Heatwave identification misses negative temperature exceedances

[dkreg]

Hi Tammas,

Per my email a while ago, the first line of both the identify_hw(ehfs:np.ndarray) and identify_semi_hw() functions identifies days and locations for which a heat event occurs. That is,

events = (ehfs>0.).astype(int)

. The ehfs > 0 condition is correct when ehfs is a view into the excess heat factor array, since EHF values are non-negative. However, when ehfs is instead a view into either the tmax or tmin exceedance arrays, the condition ehfs > 0 causes negative exceedances to be ignored. (Naturally, exceedances are temperature values, e.g., the 90th percentiles, and could, in principle, be negative). This issue might not matter a great deal in practice, since, presumably, exceedances are usually positive.

A simple fix would be to use np.nan values to mask-off values that are to be ignored during heatwave calculations, namely, non-positive ehfs values and values of tmax and tmin that are below the cutoff percentile. The condition ehfs > 0 would then be replaced with ehfs != np.nan.

[tammasloughran]

There are two steps of changes that need to be made here:

• Ensure that tmax or tmin exceedance values contain nan when not exceeding the threshold
• Apply the correct identification operation to the two versions of identify_hw

As an aside, it's supposed to be bad practice in Python to say ehf != np.nan, and instead we should use np.logical_not(np.isnan(ehf)).

[tammasloughran]

Many thanks for this @dkreg! It's great to have another set of eyes on this.