2 of 16

Budgeting assumptions

The resources needed for open science data management include hardware, software, and personnel with the expertise to execute all aspects of data management (full path data management) as described the proposal.
Your team includes experienced personnel who are allocated sufficient time to execute data management activities throughout the research data life cycle.
Grant funder/funding allows purchase of resources necessary for open science data management.
Use of open source software whenever possible.

3 of 16

Data Management activities that need to be budgeted for:

Active Research

Storage for active work (cloud, local)
Compute resources (cloud, HPC, local)
Data management

Implementing backup strategy
Lab management, ELNs

Training (time mainly)

Could involve travel, e.g., this workshop

Documentation

File, folder structure
Data
Code comments, code cleaning

DM Planning itself!

Sharing outputs openly and reproducibly

Data curation

Documentation, organization
Format conversions
Packaging for reproducibility (Docker, VMs, Singularity, Binder, Jupyter, etc)
Metadata
Working with a curator

Long-term archiving

Trusted repository
Some free some are not

Open access publishing costs (if applicable)

Mostly need to budget someone’s time to take ownership of implementing data operating procedures, do curation, documentation, and sharing

4 of 16

Under-budgeting of time...

80% is spent finding data and cleaning it

5 of 16

Ballpark amounts

The EU Open Science Cloud recommends 5% of allocated funds be dedicated to data stewardship
NOAA Climate Observation Division: projects spend 2-21% on data stewardship
Bottom line: there is no target percentage, just ballpark or guidance; so when budgeting for data management: “it depends”!

European Commission High Level Expert Group on the European Open Science Cloud “Realising the European Open Science Cloud, First report and recommendations of the Commission High Level Expert Group on the European Open Science Cloud”. 2016.

Saleem Arriago. “Facilitating and Ensuring Data STewardship: Data Challenges of NOAA’s Climate Observation Division”. 2016

6 of 16

Costs at different points in the research lifecycle

Credit: Digital Curation Centre

Time/Personnel,�Data licensing costs

Time/Personnel (experience vs cost)�Storage

Compute

Documentation

Long-term storage

Data curation &

Archiving costs

7 of 16

Active-phase research

8 of 16

Personnel budgets should...

...consider:

Costs at all phases of data life cycle
Qualifications/experience of personnel [data science, stats, standards, version control, identifiers, FAIR data, data management, tools for reproducible research (e.g., OSF, containers)]
Cost vs. experience matrix (more experience = time saving)

...include time and experience for:

Necessary training or certifications
Researching and implementing standards that will be used for all phases
Collecting and formatting metadata
Quality assurance and control (before, during and after data collection)
Data cleaning, organization, and integration (AKA data munging)

9 of 16

Hardware and software budgets should include:

Hardware (e.g., laptop[s] with adequate storage and compute power for your data generation and potentially analysis needs)
Connection/access costs for ethernet or wifi of sufficient bandwidth for uploading data
Free Open Source Software (FOSS) costs nothing (but this often requires personnel with sufficient knowledge/expertise to use it and/customize it to your needs)
Any enterprise-level software or services you consider essential to the project operations (e.g., G-suite, Dropbox, Slack, Zoom, AWS, Azure, etc.)
Long-term storage and archival fees for trustworthy repositories
Data publication fees

10 of 16

Sample Storage Costs

Assume 100 TB storage, 10 TB egress per year.

Think about what the storage time frame and purpose are

Storage Type	Storage Cost�	Egress (10 TB/yr)	Total (1 yr, 100TB)	Note
Institutional research computing	$3900 - $4500	$0	$3900 - $4500	Meant for active Storage
Amazon S3	~$1500	~$900	$2500	Glacier Deep Archive (infrequent access)
LTO-8 Tape	~$1500 + $7500	0	$9000	10 x 12TB tapes + drive. Long-term storage. Does not include labor or storage

11 of 16

Wrapping up research

12 of 16

What needs to happen?

Publishing - the steps that need to happen even before publishing

Understanding and meeting requirements for repository storage

Metadata
Permanent Identifiers (PIDs)
Curating your data (quality control again!)
Submitting data into repos (archiving)
Documenting analytical methods and parameters for reproducibility

Submitting to journal(s)

Fees
Publication fees
Data Accessibility Statement requirements by publishers (paywall or open access)

13 of 16

Data Archiving Costs

Easier to budget for than other costs

Public vs non-public�Non-public: See previous slide on storage costs

Choice of archive matters

Disciplinary may be better but less sustainable due to size - check first

Some cost money, some are free
More at osf.io/jdzgc/

14 of 16

Data Curation Costs

Highly variable.

Type of data, complexity of project, experience of researcher with open science, metadata etc.

Example 1: inexperienced student in the discipline that requires training on data sharing and publishing

$5000 + travel/training/stipend for summer student for “just” preparing a data package for publication

TERRA-REF project job posting for student at UA

A lot of the work is documentation

15 of 16