IELTS Reading 練習問題 — Set 8

A Soils contain more carbon than the atmosphere and vegetation combined, which makes them central to climate projections. Plants remove carbon dioxide through photosynthesis, and some carbon enters soil through roots, fallen leaves and microbial products. It may remain for days or centuries before returning to the air. This wide range has encouraged proposals to increase soil carbon as a climate solution, but it also makes measurement and prediction unusually difficult.

B The old picture divided organic matter into pools that decomposed at fixed fast, medium or slow rates. New research emphasises interaction. A molecule may persist not because it is chemically resistant but because it is attached to a mineral surface, trapped inside an aggregate or physically separated from microbes. Disturbance can break that protection. Stability is therefore a property of the soil environment as much as of the carbon compound itself.

C Microbes complicate expectations about warming. Higher temperature can accelerate their metabolism, potentially releasing more carbon. Yet microbial communities may adjust, available food may be exhausted, and dry conditions can restrict activity. Short laboratory incubations often show a strong initial response that weakens over time. Translating such results to fields requires attention to roots, changing moisture and fresh inputs that a sealed jar cannot reproduce.

D Measurement begins with sampling, and soil is heterogeneous over remarkably short distances. Carbon concentration differs between a plant row and the gap beside it, between surface material and deeper layers, and even around individual roots. Detecting a small annual change against this spatial variation requires many cores and consistent methods. If a project samples only the upper soil, it may miss carbon that moved downward—or a loss at depth that cancels a surface gain.

E Bulk density is an easily overlooked variable. Researchers often report carbon as a percentage of dry soil, but management can loosen or compact the ground. A higher percentage after treatment does not necessarily mean more carbon per hectare if the mass of soil in the sampled volume has changed. Equivalent-soil-mass calculations improve comparison, though stones, swelling clays and uncertain depth boundaries continue to create error.

F Practices such as cover cropping, adding compost and reducing tillage can raise carbon in some contexts. Their effect depends on climate, soil type, previous management and duration. Compost also transfers carbon from somewhere else; counting the receiving field’s gain without the source and transport can overstate the net benefit. Reduced tillage may concentrate carbon near the surface while producing little change when the whole profile is measured.

G Saturation places another limit on extrapolation. Mineral surfaces have finite capacity to protect organic matter, so gains may slow as available binding sites fill. This does not make improved management pointless: healthier soil can retain water, reduce erosion and support yields even when carbon accumulation becomes modest. It does mean that a rate measured during early recovery should not be projected indefinitely.

H Verification over time introduces the risk of reversal. A drought, fire or change in management can release carbon previously credited as a removal. Contracts sometimes create buffer pools in which a share of claimed gains is withheld to cover later losses. The approach spreads risk but cannot solve every systemic event affecting an entire region. Permanence is therefore not a property certified once; it is a monitoring obligation extending beyond the initial project.

I Remote sensing may reduce sampling costs by linking field measurements to vegetation and moisture patterns visible from satellites. It cannot measure soil carbon directly at the accuracy required for small annual changes. Models must be trained against cores, and their reliability falls when applied to soils or management systems absent from the training data. Broad coverage is useful, but it does not remove the need to touch the ground.

J Soil carbon policy must therefore separate three claims: that a practice improves soil, that carbon stock increased on a site, and that the change caused a durable net removal from the atmosphere. Each requires different evidence. Long-term monitoring, depth-consistent sampling and transparent treatment of displaced emissions can strengthen confidence. Soil can contribute to mitigation, but presenting it as a precisely countable substitute for rapid emission cuts asks the science to promise more than it can deliver.

A A near miss is an event that could have caused harm but did not, either by chance or through timely intervention. Aviation, medicine and industrial operations collect such incidents because disasters are rare and therefore provide limited data. A tool dropped beside a worker or a dosage error caught before administration contains information about weak controls. Yet organisations find near misses difficult to learn from precisely because no visible injury forces attention.

B Reporting depends on psychological safety. Workers must believe that describing an error will not automatically produce punishment or ridicule. “No-blame” policies are sometimes proposed, but the phrase can be misleading: reckless behaviour and deliberate violation still require accountability. A fair system distinguishes human error, risky shortcuts encouraged by working conditions and conscious disregard of substantial danger. Without that distinction, either fear suppresses reports or immunity destroys trust.

C The number of reports is an ambiguous performance measure. An increase may indicate deteriorating safety, or it may show that staff have become more willing to speak. A decrease could mean improvement, silence or changes in classification. Mature programmes examine the content and reporting culture instead of rewarding managers for the smallest total. Otherwise incidents are relabelled or quietly resolved to protect a target.

D Selection creates another bias. Memorable events involving dramatic last-minute recovery receive attention, while ordinary weak signals are ignored. Investigators may focus on the person closest to the event because their action is visible, overlooking scheduling, software design or procurement decisions made months earlier. A useful review reconstructs how the situation appeared before the outcome was known. This reduces hindsight bias—the tendency to treat warning signs as obvious after the event.

E Some near misses reveal resilience rather than only failure. A nurse questions an unusual prescription, or a maintenance team improvises safely when a component is unavailable. Studying these recoveries shows which expertise and relationships keep a system functioning. However, celebrating adaptation can normalise chronic shortages. If staff succeed only through repeated extra effort, the organisation may mistake exhausted workers for a reliable safety barrier.

F Information must be aggregated carefully. One report may seem trivial, but similar reports across sites can reveal a recurring interface problem. Standard categories support comparison, while free-text accounts preserve context and unexpected factors. Automated text analysis can identify themes in thousands of reports, yet rare terms and local jargon are easily misclassified. Human review remains necessary, especially when algorithms were trained on the very categories that investigators hope to question.

G Learning requires feedback. Employees stop reporting if submissions vanish into a database and no change is visible. Feedback need not promise that every suggestion will be adopted; it should explain what was reviewed, what action followed and why some risks were accepted. Sharing de-identified cases across organisations can spread lessons, but excessive anonymisation may remove the operational details needed to recognise the same hazard elsewhere.

H Differences in exposure also matter when near-miss rates are compared. One factory may report more incidents because it performs more hazardous operations, employs more people or observes work more closely. Rates adjusted by hours or tasks improve comparison but still omit complexity. Benchmarking should prompt questions, not automatically label the highest reporter unsafe or the lowest reporter exemplary. Local investigation must interpret the denominator as carefully as the event.

I Changes must themselves be monitored. Adding an alarm may reduce one risk but create alarm fatigue; a second verification may encourage the first checker to become less attentive. Investigators should state the mechanism by which an intervention is expected to work and look for displacement as well as improvement. Closing a report when an action is assigned confuses administrative completion with safer performance.

J Near-miss systems succeed when they alter ordinary work rather than merely produce reports. That may mean redesigning an interface, changing staffing, adding a verification step or removing a target that encourages shortcuts. The absence of harm should be treated as an opportunity for investigation, not proof that controls were adequate. Learning before disaster is inevitably based on incomplete evidence, but waiting for certainty means allowing consequences to supply the missing data.

A New infrastructure attracts ceremonies, architectural images and political credit. Maintenance rarely does. Its success is an absence: the bridge does not close, the drainage system does not overflow, the public database remains accessible. Because avoided failure is difficult to photograph, budgets often favour new construction even while existing assets deteriorate. This is not simply poor accounting. It reflects a political culture that treats repair as returning to normal and novelty as progress.

B The distinction is misleading because maintenance changes objects. Replacing a component with a modern equivalent may alter performance; adapting a station for accessibility changes who can use it; patching software can remove one risk while introducing another. Maintainers interpret original designs under conditions their creators did not foresee. Their work is therefore a form of continuous redesign, although procurement documents may describe it as routine service.

C Deferred maintenance creates nonlinear costs. A small roof leak damages plaster first, then wiring and structure. Inspection postponed for one budget cycle may require a complete replacement later. Financial systems nevertheless separate capital expenditure, which funds new assets, from operating expenditure, which supports care. Leaders can announce investment through the former while restricting the latter, producing facilities that are expanded and simultaneously less reliable.

D Knowledge of condition is uneven. Large authorities use asset registers, sensor data and risk models to prioritise work. These tools are valuable, but they record what has been defined as an asset and what sensors can detect. A caretaker may notice a changing smell or pattern of complaints long before an alarm threshold is reached. When contracts split cleaning, security and engineering among companies, such observations may no longer travel to someone who can act.

E Maintenance also distributes inconvenience. Closing a road for planned repair delays today’s users in order to prevent a larger future failure. Officials tempted to avoid immediate complaints may schedule work at night, transferring noise to residents, or postpone it until an emergency leaves no choice. Decisions about whose time, sleep and safety are protected are political even when presented as technical optimisation.

F Standards can protect minimum quality, yet rigid schedules sometimes waste resources. Servicing every component at the same interval ignores differences in use and condition. Predictive maintenance uses monitoring to intervene when deterioration appears, but it depends on reliable data and models. A sensor failure may itself go unnoticed, and rare failure modes may be absent from training records. Combining scheduled inspection, condition monitoring and local judgement is safer than declaring one method universally superior.

G There is increasing interest in maintainability at the design stage. Components can be made accessible, manuals supplied in open formats and replacement parts standardised. Manufacturers may prefer sealed products and proprietary tools that secure future revenue. Purchasers focused on the lowest initial bid overlook these dependencies. Evaluating lifetime cost requires assumptions about future labour, energy, parts and system change—uncertain estimates, but still more informative than treating purchase price as total cost.

H Maintenance supply chains can fail even when money is available. A specialised chip, seal or software certificate may no longer be produced, turning a minor replacement into a redesign. Organisations that map critical dependencies and keep interoperable alternatives are better prepared, but stockpiling every component is costly and parts themselves age. Resilience comes from knowing where substitution is possible and retaining the authority to make it safely.

I Metrics influence what receives care. Response time is easy to count, while the quality of a diagnosis or the patience required to prevent recurrence is harder to capture. Contractors rewarded for closing tickets may make temporary fixes that reopen later under a new number. Useful performance systems follow repeated faults and user experience, not just the speed with which each isolated record disappears.

J A maintenance-centred politics would ask different questions of public investment. It would value the workers who preserve continuity, publish the condition of existing systems and attach long-term care funding to new projects. It would also accept that not everything should be maintained forever: obsolete or harmful systems may need retirement. The central choice is what society commits to sustaining, for whom and at whose expense. Repair becomes visible as collective priority-setting rather than evidence that progress has stopped.