Currently in point initialization, we force the final WindNinja field to match the input points using a outer iteration loop that basically loops over successive WindNinja simulations and changes the "input" points (hopefully in the correct direction), and then stop when the solution field is close enough to the input points.
There may be a way to eliminate this outer loop... Since the WindNinja method involves a cost minimization, we might be able to add another "cost" here, which would involve something like an inverse distance from each of the points. So essentially, we would make it very "costly" to modify the flow field at the points compared to areas far from points.
Someday we should test this. It wouldn't require a lot of additional coding (I think the distance cost would just be multiplied by the alpha costs for stability) and, if successful, would likely make the solution much faster and probably also much more stable.
Currently in point initialization, we force the final WindNinja field to match the input points using a outer iteration loop that basically loops over successive WindNinja simulations and changes the "input" points (hopefully in the correct direction), and then stop when the solution field is close enough to the input points.
There may be a way to eliminate this outer loop... Since the WindNinja method involves a cost minimization, we might be able to add another "cost" here, which would involve something like an inverse distance from each of the points. So essentially, we would make it very "costly" to modify the flow field at the points compared to areas far from points.
Someday we should test this. It wouldn't require a lot of additional coding (I think the distance cost would just be multiplied by the alpha costs for stability) and, if successful, would likely make the solution much faster and probably also much more stable.