A collection of personal projects in ecology, data science, and ecological modeling.
I am an ecologist focused on the impacts of land-use change on ecosystem services, particularly plant-pollinator interactions and plant reproduction. With expertise in statistical analysis, I apply advanced techniques like mixed models and multivariate analyses to explore ecological patterns. I am passionate about integrating data science with ecology to inform conservation and enhance our understanding of ecological dynamics. I'm all about continuous learning and improving my skills.
Victoria Marquez, Argentina
Email: [email protected]
LinkedIn: Victoria Marquez
Local Ecological Knowledge of Forage Plants for Goat Farming and Perceptions about Pollination of Tree Species in the Arid Chaco
Marquez, V., Carbone, L.M., Jimenez-Escobar, D., Britos, H., Aguilar, R., & Zamudio, F.
In this work, I document the goat farming strategies of local peasants, their knowledge about forage plants, and how they perceive pollination of native tree species. Interesting stuff, right? We used semi-structured interviews to gather ethnoecological info and a cognitive approach to learn about the forage plants and their importance. Local peasants shared their insights on the plants they use as fodder. The top forage plants were Neltuma spp., Sarcomphalus mistol, and Castela coccinea, which provide high-quality fruits and leaves at different times of the year. Producers didn’t see pollination as a key factor for forage fruit production but highlighted the importance of climate factors. Understanding this ethnoecological info helps us grasp the peasant management systems that support local communities and play a key role in forest sustainability.
Marquez, V., Garcia Tàcite, J., Wajner, M., Zamudio, F., & Medrano, C.
World compositions have always been multispecies. A solely human world is an impossible purification to find. We're just beginning to explore this field of 'emerging studies.' In this context, we present two examples of ethnographies—one in the rural world of the Cordoba Salt Flats, another in an urban neighborhood of the City of Cordoba. This work isn't just about new methodologies for developing multispecies ethnographies but also tackles the challenge of how to write these journeys. We exercised ways to convey the interspecies relationships lived 'in the field' using techniques from artistic practices. This approach helps us dive into the tangled webs of multispecies interactions and reflect on the use of art to communicate these
Carbone, L.M., Tavella, J., Marquez, V., Ashworth, L., Pausas, J.G., & Aguilar, R.
In this study, we explore the global impact of fires on plant pollination and reproduction. We conducted a systematic review and meta-analyses to investigate how fire influences pollination and plant reproduction. We examined how these responses differ among pollinators, plant regeneration strategies, vegetation types, and biomes. Our findings reveal that fire enhances the overall production of fruits and seeds, though it doesn't necessarily improve reproduction efficiency. My participation in this meta-analyses triggered my interest in conducting global review and synthesis studies and to deepen into more complex analyses
Aguilar, R., Cristóbal-Pérez, E.J., Marquez, V., Carbone, L.M., Paglia, I., Freitas, L., Ashworth, L., Martén-Rodríguez, S., Fernandes, G.W., Lobo, G., Fuchs, E., Quesada, M.
We dive into how human-driven land-use changes are affect pollination and male and female terrestrial flowering plants fitness. As humans occupy most of the Earth's land, it's crucial to understand how these plants reproduce in such environments for their long-term survival and adaptability. We conducted hierarchical and phylogenetically-independent meta-analyses to assess the overall impacts of anthropogenic land-use changes on pollination and plant fitness. We looked at various factors such as habitat loss and fragmentation and their effects on different types of plants and pollinators. Our findings highlight the negative effects of fragmented habitats on plant pollination and fitness, which can reduce recruitment, survival, and long-term viability of plant populations.
Commoning social–ecological networks through the lens of relational ontologies and other economies: How ecologists can diversify their notions of human–non-human relationships
Astegiano, J., Andrieu, J., 2 ; Wajner, M., Marquez, V., Saur Palmieri, V., Torrico Chalabe, J., Massol, F., Calviño, A., & Zamudio, F.
This publication stems from the dynamic discussions and teamwork fostered during multidisciplinary seminars within the Laboratory of Ecological Interactions and Conservation (of which I am a member). Essentially, we propose theoretical frameworks of relational ontologies and alternative economies as tools to diversify conceptions of nature and culture in socio-ecological systems. Fascinating, right? The main challenge of this work was coordinating and aligning with the numerous researchers involved.
Pollination and sexual reproduction of key dominant trees of Arid Chaco under different land-use intensities
Marquez, V., Carbone, M.L., Chiapero, L., Calviño, A.L., Ashworth,, A., Zamudio, & Aguilar, R.
In this study, we evaluated the reproductive success and pollination of two key forage tree species in northern Córdoba province (Neltuma spp and S. mistol). Our main findings highlight the importance of conserving forest cover to maintain the pollination services provided by native bees, thereby ensuring the forage production of these species. The biggest challenge of this work was analyzing the vast amount of numerical data we collected for both species. Ecological data is generally complex, with variables of different natures, often over-dispersed, and with random effects in the explanatory models.
Marquez, V., Wajner, M. & Zamudio, F.
This work is one of my favorites because it was a unique challenge. Unlike other projects, the data here is qualitative and ethnographic, requiring different tools for collection and analysis. We surveyed local practices for raising and selecting livestock guardian dogs in northern Córdoba's farming systems. These dogs are proposed as conservation tools to help mitigate conflicts between wildlife and local producers.
Silvopastoral and peasant management effects on vegetation and soil quality in the arid chaco of central Argentina
Marquez, V., Carbone, M.L., Chiapero, A.L., Ashworth, L., Calviño, A., Zamudio, & Aguilar, R.
This study aims to evaluate the relative effects on soil quality and plant communities of the two most widespread land uses in the arid Chaco (central Argentina). This comparative study is the first of its kind. The main finding highlights increased soil salinity in areas where forests were cleared and replaced with exotic grasses. This is concerning because soil salinization is a difficult environmental issue to reverse, particularly in arid regions worldwide. Monitoring these processes closely is crucial to ensure sustainable livestock production.
Marquez, V., Carbone, L. M., Aguilar, R. & Ashworth, L.
This study explores the impact of frequent fires on the sexual expression and reproduction of Vachellia caven. Our findings reveal that V. caven adapts remarkably well to high fire frequency, maintaining its sexual and reproductive functions even in nutrient-poor soils. Despite the repeated challenges posed by fires, V. caven successfully persists and supports viable populations. This research holds a special place in my heart as it was my first paper and the culmination of my undergraduate thesis.
This project was developed by Dr. Lucas Carbone. You can also find it here: Carbone et al. 2024. RI CONICET. http://hdl.handle.net/11336/218595. I re-ran part of the analysis using Python and improved the code to ensure up-to-date results and reliability.
These analyses are part of my paper titled "Silvopastoral and Peasant Management Effects on Vegetation and Soil Quality in the Arid Chaco of Central Argentina."
To evaluate the similarity in species composition across different land use conditions (silvopastoral and peasant management), we constructed a matrix using Bray-Curtis dissimilarity indices derived from species abundance data. These data were collected from sampling plots and stratified by vegetation layers (herb, shrub, and tree). Based on this matrix, we performed a one-way non-parametric analysis of similarity (ANOSIM) with 999 permutations to test for significant differences in species composition between the two land management types. Additionally, we conducted a Non-Metric Multidimensional Scaling (NMDS) ordination analysis using the calculated dissimilarity measures to visually represent and interpret the differences in species composition associated with each management type
I conduct a phylogenetic tree with all the forage plant species of the Gran Chaco region of Argentina using the U.PhyloMaker R package (Jin & Qian 2023). You can find the data here: Márquez et al., 2024. RI CONICET. http://hdl.handle.net/11336/25093.
The analyses are part of my paper titled "Pollination and sexual reproduction of key dominant trees of Arid Chaco under different land-use intensities"
To investigate the effects of land-use intensity on plant reproductive variables, including fruit set and total fruit production, we utilized generalized linear mixed models (GLMMs). For fruit set, which represents proportional data, I used a binomial error distribution. For total fruit production, which exhibited overdispersion, i applied a negative binomial error distribution. Covariates such as tree diameter and conspecific density within each site were excluded from the models due to violations of the independence assumption required for their inclusion. The significance of fixed effects was assessed using Wald-Z statistics, while the importance of random effects was determined by comparing nested models (with and without random effects) to the global model using likelihood ratio tests (LRTs).
In this code, I conducted an exploratory data analysis (EDA) on a dataset that records air quality across various cities in India (https://www.kaggle.com/datasets/rohanrao/air-quality-data-in-india) using Pandas. The dataset contains measurements of multiple pollutants, including PM2.5, PM10, NO, NH3, CO, benzene, xylene, and the Air Quality Index (AQI). The analysis began with identifying missing values and assessing data types. I then performed transformations such as converting date columns to datetime format and creating new calculated columns. Missing values were imputed with the mean for each city and year, and columns with a high proportion of missing data were removed. Outliers were detected using the Interquartile Range (IQR) method. The aim of these steps was to clean and preprocess the data, making it ready for more in-depth analysis or modeling.
I carried out this project with two friends and colleagues, Santiago Costas and Juan Ignacio Szurlewicz, as our final work for the Data Science diploma from the Faculty of Mathematics, Physics, and Astronomy. Our goal was to predict atmospheric electrical activity (LIGHT column) from the WRF-ELEC model using Machine Learning techniques. First, we explored, cleaned, and standardized the data. Then, we selected explanatory variables using various techniques, such as correlation matrices and PCA. Finally, we conducted multiple iterations of random forest and neural network models, comparing their performance.
This code is part of my final project for the Deep Learning course in the Data Science program at the Faculty of Mathematics, Physics, and Astronomy. It implements a deep learning model (neural network) for classifying sentences as sarcastic or non-sarcastic using PyTorch. The project includes data preprocessing with NLP techniques, such as text cleaning and tokenization. The model is trained and evaluated using classes and objects, ensuring a structured and modular approach to development. The training loop involves optimization, loss calculation, and performance evaluation on a validation dataset. Additionally, the model supports the implementation of early stopping and gradient clipping, which contribute to improved training stability.