Count consecutive condition in pandas

Question

I am trying to count how many times in a row a condition is happening. I read about groupby.cumcount() but it doesn't really work the way I would like to. Here is a small part of the data:

                  min        max
time                            
1970-01-02  -3.440000  -1.180000
1970-01-03  -4.830000  -0.700000
1970-01-04  -6.250000   0.250000
1970-01-05 -11.700000  -6.690000
1970-01-06 -13.000000  -3.720000
1970-01-07  -3.870000   2.070000
1970-01-08   0.320000   2.690000
1970-01-09  -5.170000   2.310000
1970-01-10  -4.540000   1.140000
1970-01-11  -7.260000   1.300000
1970-01-12  -9.870000  -0.780000
1970-01-13  -6.520000  -0.390000
1970-01-14  -8.490000  -5.090000
1970-01-15 -13.670000  -8.670000
1970-01-16 -11.080000  -4.110000
1970-01-17 -24.770000  -7.320000
1970-01-18 -29.709999 -24.230000
1970-01-19 -24.200001 -19.480000
1970-01-20 -31.000000 -13.810000
1970-01-21 -36.389999 -30.209999
1970-01-22 -39.889999 -36.990002
1970-01-23 -41.750000 -38.730000
1970-01-24 -38.259998  -8.510000
1970-01-25 -14.100000  -5.740000
1970-01-26 -12.000000  -8.540000
1970-01-27 -12.060000  -7.470000
1970-01-28 -10.230000  -7.710000
1970-01-29 -10.850000  -8.400000
1970-01-30 -15.270000  -9.870000
1970-01-31 -11.920000  -5.290000

Considering a condition : df['min'] <= -30 and a period superior or egal to 3 days.

I would like to know how many times do we have at least three consecutive days with a 'min' under -30 in year.

So the result would be something like (dummy values) :

       occurences
time   
1970   3
1971   4
1972   2
1973   3

I have toyed with a few solutions but I can't get close, any suggestion?

Answer 1

IIUC:

In [94]: x = df[df.rolling(3)['min'].max() <= -30]

In [95]: x.groupby(x.index.year)['min'].count().to_frame('occurences')
Out[95]:
      occurences
1970           3