*Altered initial conditions for circle_obcs

- The circle_obcs experiment ostensibly tests the Flather open
  boundary conditions except it was modified to use 10 layers
  and also test the Orlanski boundary conditions too. The initialization
  code appears to have been written for a 2-layer system and
  essentially only forces a very high-mode. This change produces a
  mixed mode initial condition (thickness anomaly linear with z) but
  with most energy in the first baroclinic mode.
- Changes circle_obcs experiment only.

src/user/circle_obcs_initialization.F90

@@ -93,8 +93,16 @@ subroutine circle_obcs_initialize_thickness(h, G, GV, param_file)
     ! if (rad <= 6.*diskrad) h(i,j,k) = h(i,j,k)+10.0*exp( -0.5*( rad**2 ) )
     rad = min( rad, 1. ) ! Flatten outside radius of diskrad
     rad = rad*(2.*asin(1.)) ! Map 0-1 to 0-pi
-    h(i,j,k) = h(i,j,k) + 10.0*0.5*(1.+cos(rad)) ! cosine bell
-    h(i,j,k-1) = h(i,j,k-1) - 10.0*0.5*(1.+cos(rad)) ! cosine bell
+    if (Nz==1) then
+      ! The model is barotropic
+      h(i,j,k) = h(i,j,k) + 1.0*0.5*(1.+cos(rad)) ! cosine bell
+    else
+      ! The model is baroclinic
+      do k = 1, Nz
+        h(i,j,k) = h(i,j,k) - 0.5*(1.+cos(rad)) & ! cosine bell
+            * 5.0 * real( 2*k-Nz )
+      enddo
+    endif
   enddo ; enddo
 
 end subroutine circle_obcs_initialize_thickness