University of Illinois Urbana-Champaign

Data-driven spatial modeling of historic and future land change at global scale

Meiyappan, Prasanth

Content Files
MEIYAPPAN-DISSERTATION-2016.pdf

Permalink

https://hdl.handle.net/2142/90896

Description

Title
Data-driven spatial modeling of historic and future land change at global scale
Author(s)
Meiyappan, Prasanth
Issue Date
2016-04-18
Director of Research (if dissertation) or Advisor (if thesis)
Jain, Atul K.
Doctoral Committee Chair(s)
Jain, Atul K.
Committee Member(s)
  • Nesbitt, Stephen W.
  • Sriver, Ryan L.
  • Dalton, Michael
Department of Study
Atmospheric Sciences
Discipline
Atmospheric Sciences
Degree Granting Institution
University of Illinois at Urbana-Champaign
Degree Name
Ph.D.
Degree Level
Dissertation
Date of Ingest
2016-07-07T21:14:41Z
Keyword(s)
  • land use
  • carbon cycle
  • earth system
  • cropland
  • land use change
  • biogeochemical cycles
Abstract
Assessing the historic and future impacts of land-use and land-cover change (LULCC) on climate requires spatially and temporally explicit data sets on LULCC spanning several decades to centuries, because climate change is a long-term problem. Though remote sensing data provides a globally consistent picture of land cover, these data are only available from the past four decades. Therefore, existing LULCC reconstructions are modeled estimates that combine remote sensing data with relatively coarser-resolution inventory statistics that covers longer historical period. The uncertainties in modeling assumptions, and limited availability and inconsistencies across inventory datasets among other reasons introduce uncertainties in LULCC reconstructions. These uncertainties not only limit our ability to model future LULCC, but also translate as uncertainties in both historic and future environmental assessments. The objectives of my PhD work are as follows: (1) systematically investigate the causes of uncertainties in existing historical LULCC datasets, (2) test the sensitivity of LULCC quantification uncertainty in estimating CO2 emissions from LULCC (historic and future) using a process-based land-surface model, the Integrated Science Assessment Model (ISAM), (3) compare the relative uncertainties from various drivers (e.g. LULCC datasets, model processes e.g. nitrogen cycle, environmental factors such as climate) in estimating historic and future LULCC emissions, and (4) explore statistical techniques to model future LULCC that takes into account the uncertainties in quantifying the spatial and temporal patterns of LULCC, and (5) as a case-study, identify a key regional hotspot of historic LULCC quantification uncertainty (here, India), and reduce uncertainty through improved understanding of the dynamics and drivers of land change in the case-study region. I address the above goals by integrating land-surface modeling (ISAM), remote sensing and GIS, data collected through ground transects, and geospatial data on socioeconomics. ISAM simulations show that the estimated net global emissions from LULCC (mean and range) across three different historical LULCC reconstructions are 1.88 (1.7 to 2.21) GtC/yr for the 1980’s, 1.66 (1.48 to 1.83) GtC/yr for the 1990's, and 1.44 (1.22 to 1.65) for the 2000's. The estimates are higher than other published estimates that range from 0.80 to 1.5 GtC/yr for the 1990’s and 1.1 GtC/yr for the 2000’s. These results are higher than other published estimates because they include the effects of nitrogen limitation on regrowth of forests following wood harvest and agricultural abandonment. The estimated LULUC emissions for the tropics are 0.79±0.25 for the 1980’s, 0.78±0.29 for the 1990’s and 0.71±0.33 GtC/yr for the 2000’s, and for the non-tropics regions are 1.08±0.52, 0.90±0.19 and 0.69±0.12 GtC/yr for the three decades. The model results indicate that failing to account for the nitrogen cycle underestimates LULCC emissions by about 40% globally (0.66 GtC/yr), 10% in the tropics (0.07 GtC/yr) and 70% in the non-tropics (0.59 GtC/yr). If LULCC emissions are higher than assessed, it means fossil fuel emissions would have to be even lower to meet the same mitigation target. Extending ISAM simulations to the 21st century resulted in two key insights. First, nitrogen limitation of CO2 uptake is substantial and sensitive to nitrogen inputs. In ISAM, excluding nitrogen limitation underestimated global total LULUC emissions by 34-52 PgC (~21-29%) during the 20th century and by 128-187 PgC (90-150%) during the 21st century. The difference increases with time because nitrogen limitation will progressively down-regulate the magnitude of CO2 fertilization effect on regrowing forests, due to decreasing supply of plant-usable mineral nitrogen. Second, historically, the indirect effects of anthropogenic activity through environmental changes in land experiencing LULCC (indirect emissions) are small compared to direct effects of anthropogenic LULCC activity (direct emissions). As a result, including or excluding indirect emissions had a minor influence on the estimated total LULUC emissions historically. In contrast, the indirect LULCC emissions for the 21st century are a much larger source to the atmosphere, in simulations with nitrogen limitation. This is because of the gradual weakening of the photosynthetic response to elevated (CO2) caused by nitrogen limitation. Therefore, what fluxes are including in LULCC emissions across different models is a crucial source of uncertainty in future LULCC emissions estimates. A detailed investigation of the sensitivity of different global-scale LULCC modeling techniques show that land use allocation approaches based solely on previous land use history (but disregarding the impact of driving factor), or those based on mechanistically fitting models for the spatial processes of land use change do not reproduce well long-term historical land use patterns. With an example application to the terrestrial carbon cycle, I show that such inaccuracies in land use allocation can translate into significant implications for global environmental assessments. In contrast to previous approaches, I present a statistical land use downscaling model and show that the model can reproduce the broad spatial features of the past 100 years of evolution of cropland and pastureland patterns. Therefore, the modeling approach and its evaluation provide an example that can be useful to the land use, Integrated Assessment, and the Earth system modeling communities.
Graduation Semester
2016-05
Type of Resource
text
Permalink
http://hdl.handle.net/2142/90896
Copyright and License Information
Copyright 2016 Prasanth Meiyappan

Owning Collections

Graduate Dissertations and Theses at Illinois PRIMARY
Graduate Theses and Dissertations at Illinois