How our climate analogues are calculated — and what they are not

Garden Forward does not hand-pick comparison cities. For your location we reduce the ECMWF ERA5 reanalysis — the most recent fifteen complete years, 2011 to 2025 — to nine measured numbers: average summer maximum, winter minimum, annual mean temperature, annual rainfall, wettest and driest quarter, days above 30°C, frost nights and growing-season length.

We deliberately use a recent window rather than the standard 1991–2020 normal. A thirty-year normal that begins in 1991 already understates the summers most gardeners now experience, and an analogue built on it points to the climate of the recent past. Every reference city is reduced to exactly the same nine numbers over the same years, using the same code, so matching is a weighted nearest-neighbour search in a shared space: reproducible, and checkable. Cities within 250 km are skipped, because a place is not its own analogue.

For the 2035 and 2050 horizons we shift your measured baseline by the average change projected by two downscaled CMIP6 models under the SSP2-4.5 mid-range pathway, then match again. Using model change rather than model absolutes cancels out most of each model's bias.

The limits matter. Day length at your latitude does not move with temperature, so a garden matched to a lower-latitude city will never receive that city's light pattern. Soil, pests, diseases, local ecology and the frequency of extreme events all differ. Rainfall totals can match while the storms that deliver them do not.

Use an analogue for technique first and plant selection second. The way people in a hotter, drier place mulch, shade, irrigate and time their planting transfers well. Their plant lists may not — and any species that is invasive or legally restricted where you garden is excluded from our recommendations regardless of how well it would cope.