UCL DEPARTMENT OF GEOGRAPHY
Dr. Mathias Disney
UCL Logo
Dr. Mathias Disney
>
Contact Details
>
Current Teaching
>
Publications
>
Research Interests
>
Research Students
 

PhD Research opportunities: this project is now taken (June 2012)

Quantifying forest state and degradation: exploiting new lidar measurements

Supervisors: Dr. Mat Disney, Dr. Simon Lewis, Prof. Philip Lewis.


3D model simulations of a Scots Pine canopy (l) and a mountain birch canopy (r), used for simulating and understanding lidar signals.. More details here.

Aims

This project will develop methods to exploit new lidar scanning systems for quantifying forest biomass (state and change). The project aims to establish how well airborne laser scanning (ALS) can estimate biomass via measurement of key canopy structural properties (height, cover, structure) over contrasting forest biomes (temperate, tropical). The project will test the hypotheses that: biomass can be estimated from canopy height and cover via ALS; canopy cover can be derived from ALS independent of structural assumptions, reducing the need for empirical calibration of height/biomass relationships; ALS-derived biomass estimates can be tested using stem density and diameter measurements obtained from terrestrial laser scanning (TLS).

Context

Quantifying forest biomass is increasingly important for a range of applications including forestry, terrestrial C-cycle responses to climate change and disturbance (Lewis et al., 2009) as well as and resource management, particularly in the tropics (Brown and Gurevitch, 2004). Degradation and disturbance monitoring require low-cost, rapid, repeatable estimates of biomass (REDD OAR, 2009). While satellite optical and radar remote sensing can provide large-scale coverage (Saatchi et al, 2011), they can be limited by cloud cover (optical), lack of sensitivity to high biomass (both) and perhaps most importantly, the difficulty of relating biomass to rather indirect measurements (Turner et al, 2004).Lidar is an extremely promising alternative, providing much more direct measurement of canopy properties. Newer waveform systems provide a rich source of distance-resolved information on canopy structure (gap fraction, leaf area density), potentially allowing for much tighter constraint on models for parameter retrieval, as well as a consistent framework for combining ground and airborne data (lidar, but also optical and radar more generally). ALS offers the advantage of allowing estimates of height (Lefsky et al., 2002), which can be related to biomass through field-scale calibration of allometric relationships (Hopkinson, 2007). ALS thus holds great promise for rapid characterisation of biomass, with acquisition costs reducing as the ability to cover larger areas grows. A key requirement for this is the development of parameter retrieval methods that are robust, and ideally less reliant on local calibration and validation. Understanding the key variables that affect the lidar signal can reduce the requirements for local calibration, as well as indicating how calibration information can be collected and used as efficiently as possible. TLS systems are much more limited spatially than ALS, but can provide alternative, complementary information on 3D canopy structure, particularly diameter-at-breast height (dbh) and crown shape (Cote et al., 2009; Rutzinger et al., 2010). TLS data may therefore provide a route for testing and validating ALS measurements, particularly given their increasingly wide deployment.Models provide a powerful framework for interpreting lidar signals, and developing parameter estimation methods, as they allow the canopy and system characteristics to be controlled for explicitly (Disney et al., 2010). Much lidar modelling for parameter retrieval tends to be statistical/empirical, or results in ~effective~ parameters which cannot be validated directly (Nęsset & Gobakken, 2005; Ni-Meister et al., 2008). Recent work has shown that explicit 3D models can be used to account for (and separate) the structural and biochemical components of the lidar signal and are ideally suited for testing lidar parameter retrieval methods in complex forest canopies (Calders, 2010; Disney et al., 2010; Hancock, 2010; Hancock et al., 2011). Methods relating biomass to structure which are less dependent on local calibration would therefore be a significant advance for exploiting ALS more widely.

Method

The project has three linked components: i) theoretical development of information extraction methods from lidar; ii) UK field experiments to prototype these methods; iii) field experiments in Gabon where unique ALS and ground inventory data are available. The use of two field sites will allow testing of retrievals across widely contrasting forest structures (temperate, tropical).The first component will involve using 3D canopy models developed at UCL to simulate ALS and TLS data over a range of canopy densities (understory and overstory) and disturbance histories, with varying instrument and sampling characteristics. The advantages of this approach are that the canopy structural and radiometric properties are known a priori, and variations in sensor and survey methods can be accounted for explicitly. Current lidar modelling approaches tend to assume a particular crown shape and within-crown extinction property (statistically) related to leaf area density and arrangement (Goodwin et al., 2007; Ni-Meister et al., 2010). Recent work has shown that assuming crown ~archetypes~ may be problematic (Calders, 2010; Calders et al., 2011). So while this approach may seem to work well, it may be that the crown shape and within-crown scattering properties co-vary to allow an apparently reasonable solution, but one that represents unrealistic canopy properties. Also, as the modelled crown shape and within-crown properties are ~effective~ (rather than directly physical), it is also very difficult to validate parameters retrieved this way. Recent work has shown that there is potentially a very direct route to canopy cover from ALS, dependent only on the ratio of crown-to-ground reflectance (Armston et al., 2011). This potentially provides a very strong constraint on biomass retrieval from ALS and, importantly, requires no structural assumptions. The approach here will use detailed 3D canopy scattering models explicitly representing canopy elements down to the leaf/needle scale, thus avoiding the need for simplifying assumptions. The resulting simulated signal will be used to test canopy cover, height and biomass retrieval from ALS under the assumptions of changing structure only.The second component will involve testing the understanding derived from the theoretical development using measured lidar data. ALS data already collected by the CASE partner at a UK field site in Leicestershire will be exploited. Stand structure will be characterised via manual measurement, and deployment of TLS (existing data will be used where possible). Three visits will be made to collect data across 4-6 forest stands, covering a range of density (as a proxy for degradation/change). Using the theoretical framework developed from the 3D modelling, canopy structure will be derived from the ALS and TLS data separately, and in combination. The TLS and field measurement will provide estimates of dbh and number density required to validate retrievals from ALS.The third component will involve applying parameter retrieval methods to lidar data from a tropical forest in Gabon. The student will have access to waveform ALS data being collected by the Gabonese government through their National Parks programme, including at long-term inventory plots with detailed dbh and height measurements (Lewis et al., 2009). TLS will be deployed along transects, to collect dbh and density estimates. There are currently few (if any) tropical sites where TLS, ALS and inventory data are available. Lidar-derived structural information will be used to test whether it is possible to detect differences between plots with different disturbance histories.

Training

The student will receive training in radiative transfer theory and modelling, lidar remote sensing systems, analytical and numerical methods, computing and tropical ecology during the early part of their studies. The student will visit the CASE partner at the outset to discuss project aims and planning, and to receive training in lidar equipment and processing as required. The student will meet with the other project partners during the first year to discuss degradation (Brown), forestry measurements and survey practice (Casella) and multi-spectral lidar (Danson). The student will benefit hugely from this range of commercial and academic links, giving both the broad perspective and detailed technical skills required to achieve the project aims. The student will also benefit from participating in fieldwork, particularly overseas where they will be part of a much larger project. The student will be expected to present their results at conferences, and encouraged to submit results for journal publication. The student will present project results to the CASE partner on completion.

Eligibility

See NERC website for eligibility.

 

  Maintained by Mathias Disney Last Updated: 2 March 2007

Department of Geography - University College London - Gower Street - London - WC1E 6BT - Telephone: +44 (0)20 7679 5500
Copyright © 1999-2005 UCL


Search by Google