The drive would not spin for me. It held the original sensor records from three floods, and the public summary said a restored wetland had reduced every peak. The summary had become the accepted account in my county. Farmers repeated it at meetings. The water district used it in planning. A restored marsh had become the answer to an argument that still flooded people’s fields.
I had been sent to check the claim. The assignment came from a county archivist and three families whose land bordered the wetland. They did not agree about what the old study proved. They agreed that a claim with consequences deserved more than another reading of its conclusion.
The archive box looked ordinary. It contained a printed report, two external drives, a folded site map, a bundle of cables, and a paper log. The report described rain gauges placed above and below the wetland before its restoration. Their readings had been compared during three storms, one before the work and two after it.
The third event mattered most to the families because it had tested the claim against their own recent experience. It was the flood that had closed the road and backed water into an equipment shed. The report treated that flood as another success. The families remembered a night spent clearing a blocked culvert with an excavator, and they wanted to know why the successful wetland had needed that rescue.
One drive was the original. The second was labeled as a backup made during a later migration. I connected the original through a read-only adapter and supplied power. Nothing moved. The backup spun, but its file system did not mount on the county’s current machines.
The paper log offered a route through the problem. It named the original files, the sensor locations, the calibration dates, and the expected checksums. It also described a later conversion of the backup into a newer format. That entry gave me two possibilities. The backup might be damaged, or its differences might reflect an ordinary migration.
A checksum on the backup did not match the value printed in the log. The mismatch proved that the file had changed. It did not tell me whether the change came from corruption, normalization, or a later edit. Treating it as proof of damage would have been as careless as treating the report as proof of the wetland’s performance.
The county’s legacy computer could read the backup. Its folders contained reports, photographs, calibration tables, and two of the three storm records. The third storm was absent. A folder with its name contained a short note saying the raw files had not been included in that migration. The backup was a selection, and nobody had written that limitation on its label.
The original drive had become the only surviving route to the missing record. It was also the only route to the complete, unconverted record that could show how the third flood had been measured. I could finish my assignment with the summary and two storms. That would leave the disputed flood outside the check, which was the same as deciding the question before the evidence arrived.
I took the drive to a recovery technician. She examined it without opening the case and said the heads had seized. Replacing them might allow a disk image to be made. Opening the case would end the drive’s life as a sealed object, and the attempt could damage the platters. There was no version of the procedure that returned me an untouched drive with the files in my hands.
Her estimate was plain about the trade. She could attempt the recovery and document whatever survived. The drive would not be returned to service afterward. If the attempt failed, I would have an opened drive and no new evidence. If it worked, I would have files and a damaged original.
The county owned the hardware, but ownership did not answer every authority question. For the families, the claim’s effects were personal. The water district had a reason to prefer the old conclusion. I needed a decision that could be explained to all of them before I could defend it to myself.
We met in the archivist’s small office. One family member wanted the drive left alone because a failed recovery would erase the last physical route to the missing storm. Another wanted it opened because preservation had become a way to avoid the question. The water district’s representative asked whether the summary could simply stand.
I said the summary could stand as a summary. It could not answer the disagreement by itself. The original drive was the only remaining source for the third flood, and the backup’s missing folder explained why the public account had gone unchallenged for so long.
The technician joined us by speakerphone. She described the procedure, the possible outcomes, and the records she would create. There would be a photographic log, a parts inventory, a sector map, and a read-only image of every recoverable region. No estimate would be presented as a recovered measurement.
The families agreed to the attempt with those conditions. County officials accepted the recovery cost and authorized the work after a short discussion of who would review the files. The water district did not get a veto over a question it had helped frame. Their representative asked for the final files to be reviewed before any new claim reached the public.
The decision was still uncomfortable. I had exhausted the less damaging routes I could find, and the people affected by the result had accepted the risk. That did not guarantee success. It gave the loss a reason, a limit, and a record before anyone touched the case.
The technician opened the drive in a cleanroom. I watched through the window as she lifted the lid and photographed the internal parts. The original drive had ceased to be a drive in the way a future reader might expect to use it. Its value now depended on what could be recovered from the platters and on how carefully that recovery was described.
She replaced the heads and brought the assembly to the imaging station. The first attempt produced a brief burst of readable sectors and then stopped without warning. The second held long enough for the imaging software to build a map. Recoverable regions appeared in green. Unreadable regions stayed red.
The green regions covered most of the platters. They did not cover everything. When the work ended, 91 percent of the sectors had been recovered. The original drive sat open on the bench with its parts inventoried beside it. I had a new set of files and an original that could never be used again.
The files included the third storm. Their presence did not make them true. It made them available for the same scrutiny that had been applied to the summary. I needed the calibration history, the sensor positions, the time stamps, and the maintenance record before the recovered numbers could mean anything.
Both early storms matched the report closely. Each gauge had been calibrated before its event. The later storm, after the wetland was built, showed a lower downstream peak than the earlier storm with similar rainfall. This made the restored wetland appear to have changed the water’s path in a measurable way.
The third record began with a different problem. The downstream gauge had been moved during a bridge repair. Its new position was several hundred meters farther down the channel. The summary had treated the readings as if they came from the same location as the earlier events.
A moved gauge does not automatically invalidate a measurement. It changes the comparison. The report had normalized the readings as though the sensor’s position had not changed, and the normalization had concealed a pool forming behind the blocked culvert. That third flood could not serve as another clean test of the wetland.
The paper maintenance log supplied an independent account. On the night of the flood, a work crew had removed branches and broken fencing from the culvert below the wetland. The log gave the time they arrived and the time the water began to fall. It did not restore the missing comparison. It explained why the flooded shed did not answer the question the summary had assigned to it.
The recovered evidence gave the families a narrower answer. The wetland had reduced the downstream peak in the two comparable events. No conclusion about failure could be drawn from the third event because the gauge had moved and the culvert had been blocked. The report’s claim of success across all three floods went further than its own records allowed.
There was no clean victory for either side. The water district did not get proof that the wetland worked every time. The families did not get proof that it had failed during the flood they remembered. They got a reason to separate the wetland’s performance from the culvert’s condition, and a record of why that separation mattered.
The result also changed my view of the recovery decision. If the technician had found nothing, the attempt would still have been judged by the question, the alternatives, the authority, the amount of risk, and the record we had made. Success did not make those requirements irrelevant. A damaged drive with useful files was not proof that every risky recovery was wise.
I made a disk image of the recovered files and computed new checksums. The image was copied to two separate storage systems in different buildings. A working copy went to the county’s public records system. I bagged the failed drive and its parts with the photographic log and the technician’s sector map.
The handoff record named every loss I could identify, including the limits of the recovery and the gaps in the earlier account. My recovery had opened the original drive and left it unusable. Nine percent of the sectors had not been recovered. The backup omitted the third storm, and the public summary had misstated the gauge history. Even with those files recovered, the record was incomplete.
The record also named the checks that remained. Another investigator could recompute the flood peaks, inspect the calibration files, compare the maintenance log, and examine the recovered image against its checksums. A future hydrologist could use different methods on the surviving data. Nobody could repeat the original flood or restore the drive to its earlier condition.
My own choices now belonged to the evidence as well. I had selected the recovery route, chosen which files to compare, and decided which anomalies deserved attention. The raw data did not interpret themselves. My responsibility was to make those choices visible enough for another person to challenge them. A clean folder and a confident conclusion would have hidden too much of my part in the result.
Two copies in one archive box had nearly failed together. That was the lesson I had almost missed by trusting the label on the box. A backup is not a place or a label. It is a continuing relationship among copies, formats, documentation, access, and the ability to read what has been kept.
The printed report had lasted because paper asked for little beyond a dry shelf. The raw data had survived in a more demanding form. They needed power, software, compatible hardware, and someone willing to understand the difference between a file that existed and a file that could still be used.
The old hydrologist had left enough for me to begin. Her log identified the sensors and the calibration dates. The folder structure showed how the study had been assembled, migrated, and eventually reduced for a newer system. Her migration note explained the backup’s limits without explaining them on the backup itself. The records were incomplete, but they were connected.
I added my recovery record to hers. The connection now included the seized heads, the opened case, the sector map, and the missing 9 percent. It included the reasons the recovery had been authorized and the people who had accepted its risks. The next investigator would not have to invent my part of the story.
The public meeting took place in the same room where the old report had been presented. I showed the two comparable flood records and explained the moved gauge. Then I put the culvert maintenance log beside the third storm’s timeline. The room became quieter when the flood and the blockage appeared on the same page.
One of the farmers asked whether the wetland had worked. I said it had worked in the two events that could be compared. The third event could not answer that question. That old promise had been cleaner than the evidence could support, and the recovered files had not made it clean.
The water district’s representative asked whether the report should be withdrawn. I said the report needed a correction attached to it. Withdrawing the report would hide the history of the claim and make the correction harder to evaluate. Keeping it with the recovery record would show what had been asserted, what had been checked, and where the assertion had failed.
The correction carried a social cost too. District planners would have to revise a claim they had already used. The families would have to accept an explanation with more than one cause and no complete verdict. A bounded answer can be harder to live with than a confident one because it leaves work for everyone.
The next work order included the culvert. Its inspection and clearing schedule had been less visible than the wetland’s construction, but the flood had made it part of the same water system. The gauge’s new position also entered the monitoring record. Future comparisons would begin with the sensor’s location rather than assuming it had stayed still.
The recovered data did not settle every argument about the wetland. They could not predict the next storm or measure every effect of the restoration. They gave the county a better account of three events and a more honest list of what remained uncertain. That was enough to change the next decision.
My decision had a residue too. I could not return the drive to the archive box and pretend the box was unchanged. The opened case, the replaced heads, and the missing sectors belonged in the inventory. A future investigator needed to know that the original object had been spent to produce the image.
I signed the handoff record with my name and the date. The next investigator would know what I had recovered, what I had failed to recover, and why the attempt had been made. The old hydrologist’s name remained above mine. Her records had made my work possible, and mine would have to make someone else’s possible too.
The drive would not spin again. My work could still move forward. The county had a corrected summary, a culvert on its work schedule, and an archive that told the truth about its own condition. I put the failed drive and its parts back in the box with the new record.
My new book, Spent Evidence: What Science Destroys to Know, examines the same difficult bargain between discovery, preservation, and responsibility. It is available now from BolesBooks.com and on Amazon.
Leave a Reply Cancel reply