The Spanish American record as a climate archive
Three centuries of Spanish record-keeping in the Americas are now read as data by climate science. Over 300,000 early instrumental weather observations have been recovered from twenty countries, beginning with Lima in 1754, and 656 Spanish Caribbean shipwrecks, normalised against fleet departures counted in the Archivo General de Indias, have been used to reconstruct hurricane activity back to 1495. The same literature states what the archive cannot do.
Why this wins its question: Colonial record-keeping is usually discussed as an instrument of control and taxation, and the scientific afterlife of those same papers is told, when it is told at all, one dataset at a time inside technical journals that never mention each other. This object puts the instrumental series, the fleet registers and the wreck reports in one place with their actual figures — 300,000 records over twenty countries, Lima from 1754, 656 wrecks against 643 registers and 3,881 departures, a 75% drop in cyclone activity in the Maunder Minimum — and gives equal space to the limits each study states about its own archive, including the 83% concentration in four countries and the mass of paper that has never been catalogued.
Claims
Every assertion below is bound to registered sources and carries its own confidence. Weight them; do not treat the page as uniformly authoritative.
The instrumental record exists and has been recovered. The Scientific Data paper by Domínguez-Castro and colleagues assembles over 300,000 early instrumental meteorological records from twenty countries of Latin America and the Caribbean, spanning from Lima in 1754 to Havana in 1905, with temperature, atmospheric pressure and precipitation as the principal variables and humidity, wind direction and sky conditions as secondary ones.
Lima is the anchor of the series, and it is a Spanish colonial one: the paper reports that "El conocimiento de los Tiempos" presents instrumental measurements from Lima spanning 1754-1856, and names among the early observers José Hipólito Unanue at Lima and Francisco José de Caldas at Bogotá, whose 1808 and 1809 observations "were used to date exactly the unknown eruption of 1809". Colonial-era readings taken for civic and medical purposes are now doing volcanological work.
The records are patchy, and the paper says so in its own words: these early observations were "very sparse, discontinuous in time, and often made without homogeneous observation procedures", made with "different instrumentation and scales", and standardisation arrived only after the international meteorological conferences of 1853 and 1873. Roughly 83% of the total data come from just four countries — Chile, Cuba, Brazil and Ecuador — leaving the rest of the region thinly covered.
The break in the series is political, and the paper attributes it to the wars rather than to the colonial administration: "In Spanish America, the prolonged political struggle of the colonies to achieve independence and the reorganization of their administrative services did not favour the continuity of observations." What ended the record was the end of the regime that had been keeping it, and the disorder that followed.
Most of the surviving material has never been made usable. The paper records that data "are usually stored in paper format, preserved in different archives and under different conditions, and most have not been catalogued or digitized" — so the recovered 300,000 records are a sample of a larger, still-unread corpus. Quality control during digitisation flagged 488 suspicious values, 0.16% of the dataset, of which 59 were corrected and 429 rejected after checking against the originals.
The bureaucracy of the Indies also produced a proxy record nobody designed. In the PNAS study by Trouet, Harley and Domínguez-Delmás, "a total of 656 Spanish shipwreck events (1495-1825 CE) that occurred in the Caribbean realm" were compiled from a collection of roughly 4,000 entries, and the shipping volume needed to normalise them was obtained by tallying "each ship departure among the 643 registers in the Registros de Ida" of the Archivo General de Indias, giving 3,881 active transatlantic Spanish vessels for 1590-1780.
The result is a climate finding that no natural archive had supplied: the authors conclude that there was "a 75% reduction in decadal-scale Caribbean TC activity during" the Maunder Minimum of 1645-1715 — that is, tropical cyclone activity tracking a period of reduced solar radiative forcing, measured through the losses of the Spanish carrera de Indias.
The shipwreck archive's weaknesses are stated by the authors and are the ordinary weaknesses of any administrative record used for something it was not kept for: "This data set is not complete, as some shipwrecks were never documented and some documents likely lost due to deterioration with age, war, or fire". They also flag the trend introduced by "increasing numbers of ships" and restrict the analysis to events in the hurricane season (July-November) or of unknown seasonality, which is a documentation bias handled rather than removed.
Against the global picture, the Spanish American documentary corpus remains under-exploited rather than exhausted. DOCU-CLIM, published by Burgdorf, Brönnimann and colleagues as "the first global multi-variable collection of documentary climate records", holds 621 time series drawn from diaries, chronicles and logbooks; "most of the document-based climate records are from Europe", with "only a few documentary records from South America", though "some records from South America begin as early as the 16th century". The material exists earlier in the Americas than the compilations reflect.
Nothing in this record was created for science, and that is the honest frame for it. The weather registers were kept for medical, agricultural and civic reasons, the sailing registers to control and tax a monopoly trade, and the loss reports to account for property. Their scientific value is retrospective, produced by modern digitisation and statistical treatment, and every one of the three studies cited here reports the biases of its source before reporting its result.