Abstract:
The invasion by Striga in most cereal crop fields in Africa has posed a significant threat to food security and hascaused substantial socioeconomic losses. Hyperspectral remote sensing is an effective means to discriminateplant species, providing possibilities to track such weed invasions and improve precision agriculture. However,essential baseline information using remotely sensed data is missing, specifically for the Striga weed in Africa. Inthis study, we investigated the spectral uniqueness of Striga compared to other co-occurring maize crops andweeds. We used thein-situFieldSpec®Handheld 2™ analytical spectral device (ASD), hyperspectral data and theirrespective narrow-band indices in the visible and near infrared (VNIR) region of the electromagnetic spectrum(EMS) and four machine learning discriminant algorithms (i.e. random forest: RF, linear discriminant analysis:LDA, gradient boosting: GB and support vector machines: SVM) to discriminate among different levels of Striga(Striga hermonthica) infestations in maize fields in western Kenya. We also tested the utility of Sentinel-2 wa-veband configurations to map and discriminate Striga infestation in heterogenous cereal crop fields. Thein-situhyperspectral reflectance data were resampled to the spectral waveband configurations of Sentinel-2 usingpublished spectral response functions. We sampled and detected seven Striga infestation classes based on threeflowering Striga classes (low, moderate and high) against two background endmembers (soil and a mixture ofmaize and other co-occurring weeds). A guided regularized random forest (GRRF) algorithm was used to selectthe most relevant hyperspectral wavebands and vegetation indices (VIs) as well as for the resampled Sentinel-2multispectral wavebands for Striga infestation discrimination. The performance of the four discriminant algo-rithms was compared using classification accuracy assessment metrics. We were able to positively discriminateStriga from the two background endmembers i.e. soil and co-occurring vegetation (maize and co-occurringweeds) based on the few GRRF selected hyperspectral vegetation indices and the GRRF selected resampledSentinel-2 multispectral bands. RF outperformed all the other discriminant methods and produced the highestoverall accuracy of 91% and 85%, using the hyperspectral and resampled Sentinel-2 multispectral wavebands,respectively, across the four different discriminant models tested in this study. The class with the highest de-tection accuracy across all the four discriminant algorithms, was the “exclusively maize and other co-occurringweeds” (>70%). The GRRF reduced the dimensionality of the hyperspectral data and selected only 9 mostrelevant wavebands out of 750 wavebands, 6 VIs out of 15 and 6 out of 10 resampled Sentinel-2 multispectralwavebands for discriminating among the Striga and co-occurring classes. Resampled Sentinel-2 multispectralwavebands 3 (green) and 4 (red) were the most crucial for Striga detection. The use of the most relevant hy-perspectral features (i.e. wavebands and VIs) significantly (p≤ 0.05) increased the overall classification accu-racy and Kappa scores (±5% and ±0.2, respectively) in all the machine learning discriminant models. Ourresults show the potential of hyperspectral, resampled Sentinel-2 multispectral datasets and machine learningdiscriminant algorithms as a tool to accurately discern Striga in heterogenous maize agro-ecological systems