How fast different aquifers answer the rain

The question

Is there one answer to how long groundwater takes to respond to rainfall, or does it depend entirely on where you are standing?

The single-site answer invites being generalised, and generalising it is how a drought recovery gets announced for a region where half the aquifers have not recovered. Running the identical calculation across every site that qualifies turns one number into a distribution, and a distribution is the honest form of this answer.

Sites the test could be run at

0–12 months

Range of strongest-association lag

2 months

Median across qualifying sites

337

Sites excluded, with reasons below

What the spread looks like

Strongest association at	Sites	Share of qualifying sites
The same month	6	8%
1 month later	23	32%
2 months later	21	29%
3 months later	10	14%
6 months later	10	14%
12 months later	2	3%

Each row is one borehole that holds enough overlapping record to test. The lag column is the delay at which the association between rainfall and level is strongest, in months.
The spread across the column is the finding. A single national figure would sit inside it and describe few of the sites.
Sites are excluded rather than estimated where the record is too short or too gappy to test, and the count excluded is stated with the table.

Every qualifying site

Ordered by how strongly the test worked, not by name. A weak coefficient is not a failed site — it is a site where England-wide rainfall explains less of the level, which is itself a finding about that place.

Site	Overlap	Lag	Correlation	Pairs	Held
Old Stoke Farm 1966–2026	60 yrs	1 mo	0.51	643	89%
Strelley 1969–2026	57 yrs	1 mo	0.50	664	96%
Tufton Warren Farm 1966–2026	60 yrs	1 mo	0.49	675	93%
Combe 1969–2026	57 yrs	1 mo	0.48	621	91%
Westwell 1969–2026	57 yrs	1 mo	0.48	584	86%
Harfield Easton 1966–2026	60 yrs	1 mo	0.47	618	85%
Chidden Farm 1958–2026	68 yrs	1 mo	0.47	784	96%
Chilgrove House 1836–2026	190 yrs	1 mo	0.44	2,191	96%
West Dean 3 1976–2026	50 yrs	1 mo	0.42	554	93%
Oddo House 1969–2026	57 yrs	1 mo	0.42	625	91%
Baileys Down 1964–2026	62 yrs	2 mo	0.41	723	97%
Compton 1893–2026	133 yrs	2 mo	0.41	1,445	90%
Ashley Village 1959–2026	67 yrs	2 mo	0.40	698	87%
Winslade Farm 1966–2026	60 yrs	2 mo	0.39	625	86%
Bull I th thorne 1969–2020	51 yrs	1 mo	0.39	525	86%
The Old Farm 1960–2026	66 yrs	2 mo	0.38	680	86%
Down Cottages 1950–2026	76 yrs	2 mo	0.37	847	92%
Furlyns 1966–2026	60 yrs	1 mo	0.37	692	96%
Woodside Vernham Dean 1959–2026	67 yrs	2 mo	0.37	698	86%
Ford Farm Cottages 1960–2026	66 yrs	2 mo	0.36	741	93%
Sandy Lane 1966–2026	60 yrs	1 mo	0.36	620	87%
Snoddington Down 1964–2026	62 yrs	2 mo	0.36	648	87%
Ashe Warren Cottages 1966–2026	60 yrs	3 mo	0.36	624	86%
Hucklow 1969–2026	57 yrs	same month	0.35	655	95%
White Lane Oakley 1966–2026	60 yrs	2 mo	0.35	645	89%
Podmore 1966–2024	58 yrs	2 mo	0.35	601	87%
Dalton Estates 1889–2026	137 yrs	2 mo	0.34	1,593	96%
Jollys Farm Lower 1962–2026	64 yrs	3 mo	0.33	680	88%
Crondall 1950–2026	76 yrs	1 mo	0.33	778	85%
George Hill 1966–2026	60 yrs	2 mo	0.33	650	90%
Old School House 1969–2026	57 yrs	2 mo	0.32	589	85%
Polhampton Lodge Cottage 1966–2026	60 yrs	2 mo	0.32	665	92%
The Spinney 1952–2026	74 yrs	2 mo	0.32	751	85%
Stubwood Well 1966–2024	58 yrs	1 mo	0.31	655	95%
Knights Enham 1966–2026	60 yrs	3 mo	0.31	638	88%
Roke Farm Odiham 1950–2026	76 yrs	3 mo	0.31	799	87%
Springfield Road 1969–2026	57 yrs	3 mo	0.30	624	92%
Arkesden Village 1964–2026	62 yrs	1 mo	0.29	643	87%
Hognore Farmhouse 1968–2026	58 yrs	2 mo	0.29	625	90%
Putteridge Bury 1966–2026	60 yrs	3 mo	0.29	651	90%
Danes Hill 1969–2026	57 yrs	1 mo	0.27	621	91%
Hurstbourne Tarrant 1964–2026	62 yrs	2 mo	0.27	686	91%
Cresent Cottages 1966–2026	60 yrs	6 mo	0.25	613	86%
Gallowstree Common 1969–2026	57 yrs	6 mo	0.25	636	94%
Cave Gate 1966–2026	60 yrs	3 mo	0.24	615	86%
Heighington Station 1969–2026	57 yrs	2 mo	0.24	595	87%
Berden Hall 1964–2026	62 yrs	6 mo	0.23	657	89%
Erlands House 1950–2026	76 yrs	same month	0.22	812	88%
Hixham Hall 1964–2022	58 yrs	6 mo	0.21	602	87%
Crane Bridge 1958–2026	68 yrs	1 mo	0.21	745	92%
Wested Farm 1957–2026	69 yrs	6 mo	0.21	729	89%
Hop Pole Inn 1968–2026	58 yrs	3 mo	0.20	603	87%
Crockenhill 1955–2026	71 yrs	6 mo	0.20	821	96%
Therfield Rectory 1883–2025	142 yrs	6 mo	0.20	1,563	92%
Penniment Farm 1966–2026	60 yrs	1 mo	0.18	672	94%
Highlands Farm 1968–2026	58 yrs	3 mo	0.17	608	87%
Fairoak 1966–2024	58 yrs	2 mo	0.16	651	94%
Swan Carr 1969–2026	57 yrs	2 mo	0.14	615	89%
Bigmoor 1948–2026	78 yrs	1 mo	0.14	899	96%
Cherry Tree 1968–2026	58 yrs	same month	0.12	655	94%
John Players 1967–2026	59 yrs	6 mo	0.12	632	90%
Morris Dancers 1969–2026	57 yrs	same month	-0.11	647	96%
Shipstones Brewery 1963–2026	63 yrs	6 mo	0.11	683	91%
Ivy House Haugham 1968–2026	58 yrs	12 mo	-0.10	606	89%
Meridian 1967–2026	59 yrs	6 mo	0.10	637	91%
Oxton 1969–2026	57 yrs	same month	-0.10	663	97%
Duchess Plantation 1969–2026	57 yrs	1 mo	-0.10	645	95%
Peafield 1969–2026	57 yrs	1 mo	-0.09	648	95%
Kighill 1969–2022	53 yrs	1 mo	-0.09	619	98%
Watch Hill 1969–2026	57 yrs	same month	-0.07	639	94%
Gamston 1969–2026	57 yrs	12 mo	0.06	648	97%
Owletts 1968–2026	58 yrs	3 mo	0.06	615	89%

What was excluded, and why

409 English boreholes were considered

280 hold more than 15% of the common period as gaps. A lag estimated across absences is estimated from the periods that happen to survive.
56 overlap England’s rainfall record for fewer than 30 years. Below that the result describes the particular run of wet and dry years sampled rather than the aquifer.
0 left fewer than 30 usable pairs at every lag tested, which is too few for a coefficient to mean anything.
1 failed the pairing eligibility rules — geography, resolution or time overlap. Those rules are set out in the comparisons Hydrionis will not draw.

Nothing excluded was patched, gap-filled or estimated. A survey that reported only what qualified would have hidden its own selection, which is why the counts are here rather than in a footnote.

What would make this wrong

Where the figures come from

Every number on this page was computed from the observations Hydrionis holds at the moment the page was produced, using methods declared in the methodology register. None of them is typed into this page, so none of them can disagree with the data behind it. The underlying series are in the dataset catalogue and reachable through the API.