- Guidelines and best practices for the scientific use of global iNaturalist data. Citizen scientists have generated an enormous photographic record of biodiversity, but turning those observations into reliable scientific data requires some care. Here’s some practical guidance on accessing and using iNaturalist records and images, addressing taxonomic uncertainty, geoprivacy, observer bias, attribution and reproducibility.
- Mapping the spatial extent of agroforestry lands across Africa by characterizing tree spatial arrangements. Moving from individual biodiversity observations to the landscape, this study uses remote sensing to identify agroforestry across Africa by looking at the spatial arrangement of trees.
- AI tool for mapping smallholder crops proves itself in Senegal. Once we know where trees and agricultural landscapes are, the next question is what is actually being grown. Here researchers use Tessera, an open-source AI model trained on satellite imagery, to identify different crops in smallholder systems, achieving 84% accuracy while requiring considerably less computing power and labelled training data than other approaches.
- National Mapping of Monoculture and Intercropped Maize via Phenology-Assisted Sampling: Application in Kenya. Mapping crops is one thing; distinguishing how they are grown is another. This study distinguished intercropped from monocropped maize in smallholder farming systems, using satellite time series and phenological information to generate training samples for machine-learning classification.
- Smartphone-based proximal sensing as an accessible platform for improved physiological phenotyping. From mapping whole agricultural landscapes we move to the research station, and from satellites to smartphones. Using a smartphone to capture repeated images of plants in cowpea plots generated high-resolution time series, from which could be extracted information on vegetative growth, flowering and pod production.
- Determination of dry bean physiological maturity from UAS RGB time series using a stacking ensemble of color-derived spectral features. The cowpea study shows what can be extracted from repeated close-range images of individual plants. Here the emphasis shifts to using drone-mounted cameras to estimate a specific breeding trait: physiological maturity in dry beans. Combining colour-based indicators with time-series features and machine learning achieved a prediction error of only 2.68 days.
- Multi-scale spatial-temporal remote sensing fusion for phenology identification in rice germplasm resources. If maturity can be estimated from drone imagery in dry beans, can similar approaches monitor hundreds of different rice genotypes in a breeding collection? Apparently yes, by combining frequent medium-resolution images with less frequent high-resolution UAV imagery.
- Digitalization of the genotypic and phenotypic information retained within the FEM grapevine germplasm bank. We end where phenotyping meets conservation. The Fondazione Edmund Mach grapevine collection brings together more than 3,000 accessions, with genetic identity assessed using microsatellite markers and phenotypic characterisation covering vine development, ampelographic traits and responses to biotic stresses. The paper describes the digitisation of this information in FEMVitisDB, a FAIR-compliant database with links to international data infrastructures. And not a drone in sight. Yet