Data dictionary

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Yield differences between countries are driven by seed, fertiliser, mechanisation and irrigation as much as by weather, so a low yield is not evidence of a dry year. Within one country over time, the weather signal is far clearer — which is the comparison this metric supports. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. An extreme yield on a very small harvested area is usually a crop cut green and weighed fresh rather than a grain harvest, which the FAOSTAT item does not distinguish. Read yields on areas of a few hundred hectares with that in mind.

Not to be confused with: crop-production-maize, crop-area-harvested-maize

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Production reflects area and yield together, so a change in it does not say which moved. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well.

Not to be confused with: crop-yield-maize, crop-area-harvested-maize

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. Area is also not a water figure: two countries with the same area under a crop can differ by an order of magnitude in the water they apply to it.

Not to be confused with: crop-production-maize, land-equipped-for-irrigation, land-actually-irrigated

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Production reflects area and yield together, so a change in it does not say which moved. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well.

Not to be confused with: crop-yield-rice, crop-area-harvested-rice

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Yield differences between countries are driven by seed, fertiliser, mechanisation and irrigation as much as by weather, so a low yield is not evidence of a dry year. Within one country over time, the weather signal is far clearer — which is the comparison this metric supports. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. An extreme yield on a very small harvested area is usually a crop cut green and weighed fresh rather than a grain harvest, which the FAOSTAT item does not distinguish. Read yields on areas of a few hundred hectares with that in mind.

Not to be confused with: crop-production-rice, crop-area-harvested-rice

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. Area is also not a water figure: two countries with the same area under a crop can differ by an order of magnitude in the water they apply to it.

Not to be confused with: crop-production-rice, land-equipped-for-irrigation, land-actually-irrigated

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Yield differences between countries are driven by seed, fertiliser, mechanisation and irrigation as much as by weather, so a low yield is not evidence of a dry year. Within one country over time, the weather signal is far clearer — which is the comparison this metric supports. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. An extreme yield on a very small harvested area is usually a crop cut green and weighed fresh rather than a grain harvest, which the FAOSTAT item does not distinguish. Read yields on areas of a few hundred hectares with that in mind.

Not to be confused with: crop-production-wheat, crop-area-harvested-wheat

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Production reflects area and yield together, so a change in it does not say which moved. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well.

Not to be confused with: crop-yield-wheat, crop-area-harvested-wheat

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. Area is also not a water figure: two countries with the same area under a crop can differ by an order of magnitude in the water they apply to it.

Not to be confused with: crop-production-wheat, land-equipped-for-irrigation, land-actually-irrigated

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Yield differences between countries are driven by seed, fertiliser, mechanisation and irrigation as much as by weather, so a low yield is not evidence of a dry year. Within one country over time, the weather signal is far clearer — which is the comparison this metric supports. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. An extreme yield on a very small harvested area is usually a crop cut green and weighed fresh rather than a grain harvest, which the FAOSTAT item does not distinguish. Read yields on areas of a few hundred hectares with that in mind.

Not to be confused with: crop-production-sorghum, crop-area-harvested-sorghum

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Production reflects area and yield together, so a change in it does not say which moved. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well.

Not to be confused with: crop-yield-sorghum, crop-area-harvested-sorghum

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Yield differences between countries are driven by seed, fertiliser, mechanisation and irrigation as much as by weather, so a low yield is not evidence of a dry year. Within one country over time, the weather signal is far clearer — which is the comparison this metric supports. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. An extreme yield on a very small harvested area is usually a crop cut green and weighed fresh rather than a grain harvest, which the FAOSTAT item does not distinguish. Read yields on areas of a few hundred hectares with that in mind.

Not to be confused with: crop-production-sugar-cane, crop-area-harvested-sugar-cane

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Production reflects area and yield together, so a change in it does not say which moved. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well.

Not to be confused with: crop-yield-sugar-cane, crop-area-harvested-sugar-cane

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. Area is also not a water figure: two countries with the same area under a crop can differ by an order of magnitude in the water they apply to it.

Not to be confused with: crop-production-sorghum, land-equipped-for-irrigation, land-actually-irrigated

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. Area is also not a water figure: two countries with the same area under a crop can differ by an order of magnitude in the water they apply to it.

Not to be confused with: crop-production-sugar-cane, land-equipped-for-irrigation, land-actually-irrigated

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Yield differences between countries are driven by seed, fertiliser, mechanisation and irrigation as much as by weather, so a low yield is not evidence of a dry year. Within one country over time, the weather signal is far clearer — which is the comparison this metric supports. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. An extreme yield on a very small harvested area is usually a crop cut green and weighed fresh rather than a grain harvest, which the FAOSTAT item does not distinguish. Read yields on areas of a few hundred hectares with that in mind.

Not to be confused with: crop-production-seed-cotton, crop-area-harvested-seed-cotton

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Production reflects area and yield together, so a change in it does not say which moved. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well.

Not to be confused with: crop-yield-seed-cotton, crop-area-harvested-seed-cotton

Comparability: A national figure. It says nothing about any farm, field or irrigation scheme, and a national average conceals the irrigated and rain-fed halves of a country that has both. Harvested area excludes what was sown and then lost before harvest, so a severe drought can shrink the area statistic rather than the yield statistic. The worst years are the ones this measure describes least well. Area is also not a water figure: two countries with the same area under a crop can differ by an order of magnitude in the water they apply to it.

Not to be confused with: crop-production-seed-cotton, land-equipped-for-irrigation, land-actually-irrigated

Comparability: Not harvested area. A field cropped twice in a year is one hectare of cropland and two hectares harvested, so harvested area can exceed cropland where multiple cropping is common and falls below it where land lies fallow. Reported without imputation by roughly 85 of 234 countries, against about 228 for forest. Those are the countries with working agricultural statistics rather than a random sample, so a cross-country comparison is partly a comparison of statistical capacity. A national total says nothing about the land cover of any particular basin. A country half forested may have an entirely forested catchment beside an entirely cleared one, and it is the catchment that governs the water.

Not to be confused with: crop-area-harvested-wheat, arable-land-area, land-equipped-for-irrigation

Comparability: A subset of cropland, not an alternative measure of it: adding the two double-counts. The containment holds across all 3,953 published country-years where both are reported. What cropland has and this does not is permanent crops — orchards, vines, plantations — which Hydrionis does not hold as a separate series. Reported without imputation by roughly 85 of 234 countries, against about 228 for forest. Those are the countries with working agricultural statistics rather than a random sample, so a cross-country comparison is partly a comparison of statistical capacity. A national total says nothing about the land cover of any particular basin. A country half forested may have an entirely forested catchment beside an entirely cleared one, and it is the catchment that governs the water.

Not to be confused with: cropland-area, pasture-area

Comparability: Equipped area is a stock that changes slowly and is often carried forward unchanged between reporting years, so a flat series may mean "not resurveyed" rather than "unchanged". It is not a measure of water abstracted, and two countries with equal equipped area can abstract very different volumes.

Not to be confused with: land-actually-irrigated, abstraction-volume-irrigation

Comparability: The least consistently defined of these categories. Where the boundary falls between extensive rangeland, scrub and unused land is a national judgement, and countries with vast arid grazing land have revised this figure by tens of millions of hectares without the land changing. Reported without imputation by roughly 85 of 234 countries, against about 228 for forest. Those are the countries with working agricultural statistics rather than a random sample, so a cross-country comparison is partly a comparison of statistical capacity. A national total says nothing about the land cover of any particular basin. A country half forested may have an entirely forested catchment beside an entirely cleared one, and it is the catchment that governs the water.

Not to be confused with: arable-land-area, cropland-area

Comparability: Reported by very few countries and not on a common definition — some count area irrigated at least once, others area under full irrigation. Sparse coverage means an apparent global trend is usually a change in who reported. It remains an area, never a volume of water.

Not to be confused with: land-equipped-for-irrigation, abstraction-volume-irrigation

Comparability: A community index, not a count. It scores the taxa present against a sensitivity table, so two samples with identical scores can differ tenfold in the number of animals found. Comparable only between samples taken under the same protocol — the Environment Agency’s standardised three-minute kick sample. Not comparable with indices from other countries, which use different taxon tables and different sampling, even where the numeric range looks similar. Strongly seasonal. Spring and autumn samples from one site are not two readings of the same quantity, and a comparison that mixes them will find differences that are the calendar rather than the river. Describes the sampled reach, not the river. Conditions can differ sharply above and below a discharge, a weir or a tributary confluence. The scale is not anchored at zero: WHPT assigns negative scores to the most pollution-tolerant taxa, so a badly degraded sample can return a negative average.

Not to be confused with: whpt-ntaxa, life-family-index, psi-family-score

Comparability: A count of taxa, not of animals: finding one mayfly and finding four hundred both count as one taxon. Comparable only between samples taken under the same protocol — the Environment Agency’s standardised three-minute kick sample. Not comparable with indices from other countries, which use different taxon tables and different sampling, even where the numeric range looks similar. Strongly seasonal. Spring and autumn samples from one site are not two readings of the same quantity, and a comparison that mixes them will find differences that are the calendar rather than the river. Describes the sampled reach, not the river. Conditions can differ sharply above and below a discharge, a weir or a tributary confluence. Sensitive to analytical effort and to the taxonomic level the analyst worked at, so a change in laboratory practice can move it without the river changing.

Not to be confused with: whpt-aspt, life-family-index

Comparability: A count of families, not of insects: one mayfly and four hundred both count once, so richness and abundance can move in opposite directions in the same sample. Counted at family rank throughout, by rolling every species and genus record up through the publisher’s own taxonomy, because the analysts work to mixed ranks and counting raw taxa would measure the depth of identification rather than the river. Records identified no further than the order contribute nothing and are excluded. Comparable only between samples taken under the same protocol — the Environment Agency’s standardised three-minute kick sample. Not comparable with indices from other countries, which use different taxon tables and different sampling, even where the numeric range looks similar. Strongly seasonal. Spring and autumn samples from one site are not two readings of the same quantity, and a comparison that mixes them will find differences that are the calendar rather than the river. Describes the sampled reach, not the river. Conditions can differ sharply above and below a discharge, a weir or a tributary confluence. Richness is bounded by what exists: Britain holds only about thirty EPT families, so the ceiling is low and a change of one or two families is within the variation between visits.

Not to be confused with: whpt-ntaxa, whpt-aspt, life-family-index

Comparability: A community index, not a count. It scores the taxa present against a sensitivity table, so two samples with identical scores can differ tenfold in the number of animals found. A proportion of the community, not a measurement of sediment: it says what the animals indicate, not how much silt is present, and no depth or mass of sediment can be inferred from it. Comparable only between samples taken under the same protocol — the Environment Agency’s standardised three-minute kick sample. Not comparable with indices from other countries, which use different taxon tables and different sampling, even where the numeric range looks similar. Strongly seasonal. Spring and autumn samples from one site are not two readings of the same quantity, and a comparison that mixes them will find differences that are the calendar rather than the river. Describes the sampled reach, not the river. Conditions can differ sharply above and below a discharge, a weir or a tributary confluence. Zero and one hundred are both genuinely reported.

Not to be confused with: whpt-aspt, life-family-index

Comparability: A community index, not a count. It scores the taxa present against a sensitivity table, so two samples with identical scores can differ tenfold in the number of animals found. Comparable only between samples taken under the same protocol — the Environment Agency’s standardised three-minute kick sample. Not comparable with indices from other countries, which use different taxon tables and different sampling, even where the numeric range looks similar. Strongly seasonal. Spring and autumn samples from one site are not two readings of the same quantity, and a comparison that mixes them will find differences that are the calendar rather than the river. Describes the sampled reach, not the river. Conditions can differ sharply above and below a discharge, a weir or a tributary confluence. Designed to respond to flow, but it responds to anything that shifts the community towards still-water taxa — impoundment, siltation and channel works among them. A fall is consistent with low flow and is not evidence of it, and pairing it with a flow record shows whether the two moved together, never that one caused the other.

Not to be confused with: whpt-aspt, psi-family-score, river-flow-mean

Comparability: A radius is not a catchment: a fire fifty kilometres away may be in another basin, and one just outside the circle may sit directly upstream. Points within fifty kilometres of each other share fires, so their values are not independent and must not be summed. A zero means no mapped perimeter fell in the radius, not that nothing burned.

Not to be confused with: crop-area-harvested-wheat

Comparability: Juveniles, not returning adults: a rise is a good spawning year or good survival of young fish, and says nothing directly about the size of the run. A population estimate inferred from the fall in catch between successive electrofishing runs, not a count of fish seen. The inference assumes a closed population and equal catchability, and both assumptions weaken as rivers get deeper, wider or more turbid — so the estimate is least reliable where the river is largest. Held only for electrofishing depletion surveys of at least 50 m², which is about 7% of the Agency’s published fish rows. Single-catch surveys, seine netting and trap catches are real data and are not comparable with a depletion estimate, so they are excluded rather than combined. The Agency publishes a standard error for every estimate and Hydrionis cannot carry it: a series point holds a date and a value, so a density of 14 and a density of 14 give or take 9 are stored identically. Treat single points as indicative and trends as the readable signal. Describes a surveyed reach on one day, not the river or the catchment. Fish move, and a reach fished in August says little about the same reach in March. Density is an extrapolation to 100 m². The largest values in the record come from small reaches of dense juvenile habitat, where the arithmetic is correct and the extrapolation to a whole river would not be. Strongly seasonal and strongly age-structured — fry counts collapse and recover annually by nature, so a fall between two years is not a decline.

Not to be confused with: fish-density-brown-trout, fish-density-european-eel

Comparability: Resident brown trout and migratory sea trout are one species and are not separated here: the Agency records them together because juveniles of the two are not reliably distinguishable in the field. A change therefore cannot be attributed to the migratory form. A population estimate inferred from the fall in catch between successive electrofishing runs, not a count of fish seen. The inference assumes a closed population and equal catchability, and both assumptions weaken as rivers get deeper, wider or more turbid — so the estimate is least reliable where the river is largest. Held only for electrofishing depletion surveys of at least 50 m², which is about 7% of the Agency’s published fish rows. Single-catch surveys, seine netting and trap catches are real data and are not comparable with a depletion estimate, so they are excluded rather than combined. The Agency publishes a standard error for every estimate and Hydrionis cannot carry it: a series point holds a date and a value, so a density of 14 and a density of 14 give or take 9 are stored identically. Treat single points as indicative and trends as the readable signal. Describes a surveyed reach on one day, not the river or the catchment. Fish move, and a reach fished in August says little about the same reach in March. Density is an extrapolation to 100 m². The largest values in the record come from small reaches of dense juvenile habitat, where the arithmetic is correct and the extrapolation to a whole river would not be.

Not to be confused with: fish-density-atlantic-salmon, fish-density-european-eel

Comparability: Excludes elvers and glass eels, which the Agency records separately: this is the growing yellow and silver stage only, so it is not a measure of recruitment arriving from the sea. Eels hide in the substrate and are among the hardest fish to catch by electrofishing, so catchability is low and variable and the depletion estimate is correspondingly weaker than for salmonids. A population estimate inferred from the fall in catch between successive electrofishing runs, not a count of fish seen. The inference assumes a closed population and equal catchability, and both assumptions weaken as rivers get deeper, wider or more turbid — so the estimate is least reliable where the river is largest. Held only for electrofishing depletion surveys of at least 50 m², which is about 7% of the Agency’s published fish rows. Single-catch surveys, seine netting and trap catches are real data and are not comparable with a depletion estimate, so they are excluded rather than combined. The Agency publishes a standard error for every estimate and Hydrionis cannot carry it: a series point holds a date and a value, so a density of 14 and a density of 14 give or take 9 are stored identically. Treat single points as indicative and trends as the readable signal. Describes a surveyed reach on one day, not the river or the catchment. Fish move, and a reach fished in August says little about the same reach in March. Density is an extrapolation to 100 m². The largest values in the record come from small reaches of dense juvenile habitat, where the arithmetic is correct and the extrapolation to a whole river would not be.

Not to be confused with: fish-density-atlantic-salmon, fish-density-brown-trout

Comparability: Not wetland extent. Wetlands include seasonally inundated ground, peat, marsh and floodplain that is not open water for most of the year, and none of that is counted here; a country can lose most of its wetlands with this figure unmoved. Not a volume either — a shallow lake and a deep one of equal surface hold vastly different amounts of water, and this measures neither depth nor flow. Reported by 77 of 234 countries without imputation, so the absence of a country is the usual case rather than the exception. Where a country reports, the figure is almost always official: 1,227 of 1,237 held values.

Not to be confused with: reservoir-capacity, river-flow-mean

Comparability: This is the whole of which the other three are parts: naturally regenerating and planted forest sum to it, and primary forest sits inside naturally regenerating. The categories are nested, so adding any of them to this total double-counts. A land-use class, not a measure of tree canopy. Land felled, burned or otherwise temporarily unstocked but expected to regenerate is still counted at its full area, so this series does not register a fire or a clear-cut — only a change of land use. Observed at Global Forest Resources Assessment reference years only — 1990, 2000, 2010, 2015, 2020 and 2025. Values FAO interpolates between assessments are not held, so the gaps are real gaps and no change can be dated more finely than the interval it falls in. A national total says nothing about the land cover of any particular basin. A country half forested may have an entirely forested catchment beside an entirely cleared one, and it is the catchment that governs the water. Definitions were harmonised across the assessment cycle, but earlier national inventories differed in how they treated open woodland and trees outside forests, so the longest comparisons are the weakest.

Not to be confused with: forest-area-naturally-regenerating, forest-area-planted, forest-area-primary, burned-area-within-50km

Comparability: Nested, not independent: this and planted forest sum to forest area, and primary forest is contained within this. Adding the categories together double-counts. A land-use class, not a measure of tree canopy. Land felled, burned or otherwise temporarily unstocked but expected to regenerate is still counted at its full area, so this series does not register a fire or a clear-cut — only a change of land use. Observed at Global Forest Resources Assessment reference years only — 1990, 2000, 2010, 2015, 2020 and 2025. Values FAO interpolates between assessments are not held, so the gaps are real gaps and no change can be dated more finely than the interval it falls in. A national total says nothing about the land cover of any particular basin. A country half forested may have an entirely forested catchment beside an entirely cleared one, and it is the catchment that governs the water.

Not to be confused with: forest-area, forest-area-planted, forest-area-primary

Comparability: Nested with naturally regenerating forest, which it sums with to give forest area; the two must not be added to the total. A land-use class, not a measure of tree canopy. Land felled, burned or otherwise temporarily unstocked but expected to regenerate is still counted at its full area, so this series does not register a fire or a clear-cut — only a change of land use. Observed at Global Forest Resources Assessment reference years only — 1990, 2000, 2010, 2015, 2020 and 2025. Values FAO interpolates between assessments are not held, so the gaps are real gaps and no change can be dated more finely than the interval it falls in. A national total says nothing about the land cover of any particular basin. A country half forested may have an entirely forested catchment beside an entirely cleared one, and it is the catchment that governs the water. That plantations can reduce low flows is a finding from instrumented catchments elsewhere, not something this national series demonstrates or could demonstrate.

Not to be confused with: forest-area, forest-area-naturally-regenerating, forest-area-primary

Comparability: A subset of naturally regenerating forest, never additional to it. A land-use class, not a measure of tree canopy. Land felled, burned or otherwise temporarily unstocked but expected to regenerate is still counted at its full area, so this series does not register a fire or a clear-cut — only a change of land use. Observed at Global Forest Resources Assessment reference years only — 1990, 2000, 2010, 2015, 2020 and 2025. Values FAO interpolates between assessments are not held, so the gaps are real gaps and no change can be dated more finely than the interval it falls in. A national total says nothing about the land cover of any particular basin. A country half forested may have an entirely forested catchment beside an entirely cleared one, and it is the catchment that governs the water. Assessing "no clearly visible indications of human activity" is a judgement that varies between countries and between assessments by the same country, so a step change in this series may record a reassessment rather than a clearance.

Not to be confused with: forest-area, forest-area-naturally-regenerating, forest-area-planted

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. A calendar-year breeding-season index, so it reflects the conditions of that spring and the winter before it.

Not to be confused with: bird-index-fast-flowing-water, bird-index-slow-standing-water, bird-index-wet-meadow, bird-index-reedbed

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. A calendar-year breeding-season index, so it reflects the conditions of that spring and the winter before it.

Not to be confused with: bird-index-water-and-wetland, bird-index-slow-standing-water, bird-index-wet-meadow, bird-index-reedbed

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. A calendar-year breeding-season index, so it reflects the conditions of that spring and the winter before it.

Not to be confused with: bird-index-water-and-wetland, bird-index-fast-flowing-water, bird-index-wet-meadow, bird-index-reedbed

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. A calendar-year breeding-season index, so it reflects the conditions of that spring and the winter before it.

Not to be confused with: bird-index-water-and-wetland, bird-index-fast-flowing-water, bird-index-slow-standing-water, bird-index-reedbed

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. A calendar-year breeding-season index, so it reflects the conditions of that spring and the winter before it.

Not to be confused with: bird-index-water-and-wetland, bird-index-fast-flowing-water, bird-index-slow-standing-water, bird-index-wet-meadow

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. Runs on a survey year from July to June, labelled by the year it begins, so the 1975 point is the winter of 1975–76 and aligning it to a calendar year is off by up to half a year.

Not to be confused with: bird-index-wintering-wildfowl, bird-index-wintering-waders

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. Runs on a survey year from July to June, labelled by the year it begins, so the 1975 point is the winter of 1975–76 and aligning it to a calendar year is off by up to half a year.

Not to be confused with: bird-index-wintering-waterbirds, bird-index-wintering-waders

Comparability: An index, not a count and not a population. Each series is set to 100 in its own first year and describes change from that point, so it says how the group moved and never how many birds there are. Levels are not comparable between indices: each is based to its own start year, so one index standing higher than another means it grew more since 1975, not that it contains more birds. Shapes may be compared; heights may not. A composite of several species. It can stay level while some members collapse and others increase, so a flat index is not evidence that nothing changed. The unsmoothed index, as published. Defra also publishes a smoothed trend line with confidence intervals; Hydrionis holds neither, because a smoothed line is a model with the year-to-year variation removed and that variation is where a drought or a hard winter shows. Two years are absent and are meant to be: no value is published for 2001, when foot-and-mouth disease closed the countryside, or for 2020, when the pandemic cancelled fieldwork. The surveys did not happen, so there is nothing to record, and the gaps must not be filled. The publisher’s smoothed series does carry values for both years, which is one reason Hydrionis holds the unsmoothed one. The published confidence interval is not carried — a series point holds a date and a value with nowhere to put uncertainty — so single years should be read as indicative and the trend as the signal. Built from volunteer schemes with fixed sites and repeat visits, weighted for coverage by the publisher. Site coverage is not uniform across the UK, and upland and remote areas are less well covered than lowland England. Runs on a survey year from July to June, labelled by the year it begins, so the 1975 point is the winter of 1975–76 and aligning it to a calendar year is off by up to half a year.

Not to be confused with: bird-index-wintering-waterbirds, bird-index-wintering-wildfowl

Comparability: Compliance is computed over the samples a supplier was required to take, and sampling frequency scales with population served. A small supply taking twelve samples a year and a city taking twelve thousand can both report 100%, and one failure moves the small supply far further. Compliance percentages are not comparable between supplies of very different size, and not comparable between countries whose sampling regimes and parameter sets differ.

Comparability: The parameter set behind a physico-chemical compliance figure differs between regulatory regimes, so the same percentage does not represent the same test in two countries. Within one regime, sampling frequency scales with population served, so small and large supplies are not directly comparable.

Comparability: A count of results Hydrionis holds, which is the count the publisher exposed through its API. It is not a statement about how many samples a supplier was legally required to take.

Comparability: A mean over quantified samples only. Where most samples are censored this describes the detectable minority, not the supply. Not comparable with a mean computed by substituting zero or half the detection limit for censored results, which is the more common convention elsewhere.

Comparability: Depends on the detection limit laboratories achieved, which improves over time. A rise can mean more of the substance is present or that it became visible at lower concentrations, and the two cannot be separated from this series alone.

Comparability: Computed against the French regulatory limit of 50 mg/l as NO₃. Limits for the same substance differ between jurisdictions, so exceedance rates are not comparable internationally without checking both standards.

Comparability: Depends on the detection limit laboratories achieved, which improves over time. A rise can mean more of the substance is present or that it became visible at lower concentrations, and the two cannot be separated from this series alone.

Comparability: Computed against the French regulatory limit of 10 µg/l. Limits for the same substance differ between jurisdictions, so exceedance rates are not comparable internationally without checking both standards.

Comparability: Depends on the detection limit laboratories achieved, which improves over time. A rise can mean more of the substance is present or that it became visible at lower concentrations, and the two cannot be separated from this series alone.

Comparability: Computed against the French regulatory limit of 0.5 µg/l. Limits for the same substance differ between jurisdictions, so exceedance rates are not comparable internationally without checking both standards.

Comparability: A mean over quantified samples only. Where most samples are censored this describes the detectable minority, not the supply. Not comparable with a mean computed by substituting zero or half the detection limit for censored results, which is the more common convention elsewhere.

Comparability: A mean over quantified samples only. Where most samples are censored this describes the detectable minority, not the supply. Not comparable with a mean computed by substituting zero or half the detection limit for censored results, which is the more common convention elsewhere.

Comparability: Depends on the detection limit laboratories achieved, which improves over time. A rise can mean more of the substance is present or that it became visible at lower concentrations, and the two cannot be separated from this series alone.

Comparability: Computed against the French regulatory limit of 10 µg/l. Limits for the same substance differ between jurisdictions, so exceedance rates are not comparable internationally without checking both standards.

Comparability: A mean over quantified samples only. Where most samples are censored this describes the detectable minority, not the supply. Not comparable with a mean computed by substituting zero or half the detection limit for censored results, which is the more common convention elsewhere.

Comparability: Depends on the detection limit laboratories achieved, which improves over time. A rise can mean more of the substance is present or that it became visible at lower concentrations, and the two cannot be separated from this series alone.

Comparability: A mean over quantified samples only. Where most samples are censored this describes the detectable minority, not the supply. Not comparable with a mean computed by substituting zero or half the detection limit for censored results, which is the more common convention elsewhere.

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: Reported abstraction, not licensed entitlement and not consumption. Water abstracted is largely returned to the environment after use; the share consumed is a different and much smaller quantity. Coverage depends on which abstractors are required to declare, so a national total is a total of what is declared rather than of what is taken.

Not to be confused with: water-withdrawal-total, abstraction-volume-licensed

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: A component of total abstraction by the kind of water body it came from, not by what it was used for. Surface and groundwater are reported separately by the publisher and are not derived from one another here.

Not to be confused with: abstraction-volume-licensed

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: A component of total abstraction by the kind of water body it came from, not by what it was used for. Surface and groundwater are reported separately by the publisher and are not derived from one another here.

Not to be confused with: abstraction-volume-licensed

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: Not consumption. The difference between gross and net abstraction is water returned without use; the difference between net abstraction and consumption is water returned after use, which is far larger and is not reported here.

Not to be confused with: abstraction-volume-total, water-withdrawal-total

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: Reported by some countries and not by others, so an absent value means "not reported" rather than zero. Comparing gross abstraction across a country that reports this and one that does not compares two different quantities.

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: Reported abstraction for a declared purpose, not consumption. Purpose categories are defined by the reporting jurisdiction and do not map exactly onto categories used elsewhere, so a single purpose should not be compared across countries without checking both definitions.

Not to be confused with: abstraction-volume-licensed

Comparability: Saturation above 100% is not better than 100%. Strong daytime over-saturation and severe night-time depletion frequently occur in the same enriched river, and a daytime spot sample will not show the depletion.

Comparability: Includes nitrite as well as nitrate, so it is always at least as large as nitrate alone and the two must not be plotted as one series. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: nitrate-as-n, nitrite-as-n, nitrate

Comparability: Oxygen solubility falls as temperature and salinity rise, so an absolute concentration must be read alongside temperature. Percentage saturation is often the more meaningful figure.

Comparability: Reported as phosphorus, not as the phosphate ion: multiply by 3.07 to express as PO₄. Reactive orthophosphate is a smaller quantity than total phosphorus, so the two must not be compared.

Not to be confused with: phosphate

Comparability: Reported as nitrogen, not as the nitrate ion: multiply by 4.43 to compare with a drinking-water standard written as NO₃. Concentrations are specific to the sampling point and vary strongly with river flow, so a wet year and a dry year are not directly comparable.

Not to be confused with: nitrate

Comparability: Usually near or below the detection limit, so series are heavily censored. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: nitrate-as-n, total-oxidised-nitrogen

Comparability: Reported as nitrogen rather than as the ammonium ion. The proportion present as toxic un-ionised ammonia rises sharply with pH and temperature, so the same concentration is not equally harmful in different rivers.

Not to be confused with: ammonium

Comparability: Extremely flow-dependent: a single storm can carry more sediment than the preceding months, so a mean over samples taken on ordinary days understates the load a river actually moves. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: turbidity

Comparability: Measured with nitrification suppressed, which is not the same test as an unsuppressed BOD and gives a lower number; the two must not be combined. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated.

Not to be confused with: dissolved-oxygen, dissolved-organic-carbon

Comparability: Always at least as large as orthophosphate at the same point and time, because it includes it. River phosphorus standards are written against orthophosphate, so this is the wrong series to compare against them. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: orthophosphate-as-p

Comparability: Turbidity is an operational indicator, not a health parameter in itself, but it matters greatly because particles shield micro-organisms from disinfection.

Comparability: pH varies with temperature and with dissolved carbon dioxide, so a sample measured hours after collection may not reflect the source. Averaging pH values arithmetically is strictly incorrect, since the scale is logarithmic.

Not to be confused with: ph

Comparability: A major ion, describing the geology and the human inputs a catchment carries rather than a pollutant in itself. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: sulphate-river, sodium-dissolved, calcium-dissolved, magnesium-dissolved, potassium-dissolved

Comparability: Conductivity is strongly temperature dependent and is conventionally reported normalised to 25 °C. An unnormalised reading is not comparable.

Comparability: Rising across much of upland Britain over recent decades, which is a genuine and much-studied change rather than a monitoring artefact. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: bod-5-day

Comparability: A major ion, describing the geology and the human inputs a catchment carries rather than a pollutant in itself. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: chloride-river, sulphate-river, calcium-dissolved, magnesium-dissolved, potassium-dissolved

Comparability: A major ion, describing the geology and the human inputs a catchment carries rather than a pollutant in itself. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: chloride-river, sulphate-river, sodium-dissolved, calcium-dissolved, magnesium-dissolved

Comparability: A major ion, describing the geology and the human inputs a catchment carries rather than a pollutant in itself. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: chloride-river, sodium-dissolved, calcium-dissolved, magnesium-dissolved, potassium-dissolved

Comparability: A major ion, describing the geology and the human inputs a catchment carries rather than a pollutant in itself. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: chloride-river, sulphate-river, sodium-dissolved, magnesium-dissolved, potassium-dissolved

Comparability: A major ion, describing the geology and the human inputs a catchment carries rather than a pollutant in itself. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: chloride-river, sulphate-river, sodium-dissolved, calcium-dissolved, potassium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, nickel-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river. Listed as a priority substance under the Water Framework Directive, so an environmental quality standard applies — but that standard is often written against the bioavailable fraction rather than this one.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river. Listed as a priority substance under the Water Framework Directive, so an environmental quality standard applies — but that standard is often written against the bioavailable fraction rather than this one.

Not to be confused with: cadmium-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river. Listed as a priority substance under the Water Framework Directive, so an environmental quality standard applies — but that standard is often written against the bioavailable fraction rather than this one.

Not to be confused with: lead-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river.

Not to be confused with: lead-dissolved, cadmium-dissolved, mercury-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved

Comparability: The dissolved fraction only. The archive also publishes a total concentration including metal bound to suspended particles, a modelled bioavailable fraction, and a dry-weight concentration in riverbed sediment measured in milligrams per kilogram; none of those is this, and mixing them confuses quantities that differ by orders of magnitude. Frequently reported as below the laboratory detection limit. Those results are held as censored observations carrying their limit rather than as zero or as the limit itself, and any statistic computed over them records how they were treated. A river measurement, not a drinking-water one. River standards protect aquatic life and are written differently from the limits that apply to water leaving a treatment works, so this must never be read against a drinking-water limit. Specific to the sampling point and the moment the bottle was filled. River chemistry varies with flow, season and the distance below the last discharge, so a single result describes that reach on that day and not the river. Listed as a priority substance under the Water Framework Directive, so an environmental quality standard applies — but that standard is often written against the bioavailable fraction rather than this one.

Not to be confused with: lead-dissolved, cadmium-dissolved, nickel-dissolved, zinc-dissolved, copper-dissolved, arsenic-dissolved, chromium-dissolved, iron-dissolved, manganese-dissolved, aluminium-dissolved

Comparability: An air concentration, not a water concentration. Micrograms per cubic metre of air is a different quantity in a different medium from milligrams per litre of water, and no value here can be read as saying what is dissolved in a river. Measured on a filter pack exposed for a week at a rural site, so it is a weekly average of the air passing one place rather than a peak or an instantaneous reading. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. Weeks in which the substance was below the detection limit are held as censored observations: the week keeps its place, it carries no number because no exact value exists, and it carries the operator, the threshold and the fact that the limit is one of detection. A reader sees "<0.06 µg/m³ — below the analytical detection limit" rather than a gap or a number nobody measured. This affects 3.2% of nitrate weeks, 1.3% of sulphur dioxide weeks and under 1% of the rest. Interchanges with particulate nitrate on the filter with temperature, so the split between the two is less reliable than their sum.

Not to be confused with: air-particulate-nitrate, nitric-acid-deposition-dry

Comparability: An air concentration, not a water concentration. Micrograms per cubic metre of air is a different quantity in a different medium from milligrams per litre of water, and no value here can be read as saying what is dissolved in a river. Measured on a filter pack exposed for a week at a rural site, so it is a weekly average of the air passing one place rather than a peak or an instantaneous reading. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. Weeks in which the substance was below the detection limit are held as censored observations: the week keeps its place, it carries no number because no exact value exists, and it carries the operator, the threshold and the fact that the limit is one of detection. A reader sees "<0.06 µg/m³ — below the analytical detection limit" rather than a gap or a number nobody measured. This affects 3.2% of nitrate weeks, 1.3% of sulphur dioxide weeks and under 1% of the rest. Particulate nitrate and nitric acid interchange on the filter with temperature, so the two are reported separately and the split between them is less reliable than their sum.

Not to be confused with: air-nitric-acid, nitrate-deposition-dry

Comparability: An air concentration, not a water concentration. Micrograms per cubic metre of air is a different quantity in a different medium from milligrams per litre of water, and no value here can be read as saying what is dissolved in a river. Measured on a filter pack exposed for a week at a rural site, so it is a weekly average of the air passing one place rather than a peak or an instantaneous reading. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. Weeks in which the substance was below the detection limit are held as censored observations: the week keeps its place, it carries no number because no exact value exists, and it carries the operator, the threshold and the fact that the limit is one of detection. A reader sees "<0.06 µg/m³ — below the analytical detection limit" rather than a gap or a number nobody measured. This affects 3.2% of nitrate weeks, 1.3% of sulphur dioxide weeks and under 1% of the rest.

Not to be confused with: air-sulphur-dioxide, sulphate-deposition-dry

Comparability: An air concentration, not a water concentration. Micrograms per cubic metre of air is a different quantity in a different medium from milligrams per litre of water, and no value here can be read as saying what is dissolved in a river. Measured on a filter pack exposed for a week at a rural site, so it is a weekly average of the air passing one place rather than a peak or an instantaneous reading. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. Weeks in which the substance was below the detection limit are held as censored observations: the week keeps its place, it carries no number because no exact value exists, and it carries the operator, the threshold and the fact that the limit is one of detection. A reader sees "<0.06 µg/m³ — below the analytical detection limit" rather than a gap or a number nobody measured. This affects 3.2% of nitrate weeks, 1.3% of sulphur dioxide weeks and under 1% of the rest.

Not to be confused with: air-particulate-nitrate, ammonium-deposition-dry

Comparability: A modelled quantity, not a measurement. The flux is the measured air concentration multiplied by a deposition velocity the publisher models from meteorology, land cover and surface properties. The concentration is observed; the velocity is not, and the product inherits every assumption in it. The deposition velocity is substituted where the inputs to model it are missing, and the publisher reports how much of each week was substituted. Across the record 38% of weeks are fully substituted, and in 2017 to 2021 that reaches 85 to 87% — so the series carries a change in how it was produced partway through, and a trend read across that change is partly a trend in method. Dry deposition only. What rain brings down is measured by a separate network and is not held, so nothing here is a total deposition figure and total cannot be computed from it. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. A flux is not a concentration: this says how much reached a hectare of ground in a week, not how much was in the air.

Not to be confused with: sulphate-deposition-dry, sulphur-dioxide-deposition-dry, nitrate-deposition-dry, ammonium-deposition-dry, air-particulate-sulphate, air-sulphur-dioxide

Comparability: A modelled quantity, not a measurement. The flux is the measured air concentration multiplied by a deposition velocity the publisher models from meteorology, land cover and surface properties. The concentration is observed; the velocity is not, and the product inherits every assumption in it. The deposition velocity is substituted where the inputs to model it are missing, and the publisher reports how much of each week was substituted. Across the record 38% of weeks are fully substituted, and in 2017 to 2021 that reaches 85 to 87% — so the series carries a change in how it was produced partway through, and a trend read across that change is partly a trend in method. Dry deposition only. What rain brings down is measured by a separate network and is not held, so nothing here is a total deposition figure and total cannot be computed from it. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. A flux is not a concentration: this says how much reached a hectare of ground in a week, not how much was in the air.

Not to be confused with: sulphur-dioxide-deposition-dry, nitrate-deposition-dry, nitric-acid-deposition-dry, ammonium-deposition-dry, air-particulate-sulphate, air-sulphur-dioxide

Comparability: A modelled quantity, not a measurement. The flux is the measured air concentration multiplied by a deposition velocity the publisher models from meteorology, land cover and surface properties. The concentration is observed; the velocity is not, and the product inherits every assumption in it. The deposition velocity is substituted where the inputs to model it are missing, and the publisher reports how much of each week was substituted. Across the record 38% of weeks are fully substituted, and in 2017 to 2021 that reaches 85 to 87% — so the series carries a change in how it was produced partway through, and a trend read across that change is partly a trend in method. Dry deposition only. What rain brings down is measured by a separate network and is not held, so nothing here is a total deposition figure and total cannot be computed from it. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. A flux is not a concentration: this says how much reached a hectare of ground in a week, not how much was in the air.

Not to be confused with: sulphate-deposition-dry, sulphur-dioxide-deposition-dry, nitric-acid-deposition-dry, ammonium-deposition-dry, air-particulate-sulphate, air-sulphur-dioxide

Comparability: A modelled quantity, not a measurement. The flux is the measured air concentration multiplied by a deposition velocity the publisher models from meteorology, land cover and surface properties. The concentration is observed; the velocity is not, and the product inherits every assumption in it. The deposition velocity is substituted where the inputs to model it are missing, and the publisher reports how much of each week was substituted. Across the record 38% of weeks are fully substituted, and in 2017 to 2021 that reaches 85 to 87% — so the series carries a change in how it was produced partway through, and a trend read across that change is partly a trend in method. Dry deposition only. What rain brings down is measured by a separate network and is not held, so nothing here is a total deposition figure and total cannot be computed from it. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. A flux is not a concentration: this says how much reached a hectare of ground in a week, not how much was in the air.

Not to be confused with: sulphate-deposition-dry, sulphur-dioxide-deposition-dry, nitrate-deposition-dry, nitric-acid-deposition-dry, air-particulate-sulphate, air-sulphur-dioxide

Comparability: An air concentration, not a water concentration. Micrograms per cubic metre of air is a different quantity in a different medium from milligrams per litre of water, and no value here can be read as saying what is dissolved in a river. Measured on a filter pack exposed for a week at a rural site, so it is a weekly average of the air passing one place rather than a peak or an instantaneous reading. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. Weeks in which the substance was below the detection limit are held as censored observations: the week keeps its place, it carries no number because no exact value exists, and it carries the operator, the threshold and the fact that the limit is one of detection. A reader sees "<0.06 µg/m³ — below the analytical detection limit" rather than a gap or a number nobody measured. This affects 3.2% of nitrate weeks, 1.3% of sulphur dioxide weeks and under 1% of the rest. The share of weeks below detection did rise as air cleaned, from 0.3% in the 1980s to 2.3% in the 2020s, but it remains small enough that the omission cannot account for the observed decline.

Not to be confused with: air-particulate-sulphate, sulphur-dioxide-deposition-dry

Comparability: A modelled quantity, not a measurement. The flux is the measured air concentration multiplied by a deposition velocity the publisher models from meteorology, land cover and surface properties. The concentration is observed; the velocity is not, and the product inherits every assumption in it. The deposition velocity is substituted where the inputs to model it are missing, and the publisher reports how much of each week was substituted. Across the record 38% of weeks are fully substituted, and in 2017 to 2021 that reaches 85 to 87% — so the series carries a change in how it was produced partway through, and a trend read across that change is partly a trend in method. Dry deposition only. What rain brings down is measured by a separate network and is not held, so nothing here is a total deposition figure and total cannot be computed from it. CASTNET sites are deliberately rural and away from local sources, chosen to represent regional air rather than the air anybody breathes in a town. They are not urban air quality and should not be read as exposure. A flux is not a concentration: this says how much reached a hectare of ground in a week, not how much was in the air.

Not to be confused with: sulphate-deposition-dry, nitrate-deposition-dry, nitric-acid-deposition-dry, ammonium-deposition-dry, air-particulate-sulphate, air-sulphur-dioxide

Comparability: One index among several that disagree by design. It must not be combined with the others into a single drought score, and a place not in drought by this measure may well be in drought by another. PDSI is also known to behave differently between climates — it was calibrated on the central United States and is less reliable in snow-dominated and arid regions, where the moisture balance it assumes fits less well.

Not to be confused with: phdi, spi-1, spi-12

Comparability: One index among several that disagree by design. It must not be combined with the others into a single drought score, and a place not in drought by this measure may well be in drought by another. PHDI lagging PDSI is the expected behaviour, not a discrepancy: a month where the weather has recovered and the water supply has not is a real state of the world and both indices are describing it correctly.

Not to be confused with: pdsi, spi-12

Comparability: One index among several that disagree by design. It must not be combined with the others into a single drought score, and a place not in drought by this measure may well be in drought by another. A single dry month is not a drought. SPI-1 is noisy by construction and is most useful beside a longer window, where the disagreement between them is the finding.

Not to be confused with: spi-12, pdsi, precipitation-anomaly

Comparability: SPI measures meteorological drought only. It says nothing about whether reservoirs, rivers or aquifers are actually depleted, which can lag rainfall by months to years. It is one index among several that disagree by design and must not be combined with them into a single drought score. Consecutive months share eleven months of input, so a run of low values is one dry year being reported repeatedly, not a run of independent dry periods.

Not to be confused with: spi-1, phdi, pdsi

Comparability: This is an assessment, not a measurement and not a legal declaration. The Drought Monitor is produced weekly by author agencies who combine several indicators with reports from local observers, so it carries expert judgement that no single index does — and it is not a statement that any authority has declared a drought or imposed a restriction. It describes the share of *land area*, which is not the share of people, farmland or water supply affected: a state can be 80% in drought by area with most of its population unaffected.

Comparability: This is an assessment, not a measurement and not a legal declaration. The Drought Monitor is produced weekly by author agencies who combine several indicators with reports from local observers, so it carries expert judgement that no single index does — and it is not a statement that any authority has declared a drought or imposed a restriction. It describes the share of *land area*, which is not the share of people, farmland or water supply affected: a state can be 80% in drought by area with most of its population unaffected.

Comparability: This is an assessment, not a measurement and not a legal declaration. The Drought Monitor is produced weekly by author agencies who combine several indicators with reports from local observers, so it carries expert judgement that no single index does — and it is not a statement that any authority has declared a drought or imposed a restriction. It describes the share of *land area*, which is not the share of people, farmland or water supply affected: a state can be 80% in drought by area with most of its population unaffected.

Comparability: This is an assessment, not a measurement and not a legal declaration. The Drought Monitor is produced weekly by author agencies who combine several indicators with reports from local observers, so it carries expert judgement that no single index does — and it is not a statement that any authority has declared a drought or imposed a restriction. It describes the share of *land area*, which is not the share of people, farmland or water supply affected: a state can be 80% in drought by area with most of its population unaffected.

Comparability: This is an assessment, not a measurement and not a legal declaration. The Drought Monitor is produced weekly by author agencies who combine several indicators with reports from local observers, so it carries expert judgement that no single index does — and it is not a statement that any authority has declared a drought or imposed a restriction. It describes the share of *land area*, which is not the share of people, farmland or water supply affected: a state can be 80% in drought by area with most of its population unaffected.

Comparability: Minima are more sensitive to local siting than maxima: a frost hollow reads several degrees colder than a slope a few hundred metres away, and that difference is real for the site but not transferable to its surroundings.

Not to be confused with: temperature-min-monthly

Comparability: Screen height, exposure and instrument type differ between networks and have changed over long records. Urban stations warm relative to rural ones as the settlement around them grows, which is a real change at that station but not a regional climate signal.

Not to be confused with: temperature-max-monthly, water-temperature

Comparability: Strongly seasonal. Compare a month against the same month in other years, never against an adjacent month. Included because temperature drives evaporative demand and water demand, not as general climate reporting.

Comparability: A mean of daily maxima, not the highest temperature recorded in the month. The two are frequently confused and differ substantially.

Comparability: A mean of daily minima, not the lowest temperature recorded in the month.

Comparability: Held because solar radiation drives evaporation. The measurement basis changed from Campbell-Stokes recorders to electronic sensors over the record, which introduces a step that is not a climate signal.

Comparability: Included because temperature drives evaporation, evapotranspiration and water demand, not as general climate reporting. Station siting and urban heat effects both influence local records.

Comparability: A facility figure, never a company total. Two plants with the same generation can differ a hundredfold in withdrawal purely by cooling design, so withdrawal ranks cooling technology at least as much as it ranks water use.

Not to be confused with: facility-water-consumption, abstraction-volume-total

Comparability: Consumption and withdrawal are not alternative measurements of one quantity and must never be substituted for one another. A facility ranked highest on withdrawal is frequently well down the consumption ranking, and the reverse.

Not to be confused with: facility-water-withdrawal

Comparability: A high agricultural share reflects the size and irrigation intensity of the farming sector rather than inefficiency. Countries at the top of any such ranking are typically food producers in arid regions.

Comparability: The municipal share rises where agriculture and industry are small, so a high figure often indicates a service-based economy rather than high household consumption.

Comparability: Countries differ in whether thermoelectric cooling is counted as industrial withdrawal. Where it is included, the industrial share can exceed 50% in countries with large freshwater-cooled power fleets, which reflects cooling arrangements rather than manufacturing intensity.

Comparability: A component of withdrawal identified by its source, not an additional volume. Adding it to withdrawal would double-count it.

Not to be confused with: facility-water-withdrawal

Comparability: Definitions differ on whether supply pipe losses, garden use and non-household use on domestic meters are included. Metered and unmetered households also differ substantially, so a national average masks a wide distribution.

Comparability: Flow is a property of a specific gauge on a specific reach, and cannot be attributed to a river as a whole. Two gauges on the same river can differ by an order of magnitude. Where a station measures flow through a sluice, lock or pumping station — much of the Dutch network does — discharge is signed and a negative value means water moving in the reverse direction, which is a real operating state and not an error. Managed flow of that kind is not comparable with discharge at a natural gauging station, because it reflects operational decisions as well as hydrology.

Comparability: A gauge property, not a river property. Strongly seasonal, so compare a month against the same month in other years. Gauge records contain breaks where structures were rebuilt or ratings revised.

Comparability: The annual maximum series is the basis of flood frequency analysis. It belongs to a specific gauge on a specific reach, and cannot be attributed to a river as a whole. Rating curves are extrapolated at extreme flows, so the largest values carry the largest uncertainty.

Comparability: Level is measured against each station’s own zero point, which is set locally and is not a common datum. Two gauges reading 3 m are not at the same height, and a level is not a depth of water: it says where the surface sits relative to a reference, not how deep the river is. Read a level against that gauge’s own record, never against another gauge’s.

Comparability: Not interchangeable with the calendar-year mean held alongside it. They are different statistics over different twelve-month windows and they disagree: at the Yampa at Steamboat Springs, 1911 is 403 m³/s by water year and 412 by calendar year. The water year exists because a snow-fed river’s season begins with the autumn that builds the snowpack, so a calendar year cuts the accumulation off from the melt it produces.

Comparability: Catchment areas are not all the same quantity. The topographic area drains to the station on the surface; the contributing area excludes parts that do not deliver water to it in an ordinary year — closed depressions, karst losing ground, land above a diversion. On many rivers the two agree and on some they differ greatly, and dividing a discharge by the wrong one produces a runoff depth that is wrong by that ratio. Which figure a value reports is recorded on it. An area is also a statement of its own date: a catchment is redrawn when a survey improves or a transfer scheme moves water between basins.

Comparability: A count of services that reported, not of services that exist. Reporting is voluntary in practice and varies by year, so a fall in this count is a fall in coverage rather than in the number of water services.

Comparability: A French regulatory indicator (P104.3). It is not the inverse of the English leakage measures and must not be converted into them: England and Wales report leakage as a volume per day and per property, computed from a night-flow water balance under different rules. Comparing a rendement with a litres-per-property-per-day figure compares two different calculations.

Not to be confused with: leakage-volume, leakage-per-property

Comparability: Counts events per thousand subscribers. The English and Welsh measure of the same concept reports the average duration of interruption per property instead, so the two rank utilities differently and cannot be compared or converted.

Not to be confused with: supply-interruptions

Comparability: A high figure can mean high consumption, high leakage or a large non-household base, and the three cannot be told apart from this number alone. Read it alongside per-capita consumption and leakage rather than as a demand measure.

Not to be confused with: household-consumption-per-capita

Comparability: Metering share is a policy and regulatory outcome as much as an operational one — companies in water-stressed areas have been permitted compulsory metering that others have not — so it ranks circumstances rather than performance.

Comparability: Reported excluding voids — properties connected but unoccupied — so it is not comparable with a figure that includes them. The exclusion is stated because including voids depresses the percentage by several points.

Comparability: Depth is not water content. A metre of fresh powder and 200 mm of old compacted snow can hold the same water, and only the second is close to melting. Snow-water equivalent is the metric that answers "how much water is stored"; this one answers "how deep is it". Depth is also strongly affected by drifting, so a single stake is a poor estimate of the surrounding area.

Not to be confused with: snow-water-equivalent, snowfall-annual

Comparability: This is the operationally meaningful snow metric for water supply, because it states the volume of water held in storage rather than the thickness of the snow.

Comparability: One value per water year, running October to September, named for the year it ends. It is the peak of a monthly series, so it is the fullest month rather than the fullest day: a pack that peaked sharply mid-month reads lower here than a snow course visited at the right moment would record.

Comparability: Counted from the start of the water year rather than as a calendar month, so a peak in December and a peak in March are 3 and 6 and can be compared arithmetically without crossing a year boundary.

Comparability: A tenth of that year’s own peak rather than a fixed depth, so a thin year and a deep year are both measured against their own snowpack. It is a timing, not a quantity: two years can melt out in the same month having held very different amounts of water.

Comparability: A French regulatory indicator (P106.3), normalised by network length. The English and Welsh leakage-per-property measure is normalised by connections instead, so the two rank networks differently and cannot be placed on one scale.

Not to be confused with: leakage-per-property, leakage-volume

Comparability: Normalising by population rather than by property shifts what the figure reflects: a company serving many small households will look worse per property and better per person than one serving few large ones. It is a different ranking of the same losses, not a better one.

Not to be confused with: household-consumption-per-capita

Comparability: A percentage of input falls when demand rises even if the losses themselves are unchanged, so it moves with consumption as well as with network condition. Compare a company against its own record before comparing it against another.

Not to be confused with: non-revenue-water

Comparability: Normalising by property makes utilities comparable in a way absolute leakage does not, but still reflects network density, pipe age, material, pressure and ground conditions rather than management quality alone.

Comparability: Absolute leakage scales with the size of the company, so it must not be used to rank utilities of different sizes. Use litres per property per day or leakage per kilometre of main for comparison.

Not to be confused with: non-revenue-water

Comparability: A daily total belongs to the publisher’s observation day, which differs between networks and has sometimes changed within one station’s record. Comparing single days across countries is therefore unreliable even where the gauges are equally good. Totals are also under-caught in wind and snow, most severely at exposed and upland sites.

Not to be confused with: precipitation-monthly, precipitation-annual

Comparability: Monthly totals are dominated by seasonality. Compare a month against the same month in other years, or against the long-term average for that month, never against an adjacent month.

Comparability: Meteorological seasons are fixed three-month blocks — winter is December to February, and its December belongs to the previous calendar year. Comparing a season across years is valid; comparing adjacent seasons is not.

Comparability: Annual totals for different countries are not comparable as a measure of water availability. A large arid country and a small wet one can have similar national average rainfall while having entirely different renewable water resources per person. Gauge networks also differ in density, and mountain precipitation is systematically under-measured.

Comparability: Storage is strongly seasonal and is managed, not merely observed: an operator drawing a reservoir down for flood control produces the same reading as a drought would. A month should be compared with the same month in other years, and a fall in storage should not be read as a supply signal without knowing the operating rule.

Comparability: Each reservoir’s level is measured against the datum its operator uses, and those datums are not a common surface: some are national levelling datums, and some are local references established at the dam itself. Two reservoirs reading the same elevation are not at the same height, and the difference between two datums can be larger than the drawdown being described. Read a level against that reservoir’s own record, never against another reservoir’s, and never subtract one reservoir’s level from another’s.

Not to be confused with: reservoir-storage-volume, reservoir-storage-percent

Comparability: A release is an operating decision, not a catchment response, and must not be read as the natural flow of the river below the dam. Where a publisher reports releases through separate outlets, a total is not the sum of the ones Hydrionis happens to hold.

Comparability: Inflow is commonly derived from the storage change rather than gauged, so it inherits every error in the storage record and can be negative on days when the balance does not close. Compare a month against the same month in other years at the same reservoir.

Comparability: Total renewable resources include water flowing in from other countries, so a downstream country’s figure may depend on upstream decisions it does not control. Internal renewable resources exclude those inflows and are the better measure of self-sufficiency.

Comparability: A per-capita average conceals seasonal and regional distribution entirely. A country can be water-rich on this measure and still have cities that run short.

Comparability: Global stock estimates carry substantial uncertainty and are not measurements. Groundwater in particular cannot be measured directly at planetary scale, and published compilations differ by tens of percent.

Not to be confused with: renewable-water-resources

Comparability: Levels are specific to a borehole and its aquifer, and must be read against that borehole’s own historic range. Absolute elevations from different boreholes are not comparable.

Not to be confused with: groundwater-depth-below-surface

Comparability: Depth is measured from the land surface at the well, so it is specific to that well and must be read against that well’s own historic range. It must never be read against, or combined with, a groundwater level: level is an elevation above a levelling datum and rises as the water table rises, while depth is measured downwards from the ground and rises as the water table falls. The two move in opposite directions, and subtracting or comparing them directly inverts the result.

Not to be confused with: groundwater-level

Comparability: Modelled from household surveys rather than counted. The basic service definition is considerably weaker than safely managed, so high basic coverage is compatible with much lower safely managed coverage.

Comparability: This is the strictest rung of the JMP service ladder and requires all three criteria simultaneously. It is much more demanding than "improved source", and countries can score far lower on it than headline access statistics suggest.

Comparability: An all-in price including standing charges, taxes and the sanitation component, which is a different quantity from the volumetric-rate-only metric `water-price-per-m3`: the two differ by the whole of the fixed charge and must never be plotted on one axis. A unit price at a fixed reference consumption, not an average bill. It includes standing charges apportioned over the reference volume, so a household consuming less than the reference pays more per cubic metre than this figure. Not comparable with bills stated in another currency without an exchange-rate assumption Hydrionis does not make.

Not to be confused with: average-household-bill, water-price-per-m3

Comparability: Bills are not comparable across countries without stating the currency, the year, whether wastewater is included, whether the household is metered, and what average consumption is assumed. Some countries fund water partly through general taxation, which does not appear in the bill at all.

Comparability: Reporting years differ between countries, often by a decade. Withdrawal figures also vary in whether they include desalinated water and reused wastewater.

Not to be confused with: water-consumption-total

Comparability: A national figure averages over basins that can be individually far more stressed. Values above 100% mean withdrawal exceeds renewable supply, which is only possible by depleting groundwater, desalinating, or importing water.

Comparability: Modelled estimates, not measured counts. The definition requires excreta to be safely disposed of in situ or treated off site, which is far stricter than having a toilet.

Comparability: Counts depend on how many overflows are monitored. Monitoring coverage in England rose from a small minority of overflows to near-complete over roughly seven years, so an increase in recorded spills over that period is partly an increase in measurement.

Comparability: A five-year rolling average by construction (French indicator P107.2), so it responds slowly and a single year of heavy investment will not move it far. Renewal rates are not comparable between networks of very different age and material.