Accurate crop identification and spatial distribution mapping are important for crop production estimation and famine early warning, especially for food-deficit African agricultural countries. By evaluating existing preprocessing methods for classification using satellite image time series (SITS) in Kenya, this study aimed to provide a low-cost method for cultivated land monitoring in sub-Saharan Africa that lacks financial support. SITS were composed of a set of MODIS Vegetation Indices (MOD13Q1) in 2018, and the classification method included the Support Vector Machine (SVM) and Random Forest (RF) classifier. Eight datasets obtained at three levels of preprocessing from MOD13Q1 were used in the classification: (1) raw SITS of vegetation indices (R-NDVI, R-EVI, and R-NDVI + R-EVI); (2) smoothed SITS of vegetation indices (S-NDVI); and (3) vegetation phenological data (P-NDVI, P-EVI, R-NDVI + P-NDVI, and P-NDVI-1). Both SVM and RF classification results showed that the “R-NDVI + R-EVI” dataset achieved the highest performance, while the three pure phenological datasets produced the lowest accuracy. Correlation analysis between variable importance and rainfall time series demonstrated that the vegetation index SITS during rainfall periods showed higher importance in RF classifiers, thus revealing the potential of saving computational costs. Considering the preprocessing cost of SITS and its negative impact on the classification accuracy, we recommend overlaying the original NDVI with the original EVI time series to map the crop distribution in Kenya.

Lunar south polar regions have attracted considerable scientific interest due to their potential for preservation of water ice and unique mineralogical compositions. As a key scientific payload for surface composition exploration missions, hyperspectral imager faces significant challenges in the lunar polar regions. The primary issue is the low illumination conditions in these areas, where terrain-induced shadows drastically reduce the signal-to-noise (SNR) of hyperspectral images (HSIs), resulting in limited availability of reliable spectral available for polar region analysis. Previous studies have largely bypassed low-SNR spectra or filtered them out, as there has been no effective method to recover the spectral information under these harsh conditions. To tackle this problem, an effective method based on CycleGAN network is proposed to compensate hyperspectral data obtained by Moon Mineralogy Mapper (M3) under low illumination conditions in lunar south polar regions. The network was trained by constructing paired datasets of low and high SNR M3 spectra from the lunar South Pole. The efficacy of the proposed method is validated using real high SNR M3 spectral observations, with the performance of the compensated results comprehensively assessed across three dimensions: structural indicators, spectral indices, and spatial consistency analysis. The strong correlation between the M3 spectral compensation results with SELENE (Kaguya) Multiband Imager (MI) data, as well as other sensors' inversion of plagioclase abundance around the Shackleton Crater, underscores the network's potential for mineral exploration. To our knowledge, this study represents one of the first efforts to compensate illumination-limited spectra in lunar HSI. It provides an efficient method for enhancing the SNR of M3 spectra in the lunar polar region, offering a reliable tool and novel insights for future mineralogical and water ice studies.

Hyperspectral images (HSIs) from the lunar polar regions are frequently compromised by low signal-to-noise ratio (SNR) under adverse illumination, limiting their utility for scientific analysis. Existing spectral-only compensation approaches operate without spatial context, leading to speckle-like artifacts that degrade spatial consistency and constrain subsequent applications. To address this limitation, we propose SSF-3DSC, a spatial–spectral fusion 3D signal-compensation framework tailored for lunar HSIs to simultaneously restore spectral fidelity and spatial consistency under extreme low-illumination conditions. To the best of our knowledge, this represents the first deep learning framework specifically engineered for joint spatial–spectral restoration in the photon-starved regime. SSF-3DSC integrates three specialized components: a spectral compensation module (SCM) for restoring spectral fidelity, a multi-scale spatial attention (MSA) module for capturing hierarchical spatial patterns, and a cascaded 3D residual convolutional module (C3D-RCM) for refining spatial–spectral representations. Trained on paired low- and high-SNR Moon Mineralogy Mapper (M3) data cubes from the lunar south polar region, SSF-3DSC employs synergistic spatial–spectral fusion to achieve high-fidelity reconstruction, significantly outperforming a spectral-only lunar baseline (Paired-CycleGAN). Regional-scale experiments demonstrate its ability to recover both spatially coherent geological structures and spectrally reliable mineral abundance maps. By establishing a new benchmark for lunar HSI restoration under low-illumination conditions, this work enhances the scientific utility of low-signal M3 data and enables robust quantitative investigations into the Moon’s challenging polar regions.