Broken axis in Altair/Vega

Question

I have a Normalized Stacked Area Chart with huge differences between one variable and the others like:

df1=pd.DataFrame.from_dict(
    {'YEAR': {0: 2010,
      1: 2010,  2: 2010,  3: 2010,  4: 2011,  5: 2011,  6: 2011,  7: 2011,
      8: 2012,  9: 2012,  10: 2012,  11: 2012,  12: 2013,  13: 2013,  14: 2013,  15: 2013},
     'impact_FU': {0: 0.031479085164086554,  1: 5.9856927170853295e-05,  2: 1.1035885271638534e-05,  3: 5.8233509026863169e-06,
      4: 0.059176271387395112,  5: 0.00011179170132430088,  6: 1.9783914536689014e-05,  7: 1.0670218804040578e-05,
      8: 0.083935088170893221,  9: 0.00014806339884972569,  10: 2.3424374354037232e-05,  11: 1.30716950360811e-05,
      12: 0.10678138273474649,  13: 0.00016610749233828763,  14: 2.4764766148334989e-05,
      15: 1.3509464279754472e-05},
     'proc': {0: 'biogenic',  1: 'harvesting',  2: 'planting',  3: 'tending',
      4: 'biogenic',  5: 'harvesting',  6: 'planting',  7: 'tending',
      8: 'biogenic',  9: 'harvesting',  10: 'planting',  11: 'tending',
      12: 'biogenic',  13: 'harvesting',  14: 'planting',  15: 'tending'},
     'scenario': {0: 'BAU45',  1: 'BAU45',  2: 'BAU45',  3: 'BAU45',  4: 'BAU45',  5: 'BAU45',  6: 'BAU45',
      7: 'BAU45',  8: 'BAU45',  9: 'BAU45',  10: 'BAU45',  11: 'BAU45',  12: 'BAU45',  13: 'BAU45', 
                  14: 'BAU45',  15: 'BAU45'}})

Chart(df1).mark_area(stacked='normalize').encode(
    X('YEAR:T', timeUnit='year',),
    Y('sum(impact_FU)'),
    color=Color('proc:N'),
)

Is there any way in Altair/vega-lite/Vega to make broken y-axys...something like this?

Answer 1

Since you have very different value ranges for your data column, you can use a log scale transformation as follows. Altair comes with many scale transformations, and in your case, you'd use alt.Scale(type='log')

alt.Chart(df1).mark_area().encode(
    alt.X('YEAR:O'), 
    alt.Y('impact_FU', scale=alt.Scale(type='log')),
    color='proc:N',
)

which produces:

Altair is great for Facet charts, so you can keep the log scale and also facet on each proc with just one single extra line of code:

alt.Chart(df1).mark_area().encode(
    alt.X('YEAR:O'), 
    alt.Y('impact_FU', scale=alt.Scale(type='log')),
    color='proc:N',
    column='proc:N'
)

to get: