Application of multifractal theory hybridized with ICEEMDAN for investigating dry-wet conditions over Guinea Republic (1991-2020)

Understanding the intrinsic characteristics and nonlinear dynamics of dry-wet conditions in rain-fed agricultural regions like Guinea is crucial for agricultural planning and water resources management. Standardized Precipitation Index (SPI) for 1, 3, 6, 9 and 12- month timescales, derived from eight synoptic stations in Guinea between 1991 and 2020, were used to assess dry-wet conditions. By applying Multifractal Detrended Fluctuation Analysis (MFDFA) hybridized with the Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) algorithm on multiscale SPIs (SPI-1, SPI-3, SPI-6, SPI-9 and SPI-12), the multifractality, the long-term memory and the complexity of temporal fluctuations of dry-wet conditions were investigated in Guinea. The results indicate that (1) anti-persistence, anti-persistence combined with non-stationary, persistence, and randomness is present in the SPI series. In particular, the SPI-3 and SPI-6 series are persistent at all the studied stations, while the SPI-9 and SPI-12 series are persistent or anti-persistent depending on the station; (2) all SPI series exhibit multifractal dynamics, with varying strengths influenced by timescale and geographical location. Multifractality in SPI-3 is primarily due to long-term temporal correlations, while SPI-6, SPI-9 and SPI-12 are influenced by both long-term correlations and the broadness of the probability distribution function, with long-term correlation being the dominant factor. Consequently, SPI-6, SPI-9, and SPI-12 are potentially more sensitive to climate dynamics than SPI-3; (3) predictive models based on persistence and multifractal theories are effective for SPI-3, and SPI-6, while for SPI-9 and SPI-12, models based on either persistence or averaging previous values are recommended, depending on the station. Key words: Standardized precipitation index (SPI), hurst exponent, multifractal parameters, improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN), dry-wet, Guinea.