HSS misbehaves for certain data types.

[Debasish-Mahapatra]

Executing HSS gives the following Traceback,

433 def compute_bss(self, threshold, dim=None): 434 """ --> 
435 Compute the Brier Skill Score (BSS) for a given threshold. 
436 
437 Args: 
438 threshold (float): The threshold value for binary classification. 
439 dim (str, list, or None): The dimension(s) along which to compute the BSS. 
440 If None, compute the BSS over the entire data. 
441 
442 Returns: 
443 xarray.DataArray: The computed BSS values. 
444 """ 
445 # Convert data to binary based on the threshold 
446 obs_binary = (self.obs_data >= threshold).astype(int) 

...

2299 ) 
2300 result = Variable(dims, new_data, attrs=attrs) 
2301 return result

TypeError: numpy boolean subtract, the - operator, is not supported, use the bitwise_xor, the ^ operator, or the logical_xor function instead.

@harry9713 any idea why, can you check this a bit as well?

[Debasish-Mahapatra]

Scratch that, the error seems to come from here,

 tn, fp, fn, tp = self.confusion_matrix(obs_binary, model_binary, dim)
    # Calculate the HSS
---> hss = 2 * (tp * tn - fp * fn) / ((tp + fn) * (fn + tn) + (tp + fp) * (fp + tn))
    return hss

which actually makes sense.