Reconstruction of gridded summer Standardized Precipitation Evapotranspiration Index (SPEI) across China during the last millennium via assimilating an expanded proxy network

(1) Data Content: This dataset comprises spatiotemporally-complete reconstruction of the summer (June, July, August) Standardized Precipitation Evapotranspiration Index (SPEI-03) across China during the last millennium (1000-2000 CE). The reconstructed SPEI fields have a spatial resolution of approximately 250 km (≈1.875° lat-lon grid). SPEI-03 reflects water availability during crop growth stages and is a core indicator of agricultural drought. The dataset includes: 1) the mean, standard deviation (Sigma), and the 2.5th and 97.5th percentiles derived from the 5,000 SPEI reconstructions (50 realizations × 100 members) ; and 2) the ensemble mean (100 members) for each of the 50 Monte Carlo realizations, yielding 50 results. Each realization is an independent reconstruction based on a randomly drawn 100 member prior ensemble and a random 75% proxy subset (with the remaining 25% withheld for validation). Qualitative validation, cross-validation and independent verification demonstrated that, compared to SPEI derived solely from climate model simulations and previous reconstructions, this dataset exhibits higher reliability over China. (2) Data Production Method: We reconstructed this dataset using a paleoclimate data assimilation framework. The system assimilated 260 proxy records - far more than previous SPEI reconstructions for China - including 69 dry/wet grade series from Chinese historical documents. These documentary records, not used in earlier studies, help address the spatial data gap over eastern China in previous reconstructions. A time-averaged Ensemble Square Root Filter (EnSRF) algorithm was applied for assimilation. The static prior ensemble (ensemble size N = 100) was randomly sampled from the “Global SPEI over the Common Era” dataset (see dataset details at 10.5281/zenodo.18411292), which is based on monthly climate variables from MPI-ESM model simulations in CMIP6 past2k experiment. (3) Update Notes: v2 introduces the following updates and improvements relative to v1: 1) Proxy record screening: Proxy records were further screened to include only those showing a significant correlation (p < 0.1) with summer SPEI. This resulted in a refined dataset of 260 qualified proxy records. 2) Enhanced uncertainty characterization: In addition to providing the mean and standard deviation of the 5,000 ensemble reconstructions, the 2.5th and 97.5th percentiles of the 5,000 SPEI‑03 reconstructions (derived from 50 PDA realizations) are now included. 3) Metadata improvements: The metadata have been supplemented and refined. Please contact the producer Miao Fang (E-mail: [e-mail]) for more information about the details of the reconstructions.