PRESENTATION XU - Copy
Download as PPTX, PDF1 like258 views
This document discusses using MODIS time-series data and eddy-covariance flux data to analyze phenological patterns in boreal forests. The research aims to determine the reliability of using satellite-derived NDVI data compared to ground-based CO2 flux data for identifying the start and end dates of the growing season. The study area is located in Hyytiälä, Finland. 10 years of MODIS NDVI and eddy covariance CO2 flux data from 2003-2012 are analyzed to compare the growing season start and end dates derived from each dataset. The results show some differences in the timing identified by the two methods, highlighting challenges in accurately determining phenological transition dates.
PRESENTATION XU - Copy
- 1. Analysing Phenological Patterns in Boreal Forests Using MODIS Time-Series- Derived & Eddy-Covariance Flux Data Xu Teo MSc Earth Observation & Geoinformation Management
- 3. …and what is phenology? • Plant phenology “..is the study of the timing of recurring biological events in plants...” (Leith, 1974) • Allows us to assess how trees are responding to climate change through timing. • Determined via: • Ground measurements of CO2 • Satellite measurements of Normalised Difference Vegetation Index (NDVI) IntroductionPhenology | Timing CO2 NDVI
- 4. Research Question How reliable are satellite derived vegetation indices (VIs) in observing forest phenological change? IntroductionPhenology | Timing CO2 NDVI
- 5. Aim Ascertain the reliability of using satellite-derived proxy indicators, together with ground-based CO2 flux data, to identify the beginning and end of the growing season as accurately as possible. Objectives Determine the Growing Season Start Date (GSSD) and Growing Season End Date (GSED) via CO2 and NDVI data. Compare both sets of data to obtain a statistical comparison of the viability of using satellite data. IntroductionPhenology | Timing CO2 NDVI
- 7. Hyytiälä, Finland Introduction Methods Phenology | Timing CO2 NDVI
- 9. Eddy Covariance: CO2 Introduction Methods • Measure of turbulent gas fluxes • 10-year dataset: 2003 – 2012 • Altitude: 23 m • Sampled at 30 minute intervals Fine temporal resolution Phenology | Timing CO2 | Eddy Covariance NDVI
- 10. MODIS: NDVI Introduction Methods • Highly correlated to absorbed fraction of photosynthetically active radiation (PAR). • 10-year dataset: 2003 – 2012 • 16-day intervals • Altitude: 700 km • Pixel resolution: 250 m Coarse spatial resolution Phenology | Timing CO2 | Eddy Covariance NDVI | MODIS
- 11. 250 m 250 m Introduction Methods Phenology | Timing CO2 | Eddy Covariance NDVI | MODIS Hypothetical Flux Footprint Tower Pixel 250 m
- 13. Introduction Methods Results 2003 2004 2005 2006 2007
- 14. Introduction Methods Results 2003 2004 2005 2006 2007
- 15. Introduction Methods Results Example Start: 7 April (DoY: 97) Example End: 16 Oct (DoY: 289) Start here? End here?
- 16. Introduction Methods Results Discussion So what’s next? Phenology | Timing CO2 | Eddy Covariance NDVI | MODIS
- 17. What’s next? Further research to define start/end date. Other proxy indicators: • Normalised Difference Water Index • Surface Albedo Phenology | Timing CO2 | Eddy Covariance NDVI | MODIS Introduction Methods Results Discussion
- 18. To conclude Phenology | Timing CO2 | Eddy Covariance NDVI | MODIS Introduction Methods Results Discussion Conclusion • Phenology change timing central to understanding changing climate. • In-situ measurements (EC) and proxy indicators (NDVI) are complementary to each other. • Challenge in identifying transition timing. • Other possible methods available.
Editor's Notes
- #2: Hi everybody! I’d just like to begin my presentation with a very simple analogy. That is if our bodies are like the Earth, then our lungs are like OUR forests. Our lungs help us breath and they keep us alive and healthy. We have probably all heard of this analogy before somewhere. BUT, the problem is this - we probably know a lot less about how forests are responding to climate change than we do about how to treat pneumonia. AND THAT IS A BIG PROBLEM. Therefore, we need to take a closer look at how forests are reacting to an ailing planet and whether or not we can do this from space, because at the end of the day, all these feed into the larger scale of things known as global climate models. And we do this by… // ..Analysing phonological patterns (in this case) boreal forests using MODIS time-series-derived data (which is in space) and eddy covariance flux data (which is on the ground).
- #3: So why am I doing this? And first of all..
- #4: …What is phenology?? To be specific, what is plant phenology? It simply refers to “the study of the timing of recurring biological events in plants” and for a great example of phenology change, go walk through the meadows later this afternoon. And we NEED to study phenology because, just like diagnosing a patient, there is a NEED to u how trees are responding to a warming climate.// Also, it allows us to assess how trees are responding to climate change – are warmer temperatures affecting when they become active?// So looking at their timings – being able to ascertain when it is that they start or stop being active, and how long they remain active for, is vital. And this is determined via ground measurements of CO2 and satellite measurements of the Normalised Difference Vegetation Index or NDVI, which is basically how green vegetation is and is a proxy indicator of forest productivity.
- #5: So the overarching research question here is this – “How reliable are satellite derived vegetation indices (such as the NDVI I was just talking about) in observing forest phonological change?”
- #8: The site that I am looking at is in Hyytiala, Finland – with Sweden to the left and Russia to the right. This is a close-up view of the location of the flux tower and the flux tower itself, which is situated in an area with homogenous cover of Scots pine.
- #9: So, how was the data acquired?
- #10: As I mentioned earlier, firstly we have the ground data. Here, the EC method will be used to measure carbon dioxide gas fluxes in the forest, and is comprised of a 10 year dataset. This picture here on the right is the EC sonic anemometer that does the CO2 measurements. It takes measurements above the forest canopy and is about 23 m high, and samples at 30-minute intervals. Because of this, it gives us fine temporal resolution – keep this in mind.
- #11: MODIS stands for Moderate Resolution Imaging Spectroradiometer, and is the source for the NDVI. NDVI is used because it is highly correlated to the absorbed fraction, of photosynthetically active radiation (PAR) which plants use to carry out photosynthesis. Hence, NDVI is used here as a proxy for photosynthetic activity. We have a 10 year dataset as well and the NDVI product is produced at 16-day intervals, although it must be noted that MODIS obtains measurements everyday and we are using the composite product. It flies at an altitude of about 700 km and its pixel resolution is 250 m. Therefore, we are presented with a measurement which has a very coarse temporal resolution. And, somehow, we have to line them up and find some form of correlation. -So there must be some form of correlation – when the growing season starts, when the growing season ends, and how long the growing season changes, if any, over the years. This is what I am interested in as well.
- #12: So with those 2 chunks of data, this is how the analysis will be done. The red circle represents the hypothetical flux footprint – hypothetical because such a flux footprint is dynamic in nature and changes over the course of the day and especially with respect to the prevailing wind, which in this case, exists at this side. The green square in the middle represents the MODIS NDVI pixel. The flux tower has a footprint of about 250, and the dimensions of NDVI pixel are 250 m by 250 m. So even though the flux footprint is constantly changing and there is a prevailing wind, the MODIS NDVI pixel is well within the footprint. It is impossible to map the pixel to the footprint and, hence, to give it more ‘wiggle room’, I have taken a pixel to the left and right of the middle pixel where the flux tower is in and taken the average NDVI value to obtain a more representative value.
- #13: So then what is the story so far?
- #14: So just as a reminder of where we are again, this site is situated quite far up north and when it comes to winter, photosynthesis is severely inhibited because of the temperature and also snow. I have got a 10 years dataset to work with but I will only be showing 5 years here. So this first graph here shows the temperature fluctuations across 5 years.
- #16: Let’s take a closer look at the GPP and NDVI for 2003. So let’s take an example start and end date. Just to give you a feel of when those are I have put down the exact dates. In between, would be the growing season – and they look.. Reasonably accurate. HOWEVER, how were they… DEFINED?? That start date could very well be there, there or there. And similarly for the end date. In addition, what makes it more problematic is that NDVI is affected by things like So what constitutes a start or end date and how far off is MODIS at seeing this? This bring us back to the research question again of how well satellites can observe such change. Mention the ‘lag’ from the time plants become active till the time the satellite picks it up.
- #18: Find the appropriate methods to precisely define, from flux data, when GSSD and GSED begins and ends, respectively. Most likely from the analysis of slope. Then compare that with MODIS data
- #19: So, to conclude, phenology change TIMING is central to the understanding of a changing climate – one that will affect me and YOU. The use of in-situ and proxy indicators are complementary to each other – we need one, in this case the EC measurements, to validate the other, the NDVI measurements. The biggest challenge, though, is to identify the transition in timing – when a dormant forest suddenly turns active is crucial because at the end of the day these things contribute to global climate models. However, here are other methods out there that have been used before and I will utilise them when I deem necessary for my research work. Thank you.
- #20: So I have included this as an extra slide as I was anticipating a question about flux footprints and its dynamic nature.