Retrieving boreal forest structure from remote sensing data using reflectance modelling and machine learning

Eelis Halme

Retrieving boreal forest structure from remote sensing data using reflectance modelling and machine learning

Julkaisun otsikon käännös: Boreaalisen metsän rakenteen arviointi kaukokartoitusaineistosta käyttäen heijastusmallinnusta ja koneoppimista

Eelis Halme

Rakennetun ympäristön laitos

Tutkimustuotos: Doctoral Thesis › Collection of Articles

Abstrakti

Boreaalisilla metsillä, joihin ilmastonmuutos vaikuttaa yhä enenevässä määrin, on merkittävää ekologista arvoa ja ne ovat keskeisiä maailmanlaajuisessa hiilen kiertokulussa. Siksi näiden metsien tehokas seuranta on erittäin tärkeää. Tämän väitöskirjan tavoitteena on kehittää menetelmiä Pohjois-Euroopan boreaalisten metsien seurantaan keskittyen ensisijaisesti Suomen metsiin. Väitöskirjassa kehitetään uusi hybridimenetelmä, joka yhdistää metsän heijastusmallin ja koneoppimisalgoritmin metsämuuttujien arvioimiseksi passiivisesta optisesta kaukokartoitusaineistosta. Aluksi arvioitiin koneoppimisalgoritmien soveltuvuutta metsämuuttujien ennustamiseen Suomen metsissä ja määritettiin tarkkojen ennusteiden edellyttämä kaukokartoitusaineiston spektrinen resoluutio. Seuraavaksi metsän heijastusmallin mallinnustarkkuutta parannettiin määritetylle spektriselle resoluutiolle. Lopuksi kehitettiin uusi hybridimenetelmä, jossa metsän heijastusmalli yhdistettiin koneoppimisalgoritmin kanssa. Lisäksi erittäin korkean resoluution kuvia hyödynnettiin parantamaan hybridimenetelmän vakautta sekä tarkkuutta. Tämä väitöskirja edistää usealla tapaa boreaalisten metsien seurantamenetelmien kehittämistä. Ensinnäkin se osoittaa, että hyperspektrikuvantamisen lisäarvo liittyy ensisijaisesti metsämuuttujiin, jotka sisältävät lajikohtaista tietoa, kun taas perinteinen avaruuspohjainen monispektrinen kaukokartoitusaineisto riittää metsän rakennemuuttujien tarkkaan ennustamiseen. Toiseksi metsän tilajärjestys sekä oksien ja lehtien pinta-alan suhde havaittiin merkittävästi vaikuttavan heijastusmallinnuksen tarkkuuteen. Kolmanneksi tulokset osoittavat, että hybridimenetelmät ovat erittäin lupaavia metsän rakenteen arvioinnissa. Lupaavista tuloksista huolimatta haasteita on edelleen. Väitöskirjassa kehitetyn hybridimenetelmän suurin haaste oli harjoitus- ja testiaineistojen välinen eroavuus, mikä korosti spektrisen ja rakenteellisen edustavuuden tärkeyttä koneoppimiseen perustuvissa sovelluksissa. Väitöskirja esittää olennaisia havaintoja boreaalisten metsien seurantamenetelmien kehittämiseen, mutta se tuo esiin myös useita haasteita ja rajoituksia, jotka on ratkaistava. Näiden haasteiden ratkaiseminen vaatii lisäselvityksiä, mikä korostaa tarvetta uudelle tutkimukselle. Kaiken kaikkiaan tämä väitöskirja tarjoaa vankan perustan metsän heijastusmalleja käyttävien hybridimenetelmien kehittämiselle.

Julkaisun otsikon käännös	Boreaalisen metsän rakenteen arviointi kaukokartoitusaineistosta käyttäen heijastusmallinnusta ja koneoppimista
Alkuperäiskieli	Englanti
Pätevyys	Tohtorintutkinto
Myöntävä instituutio	Aalto-yliopisto
Valvoja/neuvonantaja	Rautiainen, Miina, Vastuuprofessori Mõttus, Matti, Ohjaaja, Ulkoinen henkilö
Kustantaja	Aalto University
Painoksen ISBN	978-952-64-2534-4
Sähköinen ISBN	978-952-64-2533-7
Tila	Julkaistu - 2025
OKM-julkaisutyyppi	G5 Artikkeliväitöskirja

Tutkimusalat

kaukokartoitus
Sentinel-2
hyperspektrikuvantaminen
boreaalinen metsä
heijastusmallinnus
koneoppiminen
hybridi-inversio

YK:n kestävän kehityksen tavoitteet

Tämä tuotos edistää seuraavia kestävän kehityksen tavoitteita:

Pääsy asiakirjaan

http://urn.fi/URN:ISBN:978-952-64-2533-7

Siteeraa tätä

@phdthesis{68030913edfc44a3ac5739ee4bd9cc1c,

title = "Retrieving boreal forest structure from remote sensing data using reflectance modelling and machine learning",

abstract = "Boreal forests, which are increasingly affected by climate change, hold significant ecological value and are central to the global carbon cycle. Therefore, the effective monitoring of these forests is of great importance. This thesis aims to advance methodologies for monitoring the northern European boreal forests, with a primary focus on Finnish forests. Specifically, the thesis develops a new hybrid method that combines a forest reflectance model with a machine learning algorithm to retrieve forest variables from passive optical remote sensing data. To develop this method, the research began by assessing the suitability of machine learning algorithms for retrieving forest variables in Finnish forests and determining the spectral resolution of the remote sensing data required for accurate forest variable retrievals. Subsequently, the accuracy of a forest reflectance model was improved for the determined spectral resolution. Finally, a new hybrid method was developed by integrating the forest reflectance model with a machine learning algorithm. In addition, very-high-resolution images were used to enhance the robustness and accuracy of the hybrid method. This thesis makes several contributions to the advancement of boreal forest monitoring methods. First, it demonstrates that the added value of hyperspectral imaging is primarily linked to forest variables that include species-specific information, whereas traditional spaceborne multispectral remote sensing data is sufficient for accurately retrieving common forest structural variables. Second, the spatial pattern of trees and the ratio of branch area to leaf area were found to significantly influence forest reflectance modelling accuracy. Third, the results show that hybrid methods hold great promise for retrieving forest structure. Despite the promising results, challenges remain. The main challenge of the developed hybrid method was the disparity between the datasets used for training and testing, which highlighted the importance of spectral and structural representativeness in machine learning-based applications. While the thesis presents relevant findings for advancing methodologies in boreal forest monitoring, it also highlights several challenges and limitations that need to be addressed. Overcoming these challenges requires further investigation, emphasising the need for additional research. Overall, this thesis provides a solid foundation for developing hybrid methods that use forest reflectance models, contributing to the evolving field of hybrid approaches in remote sensing research.",

keywords = "remote sensing, Sentinel-2, hyperspectral imaging, boreal forest, reflectance modelling, machine learning, hybrid inversion, kaukokartoitus, Sentinel-2, hyperspektrikuvantaminen, boreaalinen mets{\"a}, heijastusmallinnus, koneoppiminen, hybridi-inversio, remote sensing, Sentinel-2, hyperspectral imaging, boreal forest, reflectance modelling, machine learning, hybrid inversion",

author = "Eelis Halme",

year = "2025",

language = "English",

isbn = "978-952-64-2534-4",

series = "Aalto University publication series Doctoral Theses",

publisher = "Aalto University",

number = "86",

address = "Finland",

school = "Aalto University",

}

TY - BOOK

T1 - Retrieving boreal forest structure from remote sensing data using reflectance modelling and machine learning

AU - Halme, Eelis

PY - 2025

Y1 - 2025

N2 - Boreal forests, which are increasingly affected by climate change, hold significant ecological value and are central to the global carbon cycle. Therefore, the effective monitoring of these forests is of great importance. This thesis aims to advance methodologies for monitoring the northern European boreal forests, with a primary focus on Finnish forests. Specifically, the thesis develops a new hybrid method that combines a forest reflectance model with a machine learning algorithm to retrieve forest variables from passive optical remote sensing data. To develop this method, the research began by assessing the suitability of machine learning algorithms for retrieving forest variables in Finnish forests and determining the spectral resolution of the remote sensing data required for accurate forest variable retrievals. Subsequently, the accuracy of a forest reflectance model was improved for the determined spectral resolution. Finally, a new hybrid method was developed by integrating the forest reflectance model with a machine learning algorithm. In addition, very-high-resolution images were used to enhance the robustness and accuracy of the hybrid method. This thesis makes several contributions to the advancement of boreal forest monitoring methods. First, it demonstrates that the added value of hyperspectral imaging is primarily linked to forest variables that include species-specific information, whereas traditional spaceborne multispectral remote sensing data is sufficient for accurately retrieving common forest structural variables. Second, the spatial pattern of trees and the ratio of branch area to leaf area were found to significantly influence forest reflectance modelling accuracy. Third, the results show that hybrid methods hold great promise for retrieving forest structure. Despite the promising results, challenges remain. The main challenge of the developed hybrid method was the disparity between the datasets used for training and testing, which highlighted the importance of spectral and structural representativeness in machine learning-based applications. While the thesis presents relevant findings for advancing methodologies in boreal forest monitoring, it also highlights several challenges and limitations that need to be addressed. Overcoming these challenges requires further investigation, emphasising the need for additional research. Overall, this thesis provides a solid foundation for developing hybrid methods that use forest reflectance models, contributing to the evolving field of hybrid approaches in remote sensing research.

AB - Boreal forests, which are increasingly affected by climate change, hold significant ecological value and are central to the global carbon cycle. Therefore, the effective monitoring of these forests is of great importance. This thesis aims to advance methodologies for monitoring the northern European boreal forests, with a primary focus on Finnish forests. Specifically, the thesis develops a new hybrid method that combines a forest reflectance model with a machine learning algorithm to retrieve forest variables from passive optical remote sensing data. To develop this method, the research began by assessing the suitability of machine learning algorithms for retrieving forest variables in Finnish forests and determining the spectral resolution of the remote sensing data required for accurate forest variable retrievals. Subsequently, the accuracy of a forest reflectance model was improved for the determined spectral resolution. Finally, a new hybrid method was developed by integrating the forest reflectance model with a machine learning algorithm. In addition, very-high-resolution images were used to enhance the robustness and accuracy of the hybrid method. This thesis makes several contributions to the advancement of boreal forest monitoring methods. First, it demonstrates that the added value of hyperspectral imaging is primarily linked to forest variables that include species-specific information, whereas traditional spaceborne multispectral remote sensing data is sufficient for accurately retrieving common forest structural variables. Second, the spatial pattern of trees and the ratio of branch area to leaf area were found to significantly influence forest reflectance modelling accuracy. Third, the results show that hybrid methods hold great promise for retrieving forest structure. Despite the promising results, challenges remain. The main challenge of the developed hybrid method was the disparity between the datasets used for training and testing, which highlighted the importance of spectral and structural representativeness in machine learning-based applications. While the thesis presents relevant findings for advancing methodologies in boreal forest monitoring, it also highlights several challenges and limitations that need to be addressed. Overcoming these challenges requires further investigation, emphasising the need for additional research. Overall, this thesis provides a solid foundation for developing hybrid methods that use forest reflectance models, contributing to the evolving field of hybrid approaches in remote sensing research.

KW - remote sensing

KW - Sentinel-2

KW - hyperspectral imaging

KW - boreal forest

KW - reflectance modelling

KW - machine learning

KW - hybrid inversion

KW - kaukokartoitus

KW - Sentinel-2

KW - hyperspektrikuvantaminen

KW - boreaalinen metsä

KW - heijastusmallinnus

KW - koneoppiminen

KW - hybridi-inversio

KW - remote sensing

KW - Sentinel-2

KW - hyperspectral imaging

KW - boreal forest

KW - reflectance modelling

KW - machine learning

KW - hybrid inversion

M3 - Doctoral Thesis

SN - 978-952-64-2534-4

T3 - Aalto University publication series Doctoral Theses

PB - Aalto University

ER -

Retrieving boreal forest structure from remote sensing data using reflectance modelling and machine learning

Abstrakti

Tutkimusalat

YK:n kestävän kehityksen tavoitteet

Pääsy asiakirjaan

Sormenjälki

Siteeraa tätä