Search

Historical land cover maps are of high importance for scientists and policy makers studying the dynamic character of land cover change in the Sudano-Sahel, including anthropogenic and climatological drivers. Despite its relevance, an accurate high resolution record of historical land cover maps is currently lacking over the Sudano-Sahel. In this study, 30 m resolution historically consistent land cover and cover fraction maps are provided over the Sudano-Sahel for the period 1986–2015. These land cover/cover fraction maps are achieved based on the Landsat archive preprocessed on Google Earth Engine and a random forest classification/regression model, while historical consistency is achieved using the hidden Markov model. Using these historical maps, a multitude of variability in the dynamic Sudano-Sahel region over the past 30 years is revealed. On the one hand, Sahel-wide cropland expansion and the re-greening of the Sahel is observed in the discrete land cover classification. On the other hand, subtle changes such as forest degradation are detected based on the cover fraction maps. Additionally, exploiting the 30 m spatial resolution, fine-scale changes, such as smallholder or subsistence farming, can be detected. The historical land cover/cover fraction maps presented in this study are made available via an open-access platform. [ABSTRACT FROM AUTHOR]

Currently most global land cover maps are produced with discrete classes, which express the dominant land cover class in each pixel, or a combination of several classes at a predetermined ratio. In contrast, land cover fraction mapping enables expressing the proportion of each pure class in each pixel, which increases precision and reduces legend complexity. To map land cover fractions, regression rather than classification algorithms are needed, and multiple approaches are available for this task. A major challenge for land cover fraction mapping models is data sparsity. Land cover fraction data is by its nature zero-inflated due to how common the 0% fraction is. As regression favours the mean, 0% and 100% fractions are difficult for regression models to predict accurately. We proposed a new solution by combining three models: a binary model determines whether a pixel is pure; if so, it is processed using a classification model; otherwise with a regression model. We compared multiple regression algorithms and implemented our proposed three-step model on the algorithm with the lowest RMSE. We further evaluated the spatial and per-class accuracy of the model and demonstrated a wall-to-wall prediction of seven land cover fractions over the globe. The models were trained on over 138,000 points and validated on a separate dataset of over 20,000 points, provided by the CGLS-LC100 project. Both datasets are global and aligned with the PROBA-V 100 m UTM grid. Results showed that the random forest regression model reached the lowest RMSE of 17.3%. Lowest MAE (7.9%) and highest overall accuracy (72% ± 2%) was achieved using random forest with our proposed three-model approach and median vote. This research proves that machine learning algorithms can be applied globally to map a wide variety of land cover fractions. Fraction mapping expresses land cover more precisely, and empowers users to create their own discrete maps using user-defined thresholds and rules, which enables customising the result for a diverse range of uses. The three-step approach is useful for addressing the zero-inflation issue and mapping 0% and 100% fractions more accurately, and thus has already been taken up in the operational production of global land cover fraction layers within the CGLS-LC100 project. Furthermore, this study contributes to the accuracy assessment of land cover fraction maps both thematically and spatially, and these methods could be taken up by future land cover fraction mapping efforts. • 7 land cover fractions derived using machine learning models and validated globally. • Multi-step approach proposed for dealing with zero inflation in fraction data. • Independent spatial accuracy assessment based on over 20,000 reference points. • Multi-step approach gives a best guess and improves 0% and 100% fraction estimation. • Multi-fraction land cover mapping shown as an alternative to discrete classification. [ABSTRACT FROM AUTHOR]