Lecture 11 Flashcards
What land types are commonly incorrectly mapped?
Urban - often not homogeneous so multiple classes appear in the same map. Water normally more homogeneous
What are the problems with total accuracy
- Summary value is an average
- Does not reveal if error was evenly distributed between classes or if some classes were really bad and some really good.
- One badly classified landcover type will have effect on total accuracy.
What other types of accuracy are there
Producer and user
What is user accuracy
error of commission (inclusion) e.g. 5 hardwood and 2 water pixels included erroneously in conifer category. perspective of the user of the classified map
Producer’s accuracy
Producer’s accuracy corresponds to error of omission (Exclusion). perspective of the maker of the classified map. E.g. 14 hardwood and 3 water pixels excluded erroneously from conifer category.
What is the importance of landcover maps
Land Cover is a principal factor controlling the exchange of energy, water, gases, and nutrients within the Earth system.
Landcover is an integral part of lots of modelling including global change modelling - explain this
Global change modelling:
– Boundary conditions for General Circulation Models (GCM) – thinking about exchange of energy between surface and atmosphere.
– Global biogeochemical and hydrological models.
Land cover/use change impacts carbon, water and energy cycling at all spatial scales…
But also biodiversity, wildlife habitats, resource management, fire/disaster monitoring.
Land over is one of the Essential Climatic Variables (ECV) identified by ESA
Land cover maps have…
▪ There are a wide range of land cover maps available which have :
▪ Different spatial characteristics
▪ Spatial resolution (30m – 1km)
▪ Spatial extent (national, regional, global) ▪ Different land surface information
▪ Land cover types (e.g. forest or deciduous forest, coniferous forest)
▪ Percentage cover per pixel (%forest, %grass, %impervious surface etc)
describe the example of a global land cover map - Copernicus landcover dataset
❑ 100m spatial resolution, annual (2015 – 2019)
Provides ❑ land cover map (i.e. land cover classes) ❑ Percentage cover per pixel (vegetation, bare ground, built-up etc)
describe Global Land30 (landcover dataset)
❑ Provide the first global land cover map using Landsat TM imagery
❑ Global 30m spatial resolution for 2000, 2010, 2020
❑ 10 land cover classes
What are the challenges of image classification
❑ Image pixels have a given pixel area defined by their spatial resolution (e.g., 10m, 300m, 1km, 10km)
❑ This leads to ‘mixed’ pixels (can influence level of misclassification, more in urban as they are heterogeneous)
❑ Pixels that contain more than one land cover type
❑ Surface heterogeneity and spatial resolution influence extent of missed pixels
❑ Urban areas are very heterogeneous (need ~1m spatial resolution data ideally)
❑ Agricultural fields tend to be larger/less heterogeneous (e.g. America)
❑ Not always the case (e.g. rice paddies
What types of models do landcover maps use
▪ Monitoring soil erosion
▪ Impact on soil health
▪ Nutrient degradation
▪ Salinity
River sediment concentration ▪ Declines in river health/biodiversity
Universal Soil Loss Equation (USLE)
▪ mathematical model that describes soil erosion
▪ Land cover is one of the inputs
▪ Monitor how changing land cover impacts soil erosion
Describe the urban heat island effect
Urban Heat Island intensity has several direct and indirect effects:
▪ ozone concentration & air quality
▪ Influence local meteorology
▪ increases demand for energy (cooling - > increased emissions)
Urban heat island more pronounced at night – particularly with little vegetation and high density areas. Energy is absorbed and released at night
What factors effect reflectance from a leaf in the visible, NIR and MIR?
Visible – leaf chemical composition (leaf pigments such as chlorophyll)
NIR – cell structure (water and air interfaces)
MIR – moisture content
Name 2 factors that influence the magnitude of spectral reflectance from a surface.
Type of material (vegetation, soil, concrete, metal)
Nature of surface (e.g. surface 3D structure) Specular (smooth) Diffuse (rough)
Spectral wavelength range
Other factors (slope, time (of day and time of year), material condition (e.g. wet,dry)