CSEP Quadtree issueshttps://git.gfz-potsdam.de/khawaja/csep-quadtree/-/issues2021-05-07T19:39:54+02:00https://git.gfz-potsdam.de/khawaja/csep-quadtree/-/issues/6Global forecast model evaluation with different Grid settings2021-05-07T19:39:54+02:00Muhammad Asim KhawajaGlobal forecast model evaluation with different Grid settingsI intend to acquire all the previously available global forecasts for 0.1x0.1 grid. Then I map them on Quadtree global grid. Mapping 0.1x0.1 onto L11 grid is computationally not achievable, so I map 0.1x0.1 onto L9. Then further aggregat...I intend to acquire all the previously available global forecasts for 0.1x0.1 grid. Then I map them on Quadtree global grid. Mapping 0.1x0.1 onto L11 grid is computationally not achievable, so I map 0.1x0.1 onto L9. Then further aggregate / De-aggregate in onto L5, L6, L7, L8, L10, L11 and EQ10L11.
The purpose of this exercise is to analyze the performance of forecasts with different grids. This will also serve as a "Discussion" part of S test paper.Muhammad Asim KhawajaMuhammad Asim Khawajahttps://git.gfz-potsdam.de/khawaja/csep-quadtree/-/issues/5Multi-resolution grid for California testing region2021-04-06T20:52:33+02:00Muhammad Asim KhawajaMulti-resolution grid for California testing regionRegular grid at L12 is acquired for as a Base grid for California testing region. The selection criterion for each cell L12 grid cell is that the center point of each cell should be lying inside the grid.
Once the L12 grid is acquired ...Regular grid at L12 is acquired for as a Base grid for California testing region. The selection criterion for each cell L12 grid cell is that the center point of each cell should be lying inside the grid.
Once the L12 grid is acquired as a base grid. Then multi-resolution grid is built upon it. Cells are aggregated in L12 grid to get bigger cells based on following criteria:
1) Nmax: If any cell has earthquakes less than Nmax, then aggregate it (go to its parent cell).
2) If the CORNER points of aggregated (parent) cell goes outside of California Polygon, then Go back to that higher Zoom-level (child cell)
Here is the outcome grids of the above approach:
![L12_base](/uploads/45cfbb210ecadab69250b4f8ec280e04/L12_base.png)
![EQ10_L12](/uploads/eb67148868b910bd3629d0e3902550d1/EQ10_L12.png)
![EQ50_L12](/uploads/a7c781eb69139620ecf5fc0e99e01a00/EQ50_L12.png)
![EQ100_L12](/uploads/8863a1c5200607f3aeadf5ddaa28f74e/EQ100_L12.png)
![EQ200_L12](/uploads/d7d2b85ab54ffa4523855f9bae0c41f9/EQ200_L12.png)
The gridding approach and its outcomes are acceptable? Looking forward to hear comments and suggestions about it.
CheersMuhammad Asim KhawajaMuhammad Asim Khawajahttps://git.gfz-potsdam.de/khawaja/csep-quadtree/-/issues/4Evaluating uniform forecast model2021-04-10T06:42:43+02:00Muhammad Asim KhawajaEvaluating uniform forecast modelWe found that S test powerless when the resolution of grid is increased too much (too many spatial cells) and we have less earthquakes. In this issue, we will explore the log-likelihood intervals and the reasons for its ineffectiveness f...We found that S test powerless when the resolution of grid is increased too much (too many spatial cells) and we have less earthquakes. In this issue, we will explore the log-likelihood intervals and the reasons for its ineffectiveness for higher resolution gridhttps://git.gfz-potsdam.de/khawaja/csep-quadtree/-/issues/2Analyzing the power of S-test2021-02-09T11:14:07+01:00Muhammad Asim KhawajaAnalyzing the power of S-testMuhammad Asim KhawajaSebastian HainzlDanijel SchorlemmerMuhammad Asim Khawaja