Fine-resolution climate grids for species studies in data-poor regions

African climate science is still underdeveloped; the related infrastructure is limited and climate observation network is sparse. The lack of observations is usually compensated with coarse macro-climate models (1 km2), of which only few concentrates on tropical Africa and on the spatial variability of rains. The need for fine-resolution data is growing, as more accurate information (5-100m) is utilized in different sciences. There has been a surge in micro-climate studies utilizing small dataloggers to observe climate variability, as loggers can be placed to different environments regardless of the complexity of a terrain. Accurate, fine-scale climate models are needed in ecological research, where species distributions are related to environmental gradients. Most of the studies describing species occurrences rely on macro-scale climate models that have been averaged over years, although the factors restricting species occurrences in an environment are developed even under one growth season, and are dependent on climate extremes. Accurate information of species and its occurrence environment is even more needed as climate is changing; species habitats are threatened through the warming of climate and its resulted effects. Fine-resolution data of micro-climate is needed especially in the studying of epiphytes living in tropical montane cloud forests, as epiphytes are dependent on the surrounding environments' specific atmospheric composition, namely relative humidity. The knowledge of epiphytic species distributions is poorly studied, although epiphytes are known to react fast to quick changes in environmental conditions. In this study small dataloggers were placed around the Taita Hills in a mountainous region of Kenya to record spatial and temporal micro-climate variability. Annual, monthly and extreme temperature and humidity parameters were studied through statistical spatial modelling, with generalized additive models (GAMs). The significance of each thriving factor of climate parameters were studied separately with bootstrapping. Of thousands of models created, temperature was best explained by elevation, and humidities by land cover (e.g. distance to forests), a proxy of canopy cover. Modelling extremes requires more research, as extreme models proved to be most difficult to model. More weight should be put on accurate environment gradients (e.g. canopy cover) when modelling spatial climate variability in a fine-scale. Created fine-resolution grids were tested in modelling the distributions of epiphytic species through Boosted Regression Trees (BRT). Models created gave an insight to the drivers of epiphytic species distributions and to the limitations of their occurrences in a limited environment. Different explanatory variable groups were tested (topographical, in situ, climate groups), of which climate explained most of the variations in epiphytic species occurrences when considered with AUC-values. Some epiphytes seemed to be specialized to certain habitat trees, highlighting the value of not only indigenous trees but exotic ones. Epiphytes were restricted to occur under certain temperatures and humidities ranges that is an indication of the vulnerability of epiphytes to changing climate.