Modeller i remote sensing af vegetation

Kristine Bacher

Studenteropgave: Speciale


The purpose of this thesis is to investigate how the spectral signal from an agricultural field is influenced by the vertical distribution of chlorophyll through the canopy. In particular it is examined whether a uniform vertical distribution of chlorophyll improves the results of two types of models using satellite-based remote sensing data. Analysis of variance and simple linear regression is used to characterise the vertical chlorophyll profile in a canopy for field plots in five different European countries with a variety of crops. Out of a total of 68 fields, 28 fiels are found to be characterised by a uniform vertical chlorophyll distribution while 16 fields have either a decreasing or increasing vertical distribution of chlorophyll. The vertical chlorophyll profile for 6 fiels cannot be characterised by either of these models. 18 fields are not classified as observations lack homogeneity of variances. Biophysical conditions can be estimated by an empirical-statistical model where spectral vegetation indices are related to field measurements of biophysical variables such as leaf area index (LAI) or leaf chlorophyll content. The empirical relation cannot be generalised as it depends on crop species, context and time. A physically-based model such as PROSAIL can be used to estimate biophysical variables. PROSAIL uses input regarding the geometry and architecture of the canopy as well as leaf pigment content to simulate the spectral signal. The Matlab code for PROSAIL was adjusted to compensate for an error in the implentation of the ellipsoidal leaf angle distribution. Simulated reflectance data is used to evaluate the sensitivity of the vegetation indices. They are shown to be responsive of changes in leaf angle and LAI and the chlorophyll index in particular is shown to be sensitive of the leaf chlorophyll content. The empirical relation between the vegetation indices and LAI as well as total Chl content is improved for fields with a uniform vertical chlorophyll profile but remains highly context-dependent. PROSAIL is used to simulate the reflectance from a number of Danish wheat and barley fields from 2008 and 2009. The deviation between measured and simulated reflectance is evaluated in relation to LAI, plant type, landscape and vertical distribution of chlorophyll. It is shown that the deviation regarding red reflectance is influenced by the vertical chlorophyll profile however the simulation of reflectance of visible light is generally improved at low LAI. The simulation of near infrared reflectance is improved at higher LAI levels.

UddannelserMatematik, (Bachelor/kandidatuddannelse) KandidatGeografi, (Bachelor/kandidatuddannelse) Kandidat
Udgivelsesdato30 sep. 2014
VejledereCarsten Lunde Petersen


  • Statistics
  • Remote sensing