The older SNOTEL data changed at current week

Snowcast Showdown
1

Hello all
I’m comparing the SNOTEL data for 2022-03-03, 2022-02-24, 2022-02-17 loaded from AWS with an inference from 2022-03-03 and with an inference from 2022-03-10.
And I found some differences.
This is very strange. Is this normal?
Could it be that different participants who ran the inference at different time got different SNOTEL data?

                       2022-03-03   2022-03-03
index            loaded 2022-03-03  loaded 2022-03-10                                 
SNOTEL:1054_UT_SNTL        11.2        10.2
SNOTEL:1214_UT_SNTL         4.8         4.7
SNOTEL:1223_UT_SNTL         9.5         9.1
SNOTEL:1280_UT_SNTL        12.2        11.7
SNOTEL:329_UT_SNTL          7.1         6.9
SNOTEL:356_CA_SNTL         18.7        17.4
SNOTEL:508_CA_SNTL         14.5        15.1
SNOTEL:541_CA_SNTL         35.6        35.0
SNOTEL:568_MT_SNTL          NaN         3.6
SNOTEL:575_CA_SNTL          8.8         7.8
SNOTEL:649_MT_SNTL          NaN        19.0
SNOTEL:684_UT_SNTL         11.1        11.0
SNOTEL:690_MT_SNTL          NaN         1.0
SNOTEL:723_UT_SNTL          8.4         8.7
SNOTEL:795_UT_SNTL         10.1         9.9
SNOTEL:814_UT_SNTL         14.8        14.2
SNOTEL:820_UT_SNTL         12.7        12.4
SNOTEL:864_UT_SNTL          8.4         8.3
SNOTEL:896_UT_SNTL         11.3        11.1
SNOTEL:902_AZ_SNTL          2.2         1.8

                       2022-02-24   2022-02-24
index            loaded 2022-03-03  loaded 2022-03-10 
SNOTEL:1010_OR_SNTL         9.5         8.7
SNOTEL:1016_ID_SNTL         6.7         7.1
SNOTEL:1048_NM_SNTL         0.1         0.0
SNOTEL:1051_CA_SNTL        15.9        16.5
SNOTEL:1052_CA_SNTL        12.2        12.4
                        ...         ...
SNOTEL:915_ID_SNTL          7.7         8.0
SNOTEL:922_NM_SNTL          NaN         7.7
SNOTEL:926_ID_SNTL         14.2        14.4
SNOTEL:936_CO_SNTL          3.6         3.5
SNOTEL:945_OR_SNTL          0.1         0.0
[159 rows x 2 columns]

                       2022-02-17   2022-02-17
index             loaded 2022-03-03  loaded 2022-03-10 
SNOTEL:1005_CO_SNTL         6.7         6.6
SNOTEL:1010_OR_SNTL         9.2         8.8
SNOTEL:1048_NM_SNTL         0.1         0.0
SNOTEL:1051_CA_SNTL        15.3        15.9
SNOTEL:1061_CO_SNTL         6.3         6.2
SNOTEL:1065_UT_SNTL        10.7        10.6
SNOTEL:1066_UT_SNTL         8.1         8.2
SNOTEL:1129_WA_SNTL        11.7        11.4
SNOTEL:1138_NM_SNTL         6.1         6.2
SNOTEL:1169_NM_SNTL         3.9         4.0
SNOTEL:1172_NM_SNTL         7.7         7.8
SNOTEL:316_NM_SNTL          7.8         7.9
SNOTEL:327_CO_SNTL         11.7        11.5
SNOTEL:356_CA_SNTL         16.7        16.8
SNOTEL:449_WY_SNTL         10.7        10.8
SNOTEL:452_UT_SNTL          4.0         4.2
SNOTEL:486_NM_SNTL          1.5         1.6
SNOTEL:488_AZ_SNTL          4.2         4.8
SNOTEL:508_CA_SNTL         13.9        14.2
SNOTEL:532_NM_SNTL         10.2        10.3
SNOTEL:555_WY_SNTL          6.9         7.2
SNOTEL:577_WY_SNTL         18.5        18.7
SNOTEL:580_CO_SNTL          9.9        10.0
SNOTEL:597_WY_SNTL          8.8         8.9
SNOTEL:661_WY_SNTL          5.8         6.0
SNOTEL:665_NM_SNTL          2.2         2.3
SNOTEL:677_ID_SNTL          6.7         6.2
SNOTEL:694_UT_SNTL         11.9        12.2
SNOTEL:708_NM_SNTL          NaN         4.6
SNOTEL:709_CO_SNTL         15.0        14.9
SNOTEL:749_ID_SNTL          8.1         8.2
SNOTEL:755_NM_SNTL          0.1         0.0
SNOTEL:762_CO_SNTL          6.9         6.8
SNOTEL:764_WY_SNTL          9.2         9.4
SNOTEL:765_WY_SNTL          7.1         7.5
SNOTEL:779_WY_SNTL         14.9        15.0
SNOTEL:816_WY_SNTL         11.0        11.2
SNOTEL:831_WY_SNTL         14.8        15.0
SNOTEL:834_CA_SNTL         13.7        14.2
SNOTEL:837_WY_SNTL         20.0        20.1
SNOTEL:848_CA_SNTL         23.0        22.8
SNOTEL:861_AZ_SNTL          1.5         0.4
SNOTEL:868_WY_SNTL         16.0        16.4
SNOTEL:895_ID_SNTL          8.7         8.8
SNOTEL:902_AZ_SNTL          2.5         2.1
SNOTEL:920_SD_SNTL          4.1         4.4
SNOTEL:922_NM_SNTL          NaN         7.7
SNOTEL:927_AZ_SNTL         13.6        13.5
2 Likes
2

Thats interesting… it seems the diff can be wild, e.g. from NaN to 19.0!, that seems like a huge difference, in my world we call that a bug :slight_smile: with big impact on the model.

3

One of the features of my model is based on historical measurements from SNOTEL/CDEC and I noticed an anomaly where many stations would deviate by the exact same relative amount from their historical measurements, in a way that was statistically implausible. It made me think that there might be some corrective factor applied to the data of some stations, but I didn’t follow up on it.

4

This has been around for a while. I guess it has to do with data availability, maybe some stations report data later than others. Anyway, I think this data changes only three (or more) days after the submission date. Therefore, it is not possible that for a particular submission, two people have used different SNOTEL/CDEC data. My advice is always to use the last AWS file if your model is based on historical measurements.

5

@Galeros93 is correct – it’s possible for stations to add or update data. The ground measures features csv is updated once weekly, meaning that everyone will be using the same file for their submission.

6

where is train_labels.csv in data tab. I could not find it. I found ground measures train, test features data.