Biodiversity monitoring repeats measurements to detect change in living systems. The target might be a species’ abundance, occupancy or distribution; community composition; habitat extent; habitat condition; or a pressure acting on an ecosystem. These quantities are related, but none is a complete substitute for the others.

Define the observable

JNCC guidance starts with the monitoring objective: why evidence is needed, what component of biodiversity will be measured, which method will be used, and where and when sampling will occur. It separates species monitoring from ecosystem and habitat monitoring. Counting abundance, estimating occupancy and mapping habitat answer different questions and require different sampling designs.1 A recorded absence may reflect imperfect detection, while a change in mapped habitat does not establish how a particular species responded.

The UK Biodiversity Indicators combine selected measures to describe status and trends and support national and international reporting. The suite is updated annually and draws on government, research and voluntary-sector data.2 It is an accredited official-statistics compendium, but that accreditation does not automatically apply to every contributing statistic. Indicators compress complex evidence, so their taxa, geography, baseline and uncertainty need to travel with the trend.

JNCC’s monitoring work joins long-running terrestrial schemes, marine surveys, Earth observation and trend analysis.3 Repeated field surveys can identify species and measure local ecological attributes directly. They remain limited by sampling locations, observer effort, detectability, access and taxonomic expertise. Volunteer schemes add valuable duration and coverage, while their statistical models must account for the way records were collected.

Earth observation and habitat proxies

Satellites repeatedly measure reflected or emitted radiation and radar backscatter. From these data, analysts can estimate land cover, vegetation productivity, moisture, structure and change. JNCC describes using NDVI response to detect differences in the management of semi-natural habitats at regional or national scale.4 NDVI is a vegetation-response proxy. Drought, grazing, cutting, nutrient status and seasonal timing can produce similar signals, so a spectral change alone does not identify ecological condition or cause.

Earth observation is strongest for features that are spatially extensive and visible at sensor resolution. Small habitat patches, understorey, rare species, genetic diversity and many freshwater or marine attributes need other methods. Field labels are also needed to train and validate classifications. The resulting map inherits uncertainty from the reference data, pixel or segment boundaries, cloud treatment, predictor variables and model.

Living England

Natural England’s Living England system predicts broad habitat across England from targeted field data, Sentinel-1 and Sentinel-2 imagery, lidar-derived topography, soils, geology and climate data. Its algorithmic transparency record calls the system a production model run on a two-year cycle.5 Random Forest models are trained on 80% of the field dataset, with 20% held out for independent evaluation. The system calibrates relative model scores into class-specific reliability categories.

The 2022-23 model reported 87% overall accuracy, varying across habitat classes and English regions.5 The technical guide records classwise user and producer accuracy and substantial variation among biogeographic zones.6 Overall accuracy can be dominated by common classes, so it is insufficient for deciding whether a particular parcel contains a scarce habitat. Natural England describes the output as a predictive baseline that should be combined with local evidence and advice.

Cloud and shadow create gaps in the optical mosaic. Living England masks these effects, filters radar speckle and can produce predictions where predictor values are missing.5 A filled prediction is still an inference rather than an observation. Users should retain the reliability field and inspect the ground and ancillary evidence behind a decision.

Phase IV, covering 2021-22, reported 88% average habitat-classification accuracy and used a related field, satellite and machine-learning workflow.7 That dated result and the later 87% figure describe different editions and evaluation runs. Neither should be presented as a permanent accuracy for the system or every class.

Extent, condition and trend

Defra’s Environmental Indicator Framework uses Living England Phase VI to report broad habitat extent for 2023. It labels the measure interim and makes no assessment of change because a suitable time series is not yet available.8 Publication of one national map therefore establishes neither a trend nor habitat recovery. Extent also differs from condition, connectivity and species abundance, for which separate measures are required.

A biodiversity-monitoring product should identify its biological target, sampling frame, spatial and temporal support, baseline, detection or classification process, and uncertainty. It should state whether the evidence is a direct field observation, an EO proxy, a modelled class or a composite indicator. This allows national coverage to guide fieldwork and policy without treating a remotely sensed habitat label as a species census or ecological diagnosis.

References

Footnotes

  1. Joint Nature Conservation Committee, Setting biodiversity monitoring objectives. ↩

  2. Joint Nature Conservation Committee, UK Biodiversity Indicators. ↩

  3. Joint Nature Conservation Committee, Monitoring. ↩

  4. Joint Nature Conservation Committee, Habitat Condition. ↩

  5. Natural England, Living England algorithmic transparency record. ↩ ↩2 ↩3

  6. Natural England, Living England: Satellite-based habitat classification technical user guide. ↩

  7. Natural England, Living England: From Satellite Imagery to a National Scale Habitat Map. ↩

  8. Department for Environment, Food & Rural Affairs, Environmental Indicator Framework Theme D: Wildlife. ↩