Browsing by Subject "grass cover"

Savanna rangelands are vital for millions of people for their livelihoods, food, and income in addition to protecting biodiversity and ecosystem services. Savannas are grazed by both livestock and wildlife, and due to population growth, livestock production and overgrazing is increasing. In addition, more people are migrating to savannas and converting them to cropland. Cropland expansion and grazing have increased also in the lowlands of Taita Hills, Kenya. The declining grass cover leads to, e.g., lower resources for livestock and wildlife, putting the future of this unique biome and its populations at risk. The study area consists mainly of the LUMO Community Wildlife Sanctuary (LUMO) and Taita Hills Wildlife Sanctuary (THWS) of which the first is grazed by both livestock and wildlife while the latter primarily by wildlife. The region experiences bimodal precipitation with long rains in March–May and short rains in October–December. However, the region experienced low rainfall throughout most months of 2022 resulting in exceptionally low grass cover. While this was an unusual year, there is no guarantee that this would not become the new normal with climate change incxreasingly affecting the region. Satellite remote sensing is a cost-effective way of monitoring changes in savanna rangelands due to its high spatio-temporal nature. This study included the use of remotely sensed spectral information together with field photographs of 36 plots in two different months, January and May 2022, to create a model that predicts monthly grass cover in the study area in 2022. A color index, Excess Greenness, was used to model the green fractional vegetation cover of field photographs by utilizing a manual grass cover classification. These modelled photographs were then used to upscale the model to very-high-resolution satellite imagery of January and May. The resulting grass cover map of May was used to train the model against various spectral predictors to create a model for predicting grass cover using medium-resolution satellite imagery. The model was validated by the very-high-resolution grass cover map from January, and the performance was promising (R2 = 0.89). The model produced grass cover maps for each month of 2022. The maps present different spatial patterns between LUMO and THWS and temporal patterns between months. As expected, THWS has higher grass cover than LUMO as it is only grazed by wildlife. Throughout the year there is a clear declining trend in grass cover. The original hypothesis was that there is clear monthly variability in grass cover. However, the field campaign in May resulted in unprecedently similar results as in January, i.e., low grass cover, which is also evident in the resulting monthly maps. The results of the model are promising as they provide evidence of the capabilities of the used method. The results also demonstrate the usability of an ordinary digital camera as the basis for vegetation cover models. In addition, this method can be applied to other temporal scales, for example for a yearly time series analysis. The grass cover maps can be further analyzed together with information of grazing pressure and used to inform decision-makers and other entities of land use management.