Spatio-temporal assessment of agro-ecological covariates modulating puddling intensity in paddy rice systems using ensemble remote sensing and non-parametric frameworks

Abstract

Puddling, the tillage of saturated soils before rice transplanting, is a water and energy-intensive field preparation in irrigated paddy systems which remains poorly quantified. We present a spatio-temporal analysis of covariates affecting puddling intensity across the Mwea Tebere Irrigation Scheme, Kenya between 2016 and 2024. The framework integrates satellite-derived biophysical indices from Sentinel-2; Normalised Difference Vegetation Index, Normalised Difference Water Index, Land Surface Temperature, Normalised Difference Red Edge, Vegetation Stress Index, Soil Moisture Index, Plant Nitrogen Accumulation index and Vegetation Soil Salinity Index; Sentinel-1 C-band surface roughness as a proxy for post-puddling soil microrelief, climatological indices; Precipitation Condition Index, Temperature Condition Index and Temperature Humidity Index. The statistical characterisation was based on 20,000 randomly distributed sample points, encompassing the scheme’s spatial heterogeneity. The declining NDRE (τ = −0.31, p = 0.024) indicated progressive deterioration of canopy nitrogen, linked to declining SMI for optimal nitrogen mineralisation, rising LST accelerating nitrogen volatilisation from flooded soils, and expansion onto inherently lower-fertility soils. The significant declining SMI trend (τ = −0.33, Sen’s slope = − 0.009 yr⁻¹, p = 0.018) is a key finding of this study. The dominant soil texture was Clay Loam (71.1%) which is highly conducive to hardpan formation and soil pH was optimum for rice most points sampled (mean = 6.26). Spearman correlation (ρ = 0.99) showed strong coupling between soil water retention and clay content. The framework offers a reproducible blueprint for monitoring relevant agro-ecological covariates of puddling.