Testcases show strange features

[jhidding]

strange feature, documented in testcases 1 and 5: It detects an event at the "edge" of the longitude dimension in event count, i.e. in the Pacific (see notebooks Testcase 1/5 Classification; Testcase 5 shows what I expect to be correct); This also occurs for a number of model grids and variables when scanning with hypercc, example: snw, ACCESS1-0. It seems to prevent a meaningful selection of thresholds because they don't appear in the signal plot, and other plots are empty / full of spurious noise.

strange feature in testcase 5 (for example): spatial edge of the tipping region is detected and counted several times. Should not happen because the spatial edge is below the lower threshold. Varying the upper threshold very slightly leads to abrupt emergence of this bug. Also, detected area is too small in any case. See Testcase 5 Classification for example, compare to Testcase5 which is correct.

[bathiany]

I think that these are two different (probably unrelated) problems. I update info on the second part of this issue:
The spatial edge of the tipping region is detected and counted several times. I would not expect this to happen because the spatial edge is below the lower threshold. Varying the upper threshold very slightly leads to abrupt emergence of this result. Also, the spatial extent of the detected event around equator is too small.
Comparison between notebooks Testcase5 hypercc and Testcase5 reveals that there may be a bug in the hysteresis thresholding of hypercc? Testcase5.ipynb is correct, the hypercc version gives strange results.
I just uploaded both notebooks with the same parameter settings for comparison via github.