• shubhi sharma

    Graduate Student | Duke University

    Hi! I am currently a master's student at the Nicholas School of Environment, Duke University. Broadly, I am passionate about building statistical models to answer ecological questions. Over the course of my masters, I've worked on a variety of projects in ecology and statistics. Currently, my primary research focuses on modelling the impacts of climate change on forests in North America. In particular, I am modeling the relationship between forest recruitment, forest structure and climate using a Bayesian hierarchical framework. The aim of this project is to predict the impacts of climate change on recruitment and understand primary drivers of recruitment limitation. Scroll on to learn about my other projects, research interests and academic focus!

  • Experience

    Current and past projects

    Nicholas School of Environment

    Clark Lab

    Duke Statistical Science

    Steorts Lab

    Sep 2019 - Current


    NASA project

    Research Technician

    May 2019 - Current

    Full time


    Nicholas Institute

    Research Assistant

    Oct 2018 - May 2019 

    Part time



    Project Associate

    Jul 2016- May 2018

    Full time



    Project Associate

    Jul 2016- Jul 2017

    Full time


  • Research

    An introduction to past and current projects

    Modelling forest recruitment in North America

    This project aims to investigate the effect of climate and forest structure on forest recruitment using Forest Inventory Analysis (FIA) data. The aims of this analysis are (1) to answer the question of whether there have been observed changes in forest recruitment in the last 40 years, (2) build a model by empirically fitting data using climate and forest structure covariates in order to understand the drivers of change in recruitment (3) to predict change in recruitment under climate change scenarios and finally, to establish the result of a change in recruitment on a forest population.


    (Code & paper not available yet)

    Statistical detection of bi-modality in microtopographic landscapes

    The loss of bi-modality in self-organizing landscapes indicates a loosening of positive and negative feedback loops that govern the formation and regulation of landscape and are ultimately indicative of losing resilience. However, bi-modality detection is contingent on sample size and effect size. To understand the link between successful bi-model detection rate and required statistical power, we computed Monte Carlo approximations of successful proportion under different scenarios of sampling intensity, spatial autocorrelation, sampling error and effect size.


    (Code & paper not available yet)

    Entity resolution for noisy forest data

    Entity resolution or record linkage is the process by which databases with incomplete data records are probabilistically clustered where the same records in multiple databases are clustered together. This record linkage model is a joint random partition model with a downstream task of time series regression on records in a way that allows exact error propagation of the record linkage uncertainty into the downstream task. The downstream task may be generic.


    (Code not available yet)

    Forest species distribution modeling

    Using a species distribution model, we jointly model the 100 most abundant species in North America using climate and abiotic covariates to predict the habitat suitability under climate change scenarios RCP 4.5 and 8.5 for the time periods 2049-2069 and 2079-2099. In addition to predicting individual tree species, we clustered tree species into communities to project community shifts under climate change scenarios.


    Website: https://pbgjam.env.duke.edu/products

    Code :https://github.com/shubhi124081/FIA-Modeling

    Mapping ecosystem services

    With the Nicholas Institute for Policy Solutions, I built two theoretical Ecosystem Service Models for oyster bay restoration projects in North Carolina and mangrove restoration projects in Florida. Apart from conducting workshops and taking multi-stakeholder input, we did extensive literature reviews of the existing research landscape for each "node" in the ecosystem service models. To explore the models and/or go through evidence libraries, click below!


    Website: https://nicholasinstitute.duke.edu/sites/default/files/gems/Evidence_library_061219.pdf

  • Education 

    MEM (Candidate) & B.Sc

    Duke University

    Master of Environmental Management (2018-2020)

    Important coursework:

    ENVIRON710 Environmental Data Analysis

    ENVIRON823 Ecological resilience
    ENVIRON665 Bayesian Inference for Env Models*

    STA602 Bayesian and Modern Statistics

    STA610 Hierarchical Modelling

    STA710 Theory of Causal Inference

    EOS550 Climate & Society

    Durham University

    B.S. in Natural Science (2013-2016)

    Important coursework:

    GEOG1081 Physical geography

    GEOG2651 Geochemistry of the Environment

    GEOG3491 Environmental Processes: Field Case Studies

    GEOG3827 Geochemical applications

    BIOL1171 Genetics

    BIOL2461 Ecology

    BIOL2451 Evolution

    BIOL3541 Global Change Biology

  • Contact me

    To get a copy of my resumé or to contact me for more info