Operational large-area land-cover mapping: An Ethiopia case study

摘要: Knowledge of land cover and use nationally is a prerequisite many studies on drivers change, impacts climate, carbon storage other ecosystem services, allows for sufficient planning management. Despite this, regions globally do not have accurate consistent coverage at the national scale. This certainly true Ethiopia. Large-area land-cover characterization (LALCC), scale thus an essential first step in yet itself problematic. Such LALCC based remote-sensing image classification associated with spectrum technical challenges such as data availability, radiometric inconsistencies within/between images, big processing. Radiometric could be exacerbated areas, Ethiopia, high frequency cloud cover, diverse climate patterns, large variations elevation topography. Obtaining explanatory variables that are more robust can improve accuracy. To create base map future study large-scale agricultural transactions, we produced recent Of key importance was creation methodology repeatable and, such, used to earlier, comparable classifications same region. We examined effects band normalization different time-series compositing methods Both top atmosphere surface reflectance products from Landsat 8 Operational Land Imager (OLI) were tested single-time independently, where latter resulted 1.1% greater overall Substitution original spectral bands normalized difference indices additional improvement 1.0% Three approaches multi-temporal compositing, using OLI Moderate Resolution Imaging Spectroradiometer (MODIS) data, including sequential i.e., per-pixel summary measures predefined periods, probability density function distribution values, sinusoidal models. Multi-temporal composites improved accuracy up 4.1%, respect advantage OLI-driven over MODIS-driven composites. Additionally, night-time light classification. The its derivatives by 1.7%. producer’s Urban/Built class cost decreasing user’s Results this research aid producers decisions related operational large-area mapping, especially selecting input allow national-level timely fashion.