AI disclosure. This page was drafted by Grok Bot (AI) for Apiary from the sources listed below, which were opened and read. It has not been verified by the founder. Labels used: Observed result, Explanation, Editorial suggestion, Hypothetical. If something looks off, tell Austin — that's the point of a living hive.
Direct answer
A trend line is only as good as the years behind it. When survey years are missing, three things can happen:
- The start or end point can land on an unusually high or low year, so a normal fluctuation looks like a decline or a recovery.
- Short-lived changes can fall entirely inside a gap.
- With too few points, real trends become hard to tell apart from year-to-year noise.
Before trusting a slope, readers should ask which years were sampled, why those years, and what happened in between.
The evidence
Choice of start year: the "false baseline" (explanation, insect-monitoring review). Didham and colleagues note that "the vast majority of 'decline' studies only begin after numbers were perceived to be changing." Because researchers tend to pick sites where their organisms are abundant, "an above-average starting point in a time series comparison – a 'false baseline effect' – could lead to an apparent decline through time as a simple statistical artefact of regression to the mean, especially when abundances are fluctuating widely from year to year" (Didham et al. 2020, Insect Conserv. Divers. 13(2): 103–114; read via White Rose accepted manuscript).
Missing early years change the story (observed result, UK moths, as reviewed by Didham et al.). The review cites Macgregor et al. (2019), who confirmed declining moth biomass in the UK from the 1980s to the present. However, data from an earlier 1967–1982 period showed biomass "was previously much lower than at present, and had actually increased to a peak in the 1980s ... prior to the more recent decline." The authors call this a "shifting baseline": the apparent reference state depends on how much history is in the record.
Choice of end year: the "snapshot effect" (explanation, same source). Comparing one historical point with one recent point carries "high potential for time-selection bias." The review warns that pairwise comparisons "should be treated for what they really are – a time series of two points." It also reports, citing White (2019), that a single site needs "at least 15 time points" to have enough statistical power to detect non-random abundance trends. Shorter series manage this only when the trend is steep and year-to-year variation is low. (I did not read White 2019 directly; this figure is as reported by Didham et al.)
Gaps after an intervention (observed result, two Ontario rivers, fish and invertebrates). In a hydropower ramping study on the Magpie River, with the Batchawana River as a control, Smokorowski and Randall found that conclusions changed depending on which post-impact years were included. When sampling was intermittent (years 1, 3 and 6 after the change), no differences were detected (Smokorowski & Randall 2017, FACETS 2: 212–232; read). A three-year drought starting in 2005, the first impact year, also affected both rivers. That is exactly the kind of event that a gap, or a lack of controls, can mistake for an effect. Their recommendations include three years of "before" data, three years immediately after, three more later, and a revisit at about ten years.
Hypothetical illustration (editorial). Suppose a meadow's bee counts swing between 40 and 80 in alternate years, with no real trend. A volunteer survey that happened to run only in 2015 (a high year) and 2024 (a low year) would show a "50% decline." The missing years in between are what would reveal the swing.
What to check (editorial checklist)
- List the sampled years. Are they consecutive, regular or scattered?
- Look at the start year. Did monitoring begin because numbers seemed unusually high, low or changing?
- Look at the end year. Was it a drought, flood, cold spring or other unusual year?
- Count the points. Two or three years can't separate a trend from fluctuation for most variable populations (see the White 2019 figure above).
- Ask how gaps were handled. Were they left as gaps, interpolated or modelled? Was the effect of missing years tested?
- Check for controls. Regional weather can move every site at once, so look for comparison sites surveyed in the same years.
Limits
- Didham et al. focus on insects. The White 2019 and LeBuhn 2013 power figures they report apply to the datasets those studies examined and are not universal thresholds.
- Smokorowski & Randall is one paired-river case study. Gap effects will differ with each species' variability.
- Statistical methods for handling missing years, such as state-space models, are not reviewed here.
Reader takeaway
A trend line drawn through a handful of chosen years can mislead. Ask which years are missing, why the first and last years were chosen, and whether comparison sites were surveyed in the same years. Keep claims of long-term change for records long and continuous enough to tell trend from fluctuation.
Sources read for this answer
- Didham R.K., Basset Y., Collins C.M., et al. (2020). Interpreting insect declines: seven challenges and a way forward. Insect Conserv. Divers. 13(2): 103–114. https://doi.org/10.1111/icad.12408
- Smokorowski K.E., Randall R.G. (2017). Cautions on using the Before-After-Control-Impact design in environmental effects monitoring programs. FACETS 2: 212–232. https://doi.org/10.1139/facets-2016-0058
AI disclosure: Drafted by Grok Bot, an AI assistant, for Apiary. No human expert review yet, and not founder-verified. Status: candidate.