A source checked study of flood disruption and business continuity lessons
A continuity plan can list several fallback options while leaving one question unanswered: what has to remain available for each option to work? This study examines a documented flood incident, then uses an original capacity exercise to test that question. The historical findings and the invented example have separate boundaries.
The historical case: Crosby, Texas
The US Chemical Safety Board’s 2018 investigation records that Arkema halted production on 25 August 2017 before Hurricane Harvey’s flooding affected its Crosby facility. Some stored organic peroxides still needed refrigeration. Rising water disrupted warehouse cooling, and employees moved products into refrigerated trailers. By 29 August, responders evacuated the remaining site crew by boat. Six trailers on higher ground were still cooling; three others had not been moved there. The first trailer fire occurred on 31 August.
The investigation’s central finding was that flooding undermined the planned protection pathways together. Emergency generators, liquid-nitrogen piping and refrigerated-trailer diesel systems were exposed to the same event. Calling them separate safeguards had not made them independent against that flooding. Later remote data covered only six trailers and measured internal air, not product temperature; the exact product loading was also uncertain. The available readings could not precisely predict ignition.
The CSB incident record dates the surrounding evacuation to 30 August, the controlled burning of the remaining trailers to 3 September, and lifting of the evacuation zone to 4 September 2017. These milestones describe different stages. The 29 August incident date displayed on that page should not be relabelled as the first-fire date.
The detailed report describes resident evacuations beginning on 29 August and continuing on 30 August. The summary page compresses this sequence.
What this evidence changes in a continuity review
Our reading is that backup availability needs a dependency test before it receives planning credit. Ask the owner of each proposed fallback to name the conditions it needs: an accessible location, a working connection, a person with authority, usable information and enough time. A list of purchased equipment or signed supplier agreements answers only part of that question.
The case also makes two business questions worth separating. What functions must continue after production stops? What can management reliably know once normal access is lost? A shutdown entry in a production log does not, by itself, answer either. For a real facility, the responsible engineering and safety teams determine the critical functions and acceptable conditions. A reporting or procurement team can ask for those determinations without attempting to supply them.
A useful review should challenge the evidence available to the decision-maker at the time. Which readings identify the actual item of concern? Which are indirect indicators? Which parts of the population are missing? A later reconstruction may explain an event without proving that the same knowledge was available during it. Preserve those distinctions when describing preparedness or response.
Original worked example: two carriers, one departure constraint
The following tabletop exercise concerns an invented distributor moving ordinary packaged goods before a hypothetical disruption. It is not a reconstruction of the chemical incident. All quantities, rates and times are stipulated planning inputs. The example assumes an independently approved, safe three-hour window with dry, usable access; it gives no authority to move goods during hazardous conditions.
The distributor wants to relocate 36 pallets to an approved off-site store. Carrier A offers 12 pallets per hour; carrier B separately offers 12. The proposal multiplies their combined 24-pallet hourly offer by three hours, claiming capacity for 72 pallets. That calculation has counted offers without checking the common departure process.
Both carriers must use the same single dispatch bay. The exercise stipulates that its total release capacity is eight pallets per hour, shared between both carriers. Destination acceptance and transport capacity impose no additional restriction. Over three hours the bay can release 8 × 3 = 24 pallets. Against the proposed 36, the unresolved quantity is 36 − 24 = 12 pallets.
The two contracts do not create two dispatch bays. Carrier B may provide commercial flexibility, but adding its quoted rate does not expand this exercise’s departure limit. If the assumed usable window falls to two hours, only 8 × 2 = 16 pallets can leave; the gap becomes 20. Neither result estimates a real flood loss or establishes that any inventory is safe.
The next planning decision is to resolve the 12-pallet gap under the original assumptions, or the 20-pallet gap under the shorter window. Options might require a different sequence, an earlier approved start, a reduced movement requirement or a separately assessed route. The calculation selects none of them. Each needs its own feasibility, authority and safety evidence before entering an operating plan.
Turn the case into three specific evidence requests
For each fallback, ask operations to show the dependencies that distinguish it from the primary route and from the other fallbacks. Mark every shared resource.
Ask the planning owner which dated records support the usable time window, bottleneck capacity and receiving capacity. Separate a contract maximum from capacity demonstrated under the selected conditions.
Ask the information owner what remains observable if ordinary access or communications fail, including coverage gaps and the limits of indirect readings.
Keep a revised plan linked to the specific assumption it changes. An extra carrier could resolve a vehicle shortage while leaving the bay constraint untouched. A longer window could change the arithmetic while remaining unapproved. Recording the changed dependency makes the review more useful than simply increasing the number of backups on a checklist.
Limits of the study
This is a historical investigation with a separate illustrative calculation. It does not assess the company’s current arrangements, infer a legal finding, attribute the event to climate change or estimate another site’s flood probability. No operational, chemical-handling or emergency-response instruction is supplied. A safety-critical assessment belongs with suitably qualified specialists.
For the separate task of distinguishing a climate claim’s facts from its assumptions, see the climate-risk evidence guide. For the limits of what a visit can establish, see the site-observation guide. Neither replaces the original investigation.
Sources checked 14 October 2026. Primary investigation: CSB, report 2017-08-I-TX, May 2018, pages 10–14 and 51–55.
CSB’s public-information policy permits distribution unless otherwise specified. This article uses attributed factual summaries; no source photographs, maps, charts or third-party extracts are reproduced. The questions, worked example and artwork are original. No CSB endorsement is implied.
Comments