Long-Term Monitoring of Vegetation Dynamics and Water Resources with Remote Sensing in the Tropical Ecosystems of the Yucatan Peninsula, Mexico
Author
Jimenez Hernandez, EduardoIssue Date
2024Advisor
Didan, Kamel
Metadata
Show full item recordPublisher
The University of Arizona.Rights
Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.Embargo
Release after 08/15/2025Abstract
The Yucatan Peninsula, spanning southeast Mexico, Guatemala, and Belize, contains the second largest remnant tropical forest in Latin America, second only to the Amazon. This region plays a critical role in biodiversity conservation and acts as a major carbon sink. This study presents a comprehensive framework for monitoring land cover changes in the Yucatan Peninsula, consisting of five steps: 1) Generating high-resolution land cover maps using machine learning, 2) Validating these maps with field data, 3) Comparing the maps against existing datasets, 4) Analyzing land cover changes, and 5) Assessing the effectiveness of protected areas in mitigating land cover change. Three land cover maps were created for the periods 2013-2016, 2016-2019, and 2019-2022 using a Random Forest classification model based on Landsat 8 satellite images. The model’s performance was evaluated using F1-scores for eleven classes, yielding the following results: evergreen tropical forest (89%), deciduous tropical forest (81-83%), water (79-80%), mangrove (79-80%), wetland (76-78%), agriculture (72-76%), urban (60-65%), savanna (53-55%), coastal vegetation (48-50%), and oak forest (22-24%). Validation with 1,872 field data points from the National Inventory of Forests and Soils indicated high accuracy for evergreen tropical forest (90%), deciduous tropical forest (83-84%), mangrove (81-82%), and savanna (75-81%). When compared with existing products from the Monitoring Activity Data for the Mexican REDD+ program and the North American Land Change Monitoring System, similar accuracies were observed. Land cover change analysis revealed that 725,404 hectares of tropical forest (9% of deciduous tropical forest and 4.4% of evergreen tropical forest) and 7,370 hectares (13.3%) of savanna transitioned into agricultural land. Additionally, 58,066 hectares (1.5% of the area) transitioned into urban areas from various land cover classes. Despite these changes, 86-93% of tropical forest remained intact, and there was even a gain in forested areas from agricultural land (3-6%). The effectiveness of the Calakmul Biosphere Reserve (CBR) in preventing land cover changes was assessed by comparing land cover changes inside the reserve to those in a similar area outside it. Results showed that 24,853 hectares of tropical forest outside the CBR transitioned into agriculture, compared to only 4,412 hectares inside the reserve, representing a sixfold difference. Similarly, the transition to urban areas was 337 hectares outside versus 93 hectares inside, a threefold difference. These findings suggest that the reserve effectively prevents land cover changes. Most tropical forest loss occurred immediately around the CBR’s north and east sides, where roads, agriculture, and rural communities are located, highlighting the importance of protected areas in the Yucatan Peninsula. Increased efforts are needed to expand protected areas to counteract the accelerating pressure from cropland expansion, which poses the main threat to the region’s forests.The second part of this work provides a two-decade descriptive analysis of the hydrological cycle and assesses water resources in the Yucatan Peninsula’s karst aquifer. The aquifer’s geohydrological characteristics favor rapid infiltration, making surface runoff and perennial rivers practically nonexistent throughout most of the Peninsula. Therefore, a compelling assumption is the ecohydrology of this region is driven mostly by precipitation and evapotranspiration (ET). By incorporating NDVI and terrestrial water storage (TWS), the research explores the relationship between the ecosystem and hydrological processes and identifies trends in water resource availability. Using time series data from various remote sensing platforms and in-situ water level measurements from 2000 to 2023, Singular Spectrum Analysis was employed to examine trends and seasonality. Despite short-term oscillations, all variables demonstrated a slightly increasing long-term trend, with a common periodicity of 12 months across all variables and a 6-month periodicity shared by NDVI and ET. Cross-correlation analysis revealed a strong correlation between TWS and NDVI (r=0.674), and moderate between precipitation and ET (r=0.433), indicating synchronous hydrological processes, such as infiltration, recharge, and ET, within the same month. NDVI, serving as a proxy for vegetation water absorption, showed a two-month lag behind precipitation, as did TWS (r=0.707 and r=0.562, respectively). Surface water balance time series, calculated by subtracting ET from precipitation, consistently delineated water deficit and surplus areas over the study period. The monthly water balance series led TWS by two months, supported by a moderate cross-correlation (r=0.572). Linear models revealed TWS depletion rates of 7 mm year-1 (2002-2013), 36 mm year-1 (2014-2020), and 51 mm year-1 (2021-2023), corroborated by comparison with data from observation wells. In the subregion around the city of Merida, comprising 243 observation wells, moderate cross-correlation (r=0.4363 and ρ=0.4909) suggested the utility of TWS for monitoring water resources. However, region-wide analysis showed weaker correlation (r=0.1182 and ρ=0.1699) due to uneven well distribution and missing data. Continued field data collection, combined with remote sensing tools, is essential for ongoing monitoring of the forest and water resources in the Yucatan Peninsula.Type
Electronic Dissertationtext
Degree Name
Ph.D.Degree Level
doctoralDegree Program
Graduate CollegeBiosystems Engineering