Advanced

Remote Sensing of Forest Decline in the Czech Republic

Ardö, Jonas LU (1998) In Meddelande från Lunds Universitets Geografiska Institutioner, Avhandlingar 135.
Abstract (Swedish)
Popular Abstract in Swedish

Avhandlingen beskriver hur man med elektromagnetisk reflekterar strålning, registrerad från satelliter vilka cirkulerar runt jorden kan lokalisera och kvantifiera skador på barrskog i Tjeckien.



Signifikanta samband mellan reflektans och barrförluster gör det mögligt att utföra översiktliga uppskattningar av dessa varrförluster med satellitdata.



Genom att jämföra skogsskadornas rumsliga utbredning med förekomsten av luftföroreningar, klimat samt topografi kan hypoteser om olika orsakssamband testas och undersökas.
Abstract
This thesis describes the localisation and quantification of deforestation and forest damage in Norway spruce forests in northern Czech Republic using Landsat data.



Severe defoliation increases the spectral reflectance in all wavelength bands, especially in the mid infrared region. These spectral differences allow the separation of three damage categories with an accuracy of 75% using TM data and regression-based relationships. Estimating the same categories using an artificial neural network, multitemporal TM data and topographic data yields slightly higher accuracy (78%). The methods are comparable when using identical input data, but the neural network more efficiently manage large input data sets without... (More)
This thesis describes the localisation and quantification of deforestation and forest damage in Norway spruce forests in northern Czech Republic using Landsat data.



Severe defoliation increases the spectral reflectance in all wavelength bands, especially in the mid infrared region. These spectral differences allow the separation of three damage categories with an accuracy of 75% using TM data and regression-based relationships. Estimating the same categories using an artificial neural network, multitemporal TM data and topographic data yields slightly higher accuracy (78%). The methods are comparable when using identical input data, but the neural network more efficiently manage large input data sets without pre-processing.



The estimated coniferous deforestation in northern Bohemia from 1972 to 1989 reveals especially affected areas between 600 and 1000 m.a.s.l. and on slopes facing south and southeast. The sector downwind a large source of sulphur dioxide was strongly deforested.



Comparing regional forest damage statistics to three methods estimating harmful effects of sulphur dioxide on Norway spruce yielded significant relationships versus level of forest damage and accumulated salvage felling. Quantifying the effect of data uncertainties permit mapping the probabilities of areas to be significantly over or below thresholds for harmful effects on spruce forests.



Satellite based estimation of coniferous forest health is a good complement to field surveys and aerial photography. (Less)
Please use this url to cite or link to this publication:
author
opponent
  • Professor Woodcock, Curtis, Department of Geography, Boston University, Boston
organization
publishing date
type
Thesis
publication status
published
subject
keywords
Czech Republic, neural networks, sulphur dioxide, air pollution, Norway spruce, deforestation, forest decline, GIS, spectral characteristics, Remote sensing, Landsat, Ore Mountains, Geology, physical geography, Geologi, fysisk geografi
in
Meddelande från Lunds Universitets Geografiska Institutioner, Avhandlingar
volume
135
pages
150 pages
publisher
Lund University Press
defense location
Geografiska institutionernas föreläsningssal, 3:e våningen, Sölvegatan 13, Lund
defense date
1988-05-20 13:15
external identifiers
  • other:ISRN: LUNDBS/NBNG--96/1135--SE
  • scopus:0031707503
ISSN
0346-6787
ISBN
91-79-66-528-4
language
English
LU publication?
yes
id
c6bcb2fc-a545-46d7-80eb-a2c88351477f (old id 38648)
date added to LUP
2007-06-20 11:00:43
date last changed
2017-01-01 07:07:22
@phdthesis{c6bcb2fc-a545-46d7-80eb-a2c88351477f,
  abstract     = {This thesis describes the localisation and quantification of deforestation and forest damage in Norway spruce forests in northern Czech Republic using Landsat data.<br/><br>
<br/><br>
Severe defoliation increases the spectral reflectance in all wavelength bands, especially in the mid infrared region. These spectral differences allow the separation of three damage categories with an accuracy of 75% using TM data and regression-based relationships. Estimating the same categories using an artificial neural network, multitemporal TM data and topographic data yields slightly higher accuracy (78%). The methods are comparable when using identical input data, but the neural network more efficiently manage large input data sets without pre-processing.<br/><br>
<br/><br>
The estimated coniferous deforestation in northern Bohemia from 1972 to 1989 reveals especially affected areas between 600 and 1000 m.a.s.l. and on slopes facing south and southeast. The sector downwind a large source of sulphur dioxide was strongly deforested.<br/><br>
<br/><br>
Comparing regional forest damage statistics to three methods estimating harmful effects of sulphur dioxide on Norway spruce yielded significant relationships versus level of forest damage and accumulated salvage felling. Quantifying the effect of data uncertainties permit mapping the probabilities of areas to be significantly over or below thresholds for harmful effects on spruce forests.<br/><br>
<br/><br>
Satellite based estimation of coniferous forest health is a good complement to field surveys and aerial photography.},
  author       = {Ardö, Jonas},
  isbn         = {91-79-66-528-4},
  issn         = {0346-6787},
  keyword      = {Czech Republic,neural networks,sulphur dioxide,air pollution,Norway spruce,deforestation,forest decline,GIS,spectral characteristics,Remote sensing,Landsat,Ore Mountains,Geology,physical geography,Geologi,fysisk geografi},
  language     = {eng},
  pages        = {150},
  publisher    = {Lund University Press},
  school       = {Lund University},
  series       = {Meddelande från Lunds Universitets Geografiska Institutioner, Avhandlingar},
  title        = {Remote Sensing of Forest Decline in the Czech Republic},
  volume       = {135},
  year         = {1998},
}