Carlos Portillo-Quintero

Plant diversity plays a fundamental role in ecosystem functioning and is essential for sustaining ecosystem services. National forest inventories are key instruments for assessing floristic diversity. However, their measurement protocols may introduce bias by omitting smaller individuals because of the stem diameter criterion used or the minimum plant size threshold applied. Such bias is exacerbated in dryland ecosystems where small-statured plants with low-branching stems are particularly abundant. In this study, we evaluated the effects of using basal diameter (BD) instead of diameter at breast height, and of sampling small individuals (BD ≥ 2.5 cm), on the estimation of abundance, alpha and gamma diversity and community composition in different vegetation types in NW Mexico. We found substantial underestimation due to the omission of smaller individuals in xeric shrubland and tropical dry forest, where gamma diversity may be underestimated by up to 209% and 139%, respectively. Broadleaf forest also showed strong underestimation (133%), whereas mixed conifer–broadleaf forests were unaffected. We discuss these differential effects and propose a methodology to attenuate this underestimation and achieve more accurate floristic diversity estimates from national forest inventories in dryland vegetation, which encompasses roughly one-third of the Earth’s surface and more than half of Mexico’s territory.

The advancement of hyperspectral remote sensing technology has enhanced the ability to assess and characterize land cover in complex ecosystems. In this study, a linear spectral unmixing algorithm was applied to NEON hyperspectral imagery in 2018 and 2022 to quantify the fractional abundance of dominant land cover classes, namely herbaceous vegetation, mixed forbs, and bare soil, across the Marvin Klemme Experimental Rangeland in Oklahoma. UAV imagery acquired during the 2023 field campaign provided high resolution reference data for model training. The LSU results revealed a decline in herbaceous cover from 16.02 ha to 11.56 ha and an expansion of bare soil from 3.37 ha to 6.39 ha, while mixed forb cover remained relatively stable (12.38 ha to 13.82 ha). Accuracy assessment using the UAV-derived validation points yielded overall accuracy of 84% and 60% at fractional thresholds of 50% and 75%, respectively. Although statistical tests indicated no significant change in mean fractional abundance (p > 0.05), slope-based trend maps captured localized vegetation loss and regrowth patterns. These findings demonstrate the effectiveness of integrating LSU with UAV data for detecting subtle yet ecologically meaningful shifts in semi-arid grassland composition.

Accurate estimates of carbon stocks from plant biomass are crucial for climate change mitigation. Temporal changes in biomass must also be quantified to accurately assess forest carbon emissions and absorption. LiDAR remote sensing is the primary method for mapping forest aboveground biomass (AGB) and its changes, as it can penetrate the canopy to create detailed three-dimensional point clouds. This study developed statistical models to estimate AGB and its changes in a tropical dry forest in the Yucatan Peninsula for 2013 and 2021. The models utilized biomass data from field measurements as the dependent variable and LiDAR-derived metrics as explanatory variables. Data were collected using airborne LiDAR in 2013 and ground-based LiDAR in 2021, allowing for a comparison of accuracy from different perspectives. The results indicated strong model performance, with coefficients of determination (R²) of at least 0.89 for AGB estimates in both years. Cross-validation errors ranged from 24% to 29%. For models estimating AGB increases, R² values reached approximately 0.92, with errors under 27% when differences in metrics from both years were analyzed. These results demonstrate the good performance of the models for estimating aboveground biomass and its increases using LiDAR and highlight the feasibility and importance of integrating field data and remotely sensed data in biomass estimates and its increases in tropical dry forests.

In the context of the ambitious initiative to expand tropical protected areas from current coverage (~15%) to 30–50% in the next couple of decades, community-based forest management (CBFM) systems offer an alternative to expanding forest conservation while providing social, economic, and environmental well-being. This is especially important for the most threatened tropical ecosystem: the tropical dry forest (TDF). Remnants of what were once large continuous tracts of TDF in lowlands and submontane areas are now highly fragmented across the subcontinent and under high anthropogenic pressure. CBFM empowers local communities to leverage resources sustainably, building resilience to environmental, economic, and political pressures without sacrificing ecosystem services. This chapter presents an exploratory investigation of CBFM resilience in the TDF landscapes of MexicO's Yucatan Peninsula, examining forest loss trends within parceled and non-parceled ejidos before, during, and after the SARS-CoV-2 pandemic. Results show higher TDF loss during the pandemic year, though ejidos demonstrated dynamic resilience, supporting CBFM as a means for both conservation and economic stability during socio-economic challenges. Further study is needed to clarify the socio-economic and environmental factors behind differing resilience patterns, including remittance impacts on post-pandemic land use decisions.

In the United States, several land use and land cover (LULC) data sets are available based on satellite data, but these data sets often fail to accurately represent features on the ground. Alternatively, detailed mapping of heterogeneous landscapes for informed decision-making is possible using high spatial resolution orthoimagery from the National Agricultural Imagery Program (NAIP). However, large-area mapping at this resolution remains challenging due to radiometric differences among scenes, landscape heterogeneity, and computational limitations. Various machine learning (ML) techniques have shown promise in improving LULC maps. The primary purposes of this study were to evaluate bagging (Random Forest, RF), boosting (Gradient Boosting Machines [GBM] and extreme gradient boosting [XGB]), and stacking ensemble ML models. We used these techniques on a time series of Sentinel 2A data and NAIP orthoimagery to create a LULC map of a portion of Irion and Tom Green counties in Texas (USA). We created several spectral indices, structural variables, and geometry-based variables, reducing the dimensionality of features generated on Sentinel and NAIP data. We then compared accuracy based on random cross-validation without accounting for spatial autocorrelation and target-oriented cross-validation accounting for spatial structures of the training data set. Comparison of random and target-oriented cross-validation results showed that autocorrelation in the training data offered overestimation ranging from 2% to 3.5%. The XGB-boosted stacking ensemble on-base learners (RF, XGB, and GBM) improved model performance over individual base learners. We show that meta-learners are just as sensitive to overfitting as base models, as these algorithms are not designed to account for spatial information. Finally, we show that the fusion of Sentinel 2A data with NAIP data improves land use/land cover classification using geographic object-based image analysis.

In the United States, several land use and land cover (LULC) data sets are available based on Landsat. These data sets often fail to accurately represent the features on the ground despite having several advantages over other data. Detailed mapping of complex heterogeneous landscapes for informed decision-making is possible using high spatial resolution orthoimagery from the National Agricultural Imagery Program (NAIP). However, large-area mapping at this resolution remains challenging due to radiometric differences between scenes, low spectral depth of imagery, landscape heterogeneity, and computational limitations. Machine learning (ML) techniques have shown promise in improving LULC maps. The primary purposes of this study are to examine and evaluate bagging (Random Forest; RF), Boosting (Gradient Boosting Machines [GBM] and Extreme Gradient Boosting [XGB]), and stacking ensemble models on sentinel 2A fusion data on NAIP imagery. We also compared the accuracy based on random cross-validation (80% / 20% split) without accounting for spatial autocorrelation and target-oriented validation (10 clusters) accounting for spatial structures of the training data set. To make the LULC map of a portion of Tom Green and Irion counties in central Texas, we used a time series of sentinel data and NAIP orthoimagery. We created several spectral indices, structural variables, and geometry-based variables, reducing the dimensionality of features generated on Sentinel and NAIP data. The random and target-oriented validation results show that autocorrelation in the training data offers over-optimistic results. In our case, the overestimation ranged from 2 % to 3.5 %. The XGB-boosted stacking ensemble on-base learners (RF, XGB, and GBM) improved model performance over individual base-learner. The main contribution of this research is that meta-learners are just as sensitive to overfitting as base models, as these algorithms are not designed to account for spatial information. Finally, we show that the fusion of Sentinel 2A data with NAIP data improves land use land cover classification using Geographic Object-Based Image Analysis (GEOBIA).

The Puuc Biocultural State Reserve (PBSR) is a unique model for tropical dry forest conservation in Mexico. Preserving forest biodiversity and carbon within the PBSR depends on maintaining low-impact productive activities coordinated by multiple communal and private landowners. In this study, we used state-of-the-art remote sensing data to investigate past spatial patterns in forest clearing dynamics and their relation to forest carbon density and forest plant species richness and diversity in the context of the forest conservation goals of the PBSR. We used a Landsat-based continuous change detection product for the 2000–2021 period and compared it to carbon density and tree species richness models generated from ALOS-2 PALSAR 2 imagery and national scale forest inventory data. The estimated error-adjusted area of detected annual forest clearings from the year 2000 until the year 2021 was 230,511 ha in total (±19,979 ha). The analysis of annual forest clearing frequency and area suggests that although forest clearing was significantly more intensive outside of the PBSR than within the PBSR during the entire 2000–2021 period, there is no evidence suggesting that the frequency and magnitude of forest clearing changed over the years after the creation of the PBSR in 2011. However, an emergent hotspot analysis shows that high spatiotemporal clustering of forest clearing events (hotspots) during the 2012–2021 period was less common than prior to 2011, and these more recent hotspots have been confined to areas outside the PBSR. After comparing forest clearing events to carbon density and tree species richness models, the results show that landowners outside the PBSR often clear forests with lower carbon density and species diversity than those inside the PBSR. This suggests that, compared to landowners outside the PBSR, landowners within the PBSR might be practicing longer fallow periods allowing forests to attain higher carbon density and tree species richness and hence better soil nutrient recovery after land abandonment. In conclusion, our results show that the PBSR effectively acted as a stabilizing forest management scheme during the 2012–2021 period, minimizing the impact of productive activities by lowering the frequency of forest clearing events and preserving late secondary forests within the PBSR. We recommend continuing efforts to provide alternative optimal field data collection strategies and modeling techniques to spatially predict key tropical forest attributes. Combining these models with continuous change detection datasets will allow for underlying ecological processes to be revealed and the generation of information better adapted to forest governance scales.

The Puuc Biocultural State Reserve (PBSR is an unique model for tropical dry forest conservation in Mexico. Preserving forest biodiversity and carbon within the PBSR depends on the mainte-nance of low impact productive activities coordinated by multiple communal and private land-owners. In this study, we used state-of-the-art remote sensing data to investigate past spatial patterns in forest clearing dynamics and its relation to forest carbon density and forest plant species richness and diversity in the context of the forest conservation goals of the PBSR. We used a Landsat-based continuous change detection product for the 2000-2021 period and compared it to carbon density and tree species richness models generated from ALOS-2 PALSAR 2 imagery and national scale forest inventory data. The estimated error-adjusted area of detected annual forest clearings from the year 2000 until the year 2021 was 230,511 ha in total (+19,979 ha). The analysis of annual forest clearing frequency and area suggests that although forest clearing was significantly more intensive outside of the PBSR than within the PBSR during the entire 2000-2021 period, there is no evidence suggesting that the frequency and magnitude of forest clearing has changed over the years after the creation of the PBSR in 2011. An emergent hotspot analysis shows, however, that high spatiotemporal clustering of forest clearing events (hotspots) during the 2012-2021 pe-riod was less common than prior to 2011 and these more recent hotspots have been confined to areas outside the PBSR. After comparing forest clearing events to carbon density and tree species richness models, results show that land owners outside the PBSR often clear forests with lower carbon density and species diversity than land owners inside the PBSR. This suggests that, compared to land owners outside the PBSR, land owners within the PBSR might be practicing longer fallow periods allowing forests to attain higher carbon density and tree species richness and hence better soil nutrient recovery after land abandonment. In conclusion, our results show that the PBSR effectively acted as a stabilizing forest management scheme during the 2012-2021 period minimizing the impact of productive activities by lowering the frequency of forest clearing events and preserving late secondary forests within the PBSR. We recommend the continuation of efforts for providing alternative optimal field data collection strategies and modeling techniques to spatially predict key tropical forest attributes. The combination of these models with contin-uous change detection datasets will allow to reveal underlying ecological processes and generate information better adapted to forest governance scales.

Efforts to assess and understand changes in plant diversity and ecosystem functioning focus on the analysis of taxonomic diversity. However, the resilience of ecosystems depends not only on species richness but also on the functions (responses and effects) of species within communities and ecosystems. Therefore, a functional approach is required to estimate functional diversity through functional traits and to model its changes in space and time. This study aims to: (i) assess the accuracy of estimates of species richness and tree functional richness obtained from field data and Sentinel-2 imagery in tropical dry forests of the Yucatan Peninsula; (ii) map and analyze the relationships between these two variables. We calculated species richness and functional richness (from six functional traits) of trees from 87 plots of the National Forest Inventory in a semi-deciduous tropical forest and 107 in a semi-evergreen tropical forest. Species richness and functional richness were mapped using reflectance values, vegetation indices, and texture measurements from Sentinel-2 imagery as explanatory variables. Validation of the models to map these two variables yielded a coefficient of determination (R²) of 0.43 and 0.50, and a mean squared relative error of 25.4% and 48.8%, for tree species richness and functional richness, respectively. For both response variables, the most important explanatory variables were Sentinel-2 texture measurements and spectral bands. Tree species richness and functional richness were positively correlated in both forest types. Bivariate maps showed that 44.9% and 26.5% of the forests studied had high species richness and functional richness values. Our findings highlight the importance of integrating field data and remotely sensed variables for estimating tree species richness and functional richness. In addition, the combination of species richness and functional richness maps presented here is potentially valuable for planning, conservation, and restoration strategies by identifying areas that maximize ecosystem service provisioning, carbon storage, and biodiversity conservation.

For more than three decades, the remote sensing scientific community has successfully generated predictive models of tropical forest attributes and ecological processes at the leaf, canopy, patch and landscape scale by linking field-measured data to remotely sensed spectral values, as well as other variables derived from remotely sensed data. The main interest of these applications is to help describe ecological and functional patterns occurring at larger geographic scales with sufficient accuracy and precision and enable scientists to better understand ecological processes, such as the relationship between atmospheric fluxes, plant structural and ecophysiological traits, soil attributes, anthropogenic use, species occurrence and animal movement. However, as the earth’s environment suffers from ever-increasing human use and abuse, detecting spatiotemporal changes in these variables has become a necessary decision-making tool in conservation action and natural resources’ management. Moving from modeling into the study of soil, plants, wildlife and socioecological processes using remotely sensed data requires the extrapolation of single time-step models to its application on a time series of data with the same expected accuracy. The challenges in this matter are not trivial, since changes in soil moisture conditions, cloud contamination, canopy and leaf-level geometry and physiology can affect the strength of the proposed models. In this context, the term ‘Operationalization’ refers to migration from single time-step models to time series but also refers to the design and implementation of user-friendly tools to increase the efficacy of communicating spatiotemporal trends to the users.