Integration of LiDAR data and satellite imagery for biomass estimation in conifer-dominated forest

Abstract (Swedish)

Ett första steg i arbetet att uppskatta kolflödet mellan landbaserade ekosystem och atmosfären är att så precist som möjligt analysera kvantiteten ovanjordisk biomassa. LiDAR teknologi har i detta avseende visat sig vara ett värdefullt verktyg. Det överlöpande målet med denna studie är att utveckla en förenklad metod, baserad på fjärranalys, för att uppskatta ovanjordisk biomassa för individuella träd över barrträdsdominerad skog i sydvästra Sverige. Eftersom mängden biomassa är beroende av vegetationstyp, har en vegetationsklassificering utförts i studieområdet. Både en lokal maximeringsalgoritm, som använder ett konstant utvärderingsfönster, och en så-kallad invers avrinningsområdessegmentationsmetod användes för extraktion av skogsinventerings-parametrar från en LiDAR-data baserad trädkronhöjdmodell. Den slutliga uppskattningen av ovanjordisk biomassa gjordes utifrån regressionsmodeller framtagna utifrån trädparametrar mätta i fält. Fältmätningarna gjordes för 83 jordlotter, där trädart, höjd och diameter (vid brösthöjd, 1.3 m) mättes. Resultaten visar att ovanjordisk biomassa varierar mellan mindre än 1 kg/m2 för väldigt ung skog up till 94 kg/m2 för mogen barrskog med ett standardfel (RMSE) på 2 kg/m2 och 4.7 kg/m2, respektive. Linjära regressionsmodeller visade att användningen av avrinningsområdessegmentering inte förbättrar resultaten (R2 = 0.79) i jämförelse med resultat från en lokal maximeringsalgoritm baserad på ett konstant utvärderingsfönster (R2 = 0.83).

Abstract

Popular science
Globally, carbon dioxide (CO2) is of great concern, as it is partly responsible for the increasing greenhouse effect, causing global warming. It is hypothesized that conifer-dominated forest in Västra Götaland County may be a large source of greenhouse gases (GHG) including CO2. Forest biomass stores a lot of carbon, which is exchanged in the form of CO2 with atmosphere through respiration and photosynthesis in the process known as carbon cycle. In order to approve or reject the hypothesis it is important to quantify the fluxes of CO2 from this drained, highly fertile coniferous forest. As a first step for achieving this task the amount of carbon stored in the forest should be estimated. Therefore, the overall goal of this study is to develop a simplified method for assessing aboveground biomass (AGB) for individual trees using remote sensing. In this study, AGB represents the sum of stem, bark, branch and foliage biomasses. The developed model was based on the use of airborne laser scanning and an image from the SPOT-5 satellite. The capability of laser systems to directly provide height measurements allowed us to derive the tree height of individual trees, while SPOT-5 image allowed us to classify tree types of the studied area. These two inputs were used for calculating AGB over the whole area of study and compared against AGB measured in the field based on parameters such as trees’ species, height and diameter. The overall accuracy of the developed model, when comparing AGB estimates from remote sensing with field observations was equal to 80 %, which proved the validity of the methodology applied. Availability of AGB estimates allows further monitoring of this forest ecosystem for disturbance or change. Furthermore, AGB values can be directly used in further studies of carbon stocks in the area.

Abstract

As a first step in the assessment of carbon flux between the terrestrial environment and the atmosphere it is important to accurately quantify the carbon stock of forest ecosystems. LiDAR technology, in this respect, has proved to be a valuable tool, able to provide accurate estimates of aboveground biomass (AGB). The overall goal of this study was to develop a simplified method for assessing AGB for individual trees using remote sensing in conifer-dominated forest in the southwest of Sweden. Vegetation classification of SPOT-5 image has been done in order to improve AGB estimates based on biomass dependence on vegetation types. Both local maximum algorithm using a constant size evaluation window and inverse watershed segmentation methods were used for forest inventory parameter extraction from a LiDAR-derived canopy height model. Final estimation of AGB was conducted using regression models derived from measured tree parameters in the field. Field measurements were performed over 83 plots by recording trees’ species, height and diameter at breast height (1.3 m). Results showed AGB to vary from less than 1 kg/m2 in very young forests up to 94 kg/m2 in mature spruce forests with RMSE of 2 kg/m2 and 4.7 kg/m2, respectively. Linear regression models showed that the introduction of the watershed segmentation does not improve the results (R2 = 0.79) in comparison to the results derived from local maximum algorithm using a constant size evaluation window (R2 = 0.83). Availability of AGB estimates allows further studies of carbon stocks as well as monitoring of this forest ecosystem for disturbance and change.

• author: Shendryk, Iurii
• supervisor: Margareta Hellström
• organization: Dept of Physical Geography and Ecosystem Science
• alternative title: Estimation of forest biomass using remote sensing
• course: NGEM01 20122
• year: 2013
• type: H2 - Master's Degree (Two Years)
• subject: Earth and Environmental Sciences
• keywords: digital elevation model, vegetation classification, AGB, aboveground biomass, remote sensing, LiDAR, SPOT-5, DEM, canopy height model, CHM, tree segmentation
• publication/series: Student thesis series INES
• report number: 266
• funder: Landscape Greenhouse Gas Exchange (LAGGE) project
• language: English
• additional info: Funder: Landscape Greenhouse Gas Exchange (LAGGE) project; Funder: Department of Physical Geography and Ecosystem Science, Lund University
• id: 3364920
• date added to LUP: 2013-01-17 12:47:50
• date last changed: 2013-01-17 12:47:50