Date of Degree
Earth & Environmental Sciences
Biodiversity | Forest Management | Other Forestry and Forest Sciences | Terrestrial and Aquatic Ecology
Remote sensing, microwave radar, forestry, conservation, biodiversity, ecosystem structure
This dissertation focuses on the utilization of remote sensing techniques to characterize ecosystem structure and functional types to inform on biodiversity conservation. This work is motivated by a guiding hypothesis that increased degradation of tropical forest biodiversity due to human activity results in in measurable changes in ecosystem processes that can be detected using remote sensing technologies. Throughout this dissertation, testing of aspects of this hypothesis involve comparing areas with varying degrees of human-induced degradation in a biodiverse tropical ecoregion, the lowland Chocó rainforest of Ecuador. Remote sensing data is used to evaluate changes in vegetation structure, thermal regimes, and ecosystem functioning metrics associated with biodiversity. Chapter 1 introduces the topics of tropical forest biodiversity and remote sensing. Chapters 2-4 are the main research chapters of the dissertation covering the topics of tropical forest degradation mapping, structural and thermal assessments of land use change, and classification of ecosystem functional types. Chapter 5 summarizes the main research findings to guide future investigations.
The primary research objective of Chapter 2 is the mapping of forest degradation and study of natural regeneration dynamics of managed conservation sites in Canandé region of the lowland Chocó. The thematic feature of this chapter focuses on evaluating the status of forest regeneration at the landscape scale by utilizing measures of texture derived from the decomposition radar polarization imagery. A novel radar metric, the Radar Forest Regeneration Index (RFRI), is developed from texture features derived from dual-polarization imagery. The texture of land cover classes followed gradient that correlated with forest regeneration status: cleared pasture RFRI = 0.32, regenerating forest RFRI = 0.50, and old-growth forest RFRI = 0.58. Forest and non-forest areas were mapped with producer’s accuracies greater than 99% and 96%, respectively, and user’s accuracy greater than 98% and 84%, respectively, using random forest classification. A second classification effort focused on temporal mapping of transitioning forest regions. This classification yielded an overall accuracy of 93% for all mapped forest types. Post-classification analysis demonstrated that inclusion of texture as an input feature significantly improved detection of forest regeneration stages with comparable results from full-polarimetry and dual-polarimetry imagery. This highlighted the utility of SAR for in mapping methodologies for monitoring tropical forest regeneration status for biodiversity conservation.
In Chapter 3, the perspective shifts from landscape scale to the local site level to develop connections between ecosystem structure and functional processes. In this chapter, UAV structure-from-motion techniques are employed to characterize the structure of a variety of tropical sites along a transitioning gradient from primary forest to plantation agriculture. The chapter assesses the associations between measures of ecosystem structure and functional attributes by examining in-situ collections of vertical air temperature profiles and UAV collections of diurnal land surface temperature. These assessments are conducted in relation to the land cover transitioning class. This chapter addresses how canopy height models developed from structure-from-motion compare to ground-based measurements of canopy structure. Key findings from this chapter establish associations between diurnal LST patterns above the canopy and below-canopy diurnal air temperature evolution as a function of canopy height. Canopy structure was tallest and most complex at forest sites, which also exhibited a lower diurnal LST range than the oil palm plantation and built-up/residential sites. Canopy heights and LST ranges were 0 m ± 0 m and 15.9 °C at the built-up site, 10.10 m ± 0.70 and 11.8°C for oil palm, 17.75 m ± 0.93 and 3.0°C for the secondary forest. This investigation is relevant for scaling remote sensing observations of ecological processes to a regional context associated with ecosystem functional types.
Chapter 4 seeks to advance the classification of ecosystem functional types across the regional domain. In this chapter, a combined structural-functional dataset was developed from satellite-based products associated with parameters of ecosystem structure and functioning. Variables of ecosystem functioning are derived from PALSAR-2, GEDI and ECOSTRESS. An evaluation of remote sensing variables correspondence with land use classifications is also conducted. Ecosystem structure and functioning products developed at the local scale in previous chapters are validated in a regional context in Chapter 4. A correlation analysis is performed to investigate complementary variables and their associated sensors. A PCA analysis is conducted to identify the features that explain the most variance in the structural-functional dataset. Most important features in PC1 included RFDI GLCMvar, HV and canopy height. Most important features in PC2 included mean and range LST. The land cover transition process from forest to agriculture to built-up/residential is suggested by the gradient in measurements of degradation (RFDI GLCMvar), canopy height, and range LST, respectively, along these classes, forest areas: 0.21 ± 0.11, 18.9 m ± 6.8, 12.2°C ± 3.7; agricultural areas: 0.43 ± 0.17, 10.1m ± 6.5, 13.9°C ± 4.6; built-up/residential: 0.57 ± 0.16, 5.5 m ± 3.7, 16.1°C ± 4.7. The machine-learning algorithm K-Means was utilized to classify ecosystem functional types. A preliminary hypertuning exercise found k = 4 as the optimal number of clusters in the study region. We concluded that EF types 1 and 4 represent end members in terms of ecosystem functioning associated with biodiversity. We use the findings to discuss associations between the classified ecosystem functional types and designations of land use protection status to inform on biodiversity conservation.
Chapter 5 summarizes the overall work and discusses two upcoming satellite missions that are anticipated to advance the study of biodiversity from space. The NISAR mission will provide observations at L-band and S-band frequencies relevant for characterizing ecosystem structure. The Surface Biology and Geology (SBG) mission will provide multispectral/IR observations to quantify functional traits, functional types, and composition of vegetation. The findings of this dissertation demonstrate the utility of L-band SAR for characterizing tropical forest structure, degradation, and regeneration associated with biodiversity conservation. Additionally, incorporation of SAR with structural and functional measures from spaceborne lidar, i.e., GEDI and optical/IR imagery, i.e., ECOSTRESS allows for classification of ecosystem functional types associated with land use and conservation status. Building off these findings, the next generation of satellite missions are well poised to significantly advance the study of biodiversity from space.
Tesser, Derek, "Remote Sensing Characterization of Ecosystem Structure and Functional Types to Inform on Biodiversity Conservation in the Lowland Chocó Rainforest" (2023). CUNY Academic Works.
This work is embargoed and will be available for download on Tuesday, September 30, 2025
Graduate Center users:
To read this work, log in to your GC ILL account and place a thesis request.
See the GC’s lending policies to learn more.