A significant number of recovery actions are not yet initiated.

According to the ROAR dump from 06 May 2016, 38.5% of actions are ongoing, 34% have not been started, 7% are of unknown status, and 5% have been completed. Given the state of funding for threatened and endangered species, these numbers may not be surprising. But the picture is slightly more nuanced than this.

But there’s a good bit of variation between action priority number

Recovery actions are put into one of three categories:

1. Those that must be taken to prevent extinction or irreversible decline;

2. Those that must be taken to prevent significant decline or habitat loss; or

3. All other actions needed for recovery.

Even though we saw in the previous figure that a significant number of actions have not been initiated, the data shows an encouraging pattern overall: more priority 1 actions have been started than not, but the reverse is true for priority 3 actions (and priority 2 are in the middle). That is, across >30,000 actions in this dataset, priority 1 actions tend to be implemented ahead of priority 2 or 3 actions.

One downside is that more priority 2 actions have been completed than priority 1…

The common words of recovery action descriptions tell us a bit about what is to be done.

Recovery actions mention ‘habitat’ more than any other single word, followed by ‘populations,’ which suggests that the species and their primary dependency are the primary focus. Other common terms represent recurring themes, such as ‘determine’ and ‘monitor[ing],’ which are associated with research and monitoring actions; and ‘develop,’ which is associated with planning and administration (e.g., developing guidelines).

Note: This wordcloud includes the top 100 words in recovery action descriptions, excluding common stopwords. To make the figure more readable, words have not been stemmed; stemming would collapse words like ‘monitor’ and ‘monitoring’ into a single term, resulting in a higher weight.

Nearly half of recovery actions have not been categorized.

While wordclouds can give us an overview of the relative use of individual words, we would like to understand actions more broadly. TO do so, actions should be classified into a relatively small number of categories that capture the core concept of each action. Unfortunately, 45% of actions in ROAR are ‘Work type not yet selected.’

Another complication of the action classification in ROAR is that single actions are often tagged with multiple - often disparate - work types. Actions* should be atomic, i.e., indivisible to simpler components. This mixing makes analysis of recovery actions at larger scales - for example, where ‘Research: Competition’ is most common across the country - complicated.

* Actions in recovery plans are often hierarchical; high-level actions would not be atomic and analysis would focus on the sub-actions that constitute the high-level action.

A lot of lead agencies are responsible for recovery actions. Most assignments are simply given as ‘Other’, but FWS is responsible for ~36% of actions.

Recovery takes the work of many parties. Every recovery action is ‘assigned’ to one or more* lead agencies or groups. Improved resolution of lead agencies - beyond a simple ‘Other’ - would be useful for evaluating patterns of responsibility, e.g., through time and space.

* There should be one lead entity per action to ensure everyone understands who is ultimately responsible. Assigning secondary responsible parties may be appropriate, but should not be confused with primary responsibility.