Make the decision path visible
Closing a monitoring site can save resources while destroying an irreplaceable long record or opening a new blind spot. This guide shows how to retire a sentinel site without losing decision coverage, comparability or the evidence future teams will need.
Closing a monitoring site can save resources while destroying an irreplaceable long record or opening a new blind spot. This guide shows how to retire a sentinel site without losing decision coverage, comparability or the evidence future teams will need. The aim is not to force certainty where the evidence is conditional. It is to show what can be used now, what needs further testing and what must remain visible to the next person.
The chapters move from first principles to a worked case, changed-context transfer, explicit limits, practical checks and an auditable decision. Start with the route closest to your task or read in sequence.
Choose the route closest to your decision
Chapter contents
Route 1: chapters 1–3
Route 2: chapters 4–6
Route 3: chapters 7–9
1. Distinguish planned retirement from unmanaged failure
Decommissioning is a governed decision with evidence preservation, impact review and handover. Site loss is an event that may occur before any of those safeguards are ready. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Is the programme choosing to retire the site, or merely accepting loss caused by access, funding or equipment failure? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A lease ends at a wetland station. Treating the closure as inevitable would skip review of its unique salinity and bird-migration record. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the distinction to a destroyed sensor where emergency recovery must precede any permanent network decision. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Calling a site inactive does not show that its function is replaceable. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Record closure driver, timeline, responsible owner, legal duties, salvage options and the date at which irreversible steps begin. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Contents · Previous · Ecology root · Next
2. Inventory every decision that uses the site
Map alerts, trend estimates, models, permits, community reporting, calibration pairs and research commitments that depend on the observations. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Who changes a decision when this site stops reporting? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A stream station supports flood warning, nutrient-load estimates and a restoration trigger; its value cannot be represented by one programme budget line. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the inventory to biodiversity acoustic stations whose data feed occupancy models and species alerts. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Low local use may coexist with high downstream or historical value. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Keep users, products, thresholds, update frequency, contractual duties and the consequences of missing data. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Contents · Previous · Ecology root · Next
3. Measure historical continuity and irreplaceability
Long records gain value because they span rare events, regime shifts and changing baselines. Assess length, consistency, metadata quality and environmental uniqueness. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. What inference becomes weaker or impossible if the series ends now? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A thirty-year estuary record includes drought, land reclamation and heat extremes; a nearby three-year station cannot replace its temporal leverage. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the test to a forest plot that anchors succession and carbon estimates. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Record length alone is not enough if methods changed without overlap or the site no longer represents its original process. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Record periods of record, method eras, gaps, environmental changes, extreme events and analyses that require continuity. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
| Checkpoint | What to test |
|---|---|
| Decision question | What inference becomes weaker or impossible if the series ends now? |
| Worked comparison | A thirty-year estuary record includes drought, land reclamation and heat extremes; a nearby three-year station cannot replace its temporal leverage. |
| Transfer test | Transfer the test to a forest plot that anchors succession and carbon estimates. |
| Known boundary | Record length alone is not enough if methods changed without overlap or the site no longer represents its original process. |
Contents · Previous · Ecology root · Next
4. Test network coverage after removal
Recalculate spatial, pathway, habitat, seasonal and community coverage with the site absent. Include detectability and response time, not only distance to the nearest station. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Which critical stratum or stressor route loses its last observer? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. Removing an upstream tributary site leaves three downstream stations but no way to separate agricultural from urban inputs. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the analysis to pollinator monitoring where timing and habitat type may matter more than kilometres. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. A dense map can still contain a functional blind spot. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Keep before-and-after coverage metrics, residual gaps, uncertainty, vulnerable groups and the model assumptions behind substitution. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Contents · Previous · Ecology root · Next
5. Design overlap before the old record ends
When replacement or relocation is needed, run both sites long enough to estimate level shifts, seasonal differences and event response. Define success before overlap begins. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Does the overlap cover the conditions in which the sites may differ? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. Old and proposed stream stations operate together through wet and dry seasons; their relationship breaks during high flow, so a simple offset is rejected. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the overlap to new sensors by spanning calibration drift and environmental range. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. A short calm-season overlap cannot validate comparability during extremes. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Record overlap duration, event coverage, methods, calibration, acceptance criteria and the plan if equivalence fails. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
| Checkpoint | What to test |
|---|---|
| Decision question | Does the overlap cover the conditions in which the sites may differ? |
| Worked comparison | Old and proposed stream stations operate together through wet and dry seasons; their relationship breaks during high flow, so a simple offset is rejected. |
| Transfer test | Transfer the overlap to new sensors by spanning calibration drift and environmental range. |
| Known boundary | A short calm-season overlap cannot validate comparability during extremes. |
Contents · Previous · Ecology root · Next
6. Work a retirement decision case
Combine decision dependencies, historical value, coverage, overlap evidence, lifecycle cost and residual risk. Compare full retirement, partial record, seasonal operation and relocation. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Would a less drastic operating mode preserve the unique function at acceptable cost? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A remote site is expensive but uniquely detects monsoon pulses. Continuous operation ends, while event-triggered sampling and a lower-cost sensor preserve the pathway. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Change the context to a site whose habitat has transformed and test whether the old location still serves the original question. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Keeping a site forever is not automatically responsible; obsolete monitoring can divert resources from larger gaps. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Save alternatives, costs, information loss, stakeholder input, decision, conditions and the reason rejected options were not chosen. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Contents · Previous · Ecology root · Next
7. Preserve data, metadata and physical context
Archive raw and processed data, calibration, quality flags, methods, site photographs, coordinates, access history and environmental changes in durable formats. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Could a future analyst reconstruct what was measured, how and under which site conditions? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. The final package includes rating curves, sensor changes, maintenance logs and photographs showing bank movement that affected readings. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the archive to biological specimens by linking vouchers and taxonomic revisions. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. A downloadable table without methods and quality history is not a preserved monitoring record. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Keep repository identifiers, licenses, restrictions, checksums, retention, schema and the custodian responsible after closure. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Contents · Previous · Ecology root · Next
8. Protect comparability in downstream analyses
Mark the end of record, replacement relationship and uncertainty in every derived product. Refit models when the network geometry changes. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Can trend users distinguish ecological change from a change in which sites are observed? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A regional index shifts after a high-altitude site closes; analysts recompute weights and show the effect of the network change separately. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the safeguard to a rotating panel where site composition changes by design. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Splicing series without uncertainty can manufacture a trend or hide a real one. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Record model versions, weights, imputation, break indicators, sensitivity results and communication to data users. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
| Checkpoint | What to test |
|---|---|
| Decision question | Can trend users distinguish ecological change from a change in which sites are observed? |
| Worked comparison | A regional index shifts after a high-altitude site closes; analysts recompute weights and show the effect of the network change separately. |
| Transfer test | Transfer the safeguard to a rotating panel where site composition changes by design. |
| Known boundary | Splicing series without uncertainty can manufacture a trend or hide a real one. |
Contents · Previous · Ecology root · Next
9. Maintain early warning and minimum coverage
Before retirement, confirm that alert lead time, threshold confirmation and emergency response remain adequate. Temporary backups may be needed during transition. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Will the remaining network see the same event soon enough to act? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A coastal sensor gives twelve hours more warning of a low-oxygen pulse than the next station; retirement pauses until a replacement pathway is proven. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the check to invasive-species surveillance where absence of detections may reflect reduced effort. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Average coverage targets can hide loss of a rare but high-consequence signal. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Keep alert simulations, missed-event scenarios, response ownership, backup arrangements and restoration triggers. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Contents · Previous · Ecology root · Next
10. Close the site safely and respectfully
Remove or transfer equipment, restore the physical site, settle permits, preserve community agreements and document hazards. Avoid leaving abandoned infrastructure or unclear ownership. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Who is responsible for land condition, equipment, data stewardship and future access after the last sample? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A marsh platform is removed outside breeding season, disturbed ground is restored and local partners receive the final archive and contact route. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the closure plan to Indigenous or community-managed sites through agreed authority and data governance. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Technical retirement does not end ethical, contractual or environmental obligations. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Record permits, consultation, asset disposition, restoration evidence, safety sign-off and continuing commitments. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
| Checkpoint | What to test |
|---|---|
| Decision question | Who is responsible for land condition, equipment, data stewardship and future access after the last sample? |
| Worked comparison | A marsh platform is removed outside breeding season, disturbed ground is restored and local partners receive the final archive and contact route. |
| Transfer test | Transfer the closure plan to Indigenous or community-managed sites through agreed authority and data governance. |
| Known boundary | Technical retirement does not end ethical, contractual or environmental obligations. |
Contents · Previous · Ecology root · Next
11. Answer recurring network questions
Explain when a long record is long enough, whether a nearby site can replace it, how much overlap is needed and when cost justifies closure. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Can each answer show the decision-specific tradeoff instead of offering a universal duration or distance? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. A manager asks for a fixed two-year overlap. The team instead chooses seasons and events needed to test the expected differences. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the answer to a sparse network where no substitute exists and partial operation may be preferable. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. Do not turn a local stop rule into a universal ecological standard. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Publish decision questions, evidence expectations, escalation, review dates and the route for affected data users. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Can one threshold settle every case?
No. A threshold belongs to a purpose, population, evidence base and consequence structure. Recheck it when any of those change.
Does more data always make the answer safer?
No. Repeated or poorly controlled data can reproduce the same blind spot and increase confidence without increasing validity.
When should an irreversible action pause?
Pause when a plausible unresolved condition could change the action and targeted evidence or a reversible alternative is proportionate to the risk.
What makes the process auditable?
Another person must be able to find the inputs, transformations, assumptions, alternatives, authority, decision and residual uncertainty without relying on memory.
Contents · Previous · Ecology root · Next
12. Audit retirement and reopen when conditions change
Review data availability, replacement performance, blind spots, user impacts and site restoration after closure. Reopen the decision if assumptions fail. In sentinel networks that observe ecological conditions, stressor pathways and management response over long periods, the practical claim is whether a site can be retired without unacceptable loss of historical continuity or decision coverage. The chapter therefore begins with the decision boundary: what the evidence must support, what it may suggest and what it cannot safely establish.
First-principles check. Did the programme save money without losing the function it promised to preserve? Name the object being compared, the mechanism that could create error and the consequence of a wrong answer. Make the time window, population, method and authority explicit before accepting a score, label or workflow result.
Worked practice. One year later, the substitute misses brief contaminant pulses; event sampling is reinstated at the retired reach. The important sequence is visible: observations come first, alternative explanations are tested, the bounded inference is stated and the action follows. This prevents the conclusion from becoming stronger merely because it is convenient.
Changed-context transfer. Transfer the audit to climate-driven pathway change that creates a new need near the old site. Carry over the reasoning steps, not the old threshold. Replace the local data, controls, costs and consequences, then ask whether the same conclusion still holds. A copied answer is not transfer if a hidden assumption changed.
Limit and misconception. A tidy closure report can conceal later information loss unless outcomes are checked. The recurring danger is that short-term savings permanently remove a unique signal and leave the remaining network looking more complete than it is. A careful analysis marks the boundary and keeps competing explanations available until evidence genuinely excludes them.
Evidence record. Keep post-retirement metrics, complaints, missed events, corrective action, lessons and verified specialist routes. Add source, date, operator or analyst, method and software version, exclusions and uncertainty. Preserve earlier states and link raw observations to derived decisions so that a later reviewer can challenge one step without losing the chain.
Decision-quality drill. Write one sentence naming the action supported now and one naming the strongest unsupported claim. Then change one plausible condition—threshold, method, missing record or stakeholder cost. If the action reverses, label the decision sensitive and set a review trigger.
Contents · Previous · Ecology root · Next
Continue through verified routes
Return to the core Gold Standard framework or continue into the specialist routes that directly extend this decision. The external sources below define or test the concepts used here.
Authoritative sources
- USGS: What is a streamgage?
- USGS report: Evaluation of the streamgage network in South Carolina
- USGS: National Streamgaging Network
- USGS Report to Congress: Network Evaluation
Guidance, methods and datasets can change. Recheck the linked source, local evidence and decision context before carrying the conclusion into a new case.