# Clarification of data download throughout the forecast stage

**URL:** <https://community.drivendata.org/t/clarification-of-data-download-throughout-the-forecast-stage/9815>\
**Category:** Water Supply Forecast Rodeo\
**Created:** [December 28, 2023, 5:44pm UTC](https://community.drivendata.org/t/clarification-of-data-download-throughout-the-forecast-stage/9815 "2023-12-28T17:44:00Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![robby](https://avatars.discourse-cdn.com/v4/letter/r/e9c0ed/32.png) [@robby](https://community.drivendata.org/u/robby)\
**Post date:** [December 28, 2023, 5:44pm UTC](https://community.drivendata.org/t/clarification-of-data-download-throughout-the-forecast-stage/9815/1 "2023-12-28T17:44:00Z")

</div>

It seems to me that the data downloaded and processed for the forecast is wiped after the forecast is produced, meaning that each weeks forecast needs to redownload and process anything relevant that was used the previous week, is that correct?

---

<div class="post-metadata">

**Author:** ![mmiron](https://avatars.discourse-cdn.com/v4/letter/m/5daacb/32.png) [@mmiron](https://community.drivendata.org/u/mmiron)\
**Post date:** [December 29, 2023, 10:38am UTC](https://community.drivendata.org/t/clarification-of-data-download-throughout-the-forecast-stage/9815/2 "2023-12-29T10:38:06Z")

</div>

@robby Yes and no. If you don’t use any data that isn’t included on the mounted data drive (i.e. the data downloaded and made available, prior to your solution even being run), then not really. If you do use other data sources though, then yes, I believe it is discarded. Anything you write to the preprocessed directory should be available throughout the lifetime of the process though (i.e. until your program exits).

---

<div class="post-metadata">

**Author:** ![jayqi](https://yyz2.discourse-cdn.com/flex028/user_avatar/community.drivendata.org/jayqi/32/924_2.png) [@jayqi](https://community.drivendata.org/u/jayqi)\
**Post date:** [December 31, 2023, 5:47pm UTC](https://community.drivendata.org/t/clarification-of-data-download-throughout-the-forecast-stage/9815/3 "2023-12-31T17:47:31Z")

</div>

Hi @robby,

@mmiron’s explanation is correct. We do not have any caching between code execution jobs.