# Is this the most efficient way to download "mtdch" from "reanalysis-era5-complete"?

**URL:** <https://forum.ecmwf.int/t/is-this-the-most-efficient-way-to-download-mtdch-from-reanalysis-era5-complete/2964>\
**Category:** API\
**Tags:** cds, era5\
**Created:** [7 June 2024 11:15 UTC](https://forum.ecmwf.int/t/is-this-the-most-efficient-way-to-download-mtdch-from-reanalysis-era5-complete/2964 "2024-06-07T11:15:00Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![KH\_L](https://forum.ecmwf.int/letter_avatar_proxy/v4/letter/k/3d9bf3/32.png) [@KH\_L](https://forum.ecmwf.int/u/KH_L)\
**Post date:** [7 June 2024 11:15 UTC](https://forum.ecmwf.int/t/is-this-the-most-efficient-way-to-download-mtdch-from-reanalysis-era5-complete/2964/1 "2024-06-07T11:15:00Z")

</div>

Hello,

I am trying to download the parameter “ **mtdch** ” at model level 137 (param code 235014) from “ **reanalysis-era5-complete** ” using the CDS API.

I read somewhere that I should not request data on more than one tapes within a single MARS request. The ERA5 Catalogue indicates that mtdch records for each calendar day are “stored in one file on tape” (see screenshot below).

 ![image](https://forum.ecmwf.int/uploads/default/original/1X/34e30e43074a4f152f2de4cbbfde2b37a6b1407e.png)

Therefore, to retrieve all records from year 2000 to 2021, I am looping my requests at the calendar day level. The following is the python code snippet I am using:

```auto
for day in calendar_days:
            c.retrieve("reanalysis-era5-complete", {
                "class": "ea",
                "date": day,
                "expver": "1",
                "levelist": "137",
                "levtype": "ml",
                "param": "235014",
                "step": "1/2/3/4/5/6/7/8/9/10/11/12",
                "stream": "oper",
                "time": "06:00:00/18:00:00",
                "type": "fc",
                'format': 'netcdf',
                "grid": [0.281, 0.281], 
                "area": [40, -180, -40, 180]
            }, output_path)

```

The requests do run to completion without problem but they are quite slow, usually taking 30 minutes or more to retrieve a single calendar day.

So my question is, is this already the most efficient way of retrieving this parameter?

Any help is much appreciated.

Many thanks,  
KH