Extremely high value of ERA5 monthly total precipitation

Datasets and Usage C3S - Datasets and Usage
1

Dear all,

I’m confused when working on weather data for Kenya using ERA5 hourly data. I first computed monthly values myself from ERA5 hourly data following the method described in:
https://confluence.ecmwf.int/pages/viewpage.action?pageId=197702790

Specifically, I used ERA5 hourly total precipitation and aggregated it to monthly values using the recommended approach. After doing this, I found that some show surprisingly high monthly precipitation values, for example:

country year month lat lon tp_monthly_mm
Kenya 2014 7 1.25 35.25 1179.508
Kenya 2014 6 1.25 35.5 1436.107
Kenya 2014 10 1.25 35.5 1039.039
Kenya 2014 8 1.25 35.5 2039.74
Kenya 2014 8 1.25 35.5 2039.74

This implies that some regions reach around 2000 mm in a single month, which is quite surprising to me. It feels extremely high. If this were annual precipitation, the magnitude would seem more reasonable, but as monthly totals I am unsure whether this is realistic for this region.

To double-check, I also used ERA5 monthly data instead of aggregating hourly data myself. I perform the calculations in Stata:

* Calculate monthly total precipitation by converting from daily values
rename tp tp_perday

* Create number of days in each month
gen d = .
replace d = 31 if inlist(month, 1, 3, 5, 7, 8, 10, 12)
replace d = 30 if inlist(month, 4, 6, 9, 11)
replace d = 28 if month == 2

* adjustment for February
replace d = 29 if month == 2 & inlist(year, 2012, 2016, 2020)

* Convert to monthly total precipitation (mm)
gen tp = tp_perday * 1000 * d

The results are very similar to my hourly-based aggregation, which suggests that my calculation procedure is consistent.

country year month lat lon tp_monthly_mm tp
Kenya 2014 7 1.25 35.25 1179.508 1179.6
Kenya 2014 6 1.25 35.5 1436.107 1434.402
Kenya 2014 10 1.25 35.5 1039.039 1038.994
Kenya 2014 8 1.25 35.5 2039.74 2039.673
Kenya 2014 8 1.25 35.5 2039.74 2039.673

However, I still have some concerns:

  1. Does this mean that ERA5 total precipitation can realistically reach ~2000 mm/month in some locations in Kenya?
  2. Or could these high values be related to known biases or characteristics of ERA5 precipitation?
  3. I could not find any reports of extreme flooding events corresponding to such high values, so I am unsure whether this magnitude is climatologically reasonable.

I would appreciate any clarification on whether this level of monthly precipitation is expected in ERA5 for East Africa, or whether there might be something I am missing in the interpretation of tp in either hourly or monthly datasets.

Many thanks

Wenqi

2

Same here. I found the following explanation:

The issue: The model calculates an average rainfall for that entire 30 × 30 km block. If a heavy, isolated thunderstorm only covers 10% of that area, ERA5 spreads that rainfall volume across the entire grid cell.

The result: Your localized area gets a high hourly volume, even though the actual point measurement might have been much lower.

3

Please ignore my previous answer. I was getting double the expected precipitation just because 00:00 belongs to the previous day, not to the next one. The next day’s readings actually start from 01:00.

4

On more nuance here. The docs say “Total precipitation over the one hour ending at forecast step.” but this seems to apply for regular ERA5 hourly accumulated precipitation. For ERA5-Land, which I use, summing hourly values would overcount badly because each hour is cumulative from midnight.