by K. Pressel et al.;
+       see https://github.com/pnnl/pinacles.git and  This version utilizes global
+       min/max values for simplicity rather than the local average around each
+       cell. Motivation for this modification is the compressible dycore in ERF
+       as opposed to incompressible/analestic in PINACLES.
+       ======================================================================= */
     if (use_mono_adv) {
         // Copy flux data to flx_arr to avoid race condition on GPU
         ParallelFor(bx, ncomp, [=] AMREX_GPU_DEVICE (int i, int j, int k, int n) noexcept