Tests results of first five experiments on temperature networks

Problem networks of 10 series.
Five experiments:

  1. A few constant (2 °C) shifts are added to 5 series, the other 5 remaining as homogeneous references.
  2. Idem, but with a strong seasonal variation in the shifts.
  3. Idem, but with short term platforms and local trends, without seasonal variations.
  4. Any of the 10 series can be randomly affected by inhomogeneities, without seasonal variations.
  5. Idem, but with seasonal variations in the shifts.

Labels of tested methods:

Inh: Inhomogenous (problem) series
cl1: Climatol with constant correction
Cl1: Climatol with variable correction
Acs: ACMANT with sinusoidal correction
Aci: ACMANT with irregular correction
MSH: MASH without seasonality
RHa: RHtests absolute homogenization
RHA: RHtests absolute homogenization, with quantile adjustment
RHr: RHtests relative homogenization
RHR: RHtests relative homogenization, with quantile adjustment
US1: USHCN
Hoa: HOMER with d¬ j¬ c jc m c
Hob: HOMER with d¬ j¬ c jc c jc m c

Key to HOMER operations:

d: Pairwise detection (d¬: on raw/qc files; dc: on corrected files)
j: Joint detection (j¬: on raw/qc files; jc: on corrected files)
c: Correction
m: Assess Month of change

Tm1                         Tm2                         Tm3
Network cross-correlation
InhomogeneitiesRoot mean square errors
InhomogeneitiesTrend errors