Gibson, Carey

ABSTRACT

Development of Forest Degradation Indicators from Long-term Trajectories of Multispectral Satellite Images, and their Projections into the Future under Climate Change, in Ontario, Canada

Md. Mozammel Hoque

Ontario forests are affected by natural and anthropogenic disturbances leading to forest degradation, which significantly impact local ecosystems, health, safety, and economy. This thesis develops a methodology for the continuous assessment, mapping, and monitoring of present and historic (1972–2020) forest disturbances, and future forest degradation trends and projections, using remote sensing data, ground measurements, and predictive models in an Ontario forested area. After testing four supervised classification algorithms, support vector machine was found to be the most robust, consistent, and effective for land cover classification. Seven vegetation indices derived from Landsat and MODIS platforms were used to derive forest degradation indicators (FDIs), which were combined into one composite forest degradation indicator (CFDI) for each year, using the principal component analysis image fusion approach. The CFDI was the most informative indicator. The computed FDIs from available large multispectral image stacks were statistically related to historical climate variables. These relationships were used to project future FDIs related to climate variables derived from General Circulation Models through multiple linear regression models. Spatially-explicit maps of relevant climatic variables and of long-term historical forest degradation were developed from the LandTrendr trajectory analysis. Climate variables P, MA1, MA2, and CFDI were strongly correlated, allowing for the development of a model with a high coefficient of determination, R2 (0.93), and low RMSE (0.28) to predict future values. Forest disturbances (as CFDI) were also monitored from 1972–2020. Overall, these relationships allowed for to the creation of spatially-explicit, long-term historical forest degradation maps derived from the Landtrendr trajectory analysis. Historical and future forest degradation maps identified the areas with projected high vulnerability to climate change, as well as the actual and potential changes in forest cover under climate change. The results indicated 2050 will experience an average temperature increase of 3.0°C, projected yearly decrease in precipitation of 109.5 mm, evapotranspiration increase of 73.0 mm, and moisture deficits of 28.47 mm (MA1) and 37.60 mm (MA2), leading to increased forest degradation.

Author Keywords: Climate change impacts, Forest degradation indicators, Forest disturbance and degradation, Land cover classification, Projections of 2050 forest degradation under climate change, Remote sensing technology

Currently, UAVs are deployed to measure crop health in a timely manner by mapping vegetation indices. A study using two different fields was conducted in order to search for a relationship that may exist between crop health and soil fertility. A UAV equipped with sensor technology was used for mapping of vegetation indices which were then statistically compared to soil nutrient data collected via soil sampling. Elevation data was also collected which was then statistically compared to soil nutrients as well as crop health. Results of this study were unfortunately impacted by variables outside of the researcher's control. Moisture became the greatest limiting factor in 2016 followed by an excess of rain in 2017. Results did not show any promising correlations as moisture uncontrollably became the defining variable. Further research in a more controlled setting will need to be conducted in order to explore this potential relationship.

Author Keywords: Agriculture, Multispectral Imagery, Precision Agriculture, Proximal Soil Sensing, Remote Sensing, Unmanned Aerial Vehicle

The forestry sector has a well-developed history of using remote sensing to identify structural characteristics of forests and to detect and attribute changes that occur in forested landscapes. Monitoring the recovery of disturbed forests is an important factor in long term forest management. However, forest that is recovered spectrally may not be recovered when considered in terms of a Free to Grow assessment. A Free to Grow assessment is used in Ontario to determine whether a disturbed site will likely achieve a desired future state, i.e., is recovered according to a forestry perspective. The objective of this research was to determine the relationship between a pixel-based Landsat Time Series of spectral recovery and the results of Free to Grow assessments. Spectral trajectories were generated from representative pixels within known harvested forest areas. Results indicate that while Free to Grow sites often achieve spectral recovery (>90%), many non-Free to Grow sites were classified as spectrally recovered, suggesting that improved methods of spectral recovery monitoring are needed.

Author Keywords: forest recovery, Free to Grow, Landsat Time Series, LandTrendr, Pixel-based, spectral recovery

In this thesis, a methodological framework for the assessment and monitoring of forest degradation based on remote sensing techniques and field data, as part of the REDD+ programme, is presented. The framework intends to support the implementation of a national Monitoring, Verification and Report (MRV) system in developing countries. The framework proposed an operational definition of forest degradation and a set of indicators, namely Canopy Cover (CC), Aboveground Biomass (AGB) and Net Primary Productivity (NPP), derived from remote sensing data. The applicability of the framework is tested in a sub-deciduous tropical forest in the Southeast of Mexico. The results from the application of the methodological framework showed that the higher rates of forest degradation, 1596-2865 ha·year-1, occur in areas with high population density. Estimations of aboveground biomass in these degraded areas span from 1 to 24 Mg·ha-1, with a rate of carbon fixation ranging from 130 to 246 gC·m2·year. The results also showed that 43 % of the forests of the study area remain with no evident signs of degradation, as detected by the indicators selected, during the period evaluated. The integration of the different elements conforming the methodological framework for the assessment and monitoring of forest degradation enabled the identification of areas that maintain a stable condition and areas that change over the period evaluated. The methodology outlined in this thesis also allows for the identification of the temporal and spatial distributions of forest degradation based on the indicators selected, and it is expected to serve as the basis for operations of the REDD+ programme with the appropriate adaptations to the area in turn.

Author Keywords: Forest degradation, Monitoring, REDD+, Remote Sensing, Tropical forest