Date of Award

May 2019

Degree Type


Degree Name

Master of Science



First Advisor

Mark D Schwartz

Committee Members

Mark D Schwartz, Alison C Donnelly, Changshan Wu


In this project, autumn phenological transition dates and senescence rate (derived from field observation, satellite data and carbon flux measurements) are compared in a northern Wisconsin deciduous forest. Field data cover 2010 and 2012 for the northern site and 2010, 2012 and 2013 for the southern site, with leaf coloration and leaf fall recorded. Satellite indices are EVI and NDVI obtained from the MODIS V006 product via Google Earth Engine platform, covering 2000 to 2017. Carbon flux indices are NEE and GPP covering 1997 to 2017. Field data and normalized satellite data are fitted by a two-section logistic model while carbon data are fitted by a double-logistic model to derive three transition dates and senescence rate parameters. Comparison among these dates and parameters suggests: (a) Generally, the transition dates derived from NDVI is closest to the transitions of leaf coloration and leaf fall; (b) The senescence rate based on NDVI is also closest to the rate of leaf coloration and leaf fall; (c) In year-to-year comparisons, either NEE or GPP can be the least accurate approach in estimating leaf coloration and leaf fall progress; while in long-term comparisons, the accuracy order of EVI, NEE and GPP is variable; and (d) NDVI-based senescence rate is faster, while the senescence rate derived from the other three approaches don’t differ a lot.

Speculations on the reasons for these findings are as follows: (a) canopy senescence is asynchronous, so the timing of first observed leaf coloration from above-canopy and below-canopy can be different; (b) Compared with NDVI, EVI is more sensitive to the subtle canopy change in early autumn and is less affected by soil noise in late autumn, resulting in longer senescence duration; (c) Photosynthesis starts to decrease before visual senescence due to environmental and leaf physiological change, which leads to the bias between field data and carbon data derived transition in early autumn; and (d) The life activities of shrubs and coniferous trees cause carbon exchange to continue changing after deciduous tree senescence terminates.

Included in

Geography Commons