A Global Assessment of Bimodality Underlying Normalized Drought Indices (SPI and SPEI)

<title>Abstract</title> <p>Normalized drought indices, such as the Standardized Precipitation Index (SPI) and Standardized Precipitation Evapotranspiration Index (SPEI), rely on fitting statistical distributions to meteorological measures. This implicitly assumes a unimodal distribution; an assumption that has not yet been comprehensively evaluated. This study presents a global assessment of bimodality using two global climate datasets (CRU TS 4.06 and GSWP3 v0.5b), evaluated for 3-, 6-, and 12-month accumulation periods using three complementary bimodality tests. Results show that bimodality is rare, occurring primarily for short accumulation periods in arid and semi-arid regions with strong wet-dry seasonal contrasts. Typical bimodal distributions involved a primary mode clustered near a lower bound (zero for the SPI), followed by a secondary peak reflecting occasional wet years. The frequency of bimodality decreased markedly with longer moving windows and was negligible for 12-month accumulation periods. Case studies further showed that many apparent bimodal distributions were associated with data artifacts rather than true climatic processes. Overall, the unimodal assumption underlying SPI and SPEI is valid for most locations and applications, with visual screening followed by formal bimodality testing recommended for short accumulation periods in strongly seasonal dry climates between 30°N and 30°S.</p>