Carbon dynamics seasonality interannual variability and the future

Carbon dynamics: seasonality, interannual variability, and the future under climate change what we are learning from remote sensing, surface measurements, and modeling NASA Carbon Cycle & Ecosystems Workshop University of Maryland, April 28 to May 2, 2008 Amazon Eddy flux tower MODIS sensor on Terra satellite

Carbon dynamics: seasonality, interannual variability, and the future under climate change Scott Saleska, University of Arizona Mike Behrenfeld, Sangram Ganguly, Mike Goulden, Kamel Didan, Mark Friedl, Scott Goetz, Alfredo Huete, Ranga Myneni, Piyachat Ratana, Natalia Restrepo-Coupe, Joellen Russell, Humberto da Rocha, Yosio Shimabukuro, Xiaoyang Zhang Amazon Eddy flux tower MODIS sensor on Terra satellite

Outline 1. Terrestrial Systems a. High latitude trends with climate change b. Tropical seasonality and response to drought 2. Ocean Systems 3. Summary, Outstanding questions, and future work

What is the effect of climate trends on vegetation seasonality and productivity?

Seasonal NDVI amplitude What is the effect of climate trends on vegetation seasonality and productivity? Answer 10 years ago (Myneni et al. , 1997): Consistent INCREASE in seasonal amplitude of satellite-derived vegetation “greenness” in Northern Hemisphere (NDVI from AVHRR) (earlier spring green-up bigger NDVI amplitude)

Seasonal NDVI amplitude What is the effect of climate trends on vegetation seasonality and productivity? Answer 10 years ago (Myneni et al. , 1997): Consistent INCREASE in seasonal amplitude of satellite-derived vegetation “greenness” in Northern Hemisphere (NDVI from AVHRR) (earlier spring green-up bigger NDVI amplitude) Also was consistent with increasing amplitude of atmospheric CO 2 oscillation

What happened since 1991?

Unburned Areas – Photosynthesis Trends Goetz et al. PNAS 2005 See also: Zhou et al. , 2001; Angert et al. , 2005; Ganguly et al. , in review NDVI Changes in Unburned Areas, 1982 -2003

Seasonal NDVI Unburned Areas – Photosynthesis Trends Goetz et al. PNAS 2005 See also: Zhou et al. , 2001; Angert et al. , 2005; Ganguly et al. , in review NDVI Changes in Unburned Areas, 1982 -2003

Seasonal NDVI Unburned Areas – Photosynthesis Trends Pinatubo cooling? Goetz et al. PNAS 2005 See also: Zhou et al. , 2001; Angert et al. , 2005; Ganguly et al. , in review NDVI Changes in Unburned Areas, 1982 -2003

Seasonal NDVI Trend Unburned Areas – Photosynthesis Trends Tundra m ha (%) Unburned Forest Negative 2. 7 ( 4%) 25. 8 (22%) Near Zero 45. 1 (62%) 87. 9 (74%) Positive 24. 5 (34%) 5. 1 ( 4%) Goetz et al. PNAS 2005

“Drier summers cancel out the CO 2 uptake enhancement induced by warmer springs” Angert, et al. (2005), PNAS. Anomaly (ppm/yr or C/yr) 1. 5 Trend toward earlier spring uptake with warming continues post-Pinatubo… 0 -1. 5 1985 1990 1995 2000 1. 5 0 -1. 5 1985

“Drier summers cancel out the CO 2 uptake enhancement induced by warmer springs” Angert, et al. (2005), PNAS. Anomaly (ppm/yr or C/yr) 1. 5 Trend toward earlier spring uptake with warming continues post-Pinatubo… 0 -1. 5 1985 1990 1995 2000 But trend towards increased CO 2 uptake over whole-growing season decouples from warming. 1. 5 0 -1. 5 1985 1990 1995 2000

Effects of 2003 European Heatwave reveal mechanisms consistent with long-term trends (Jolly et al. , 2005) 2003 MODIS summer FPAR (relative to longterm mean)

B. Tropics: What is the seasonality of ecosystem metabolism in Amazônia?

B. Tropics: What is the seasonality of ecosystem metabolism in Amazônia? • Previous consensus answer: photosynthesis and/or transpiration decline in dry seasons: Climate and/or ecosystem models Dickenson & Henderson-Sellars (1988) Nobre et al. (1991) Tian et al. (1998) [TEM] Botta et al. (2002) [ IBIS] Werth & Avissar (2002) [ GISS GCM ] Lee et al. (2005) [ NCAR CLM] But see Potter et al. (1998) modeling study (CASA model)

B. Tropics: What is the seasonality of ecosystem metabolism in Amazônia? • Previous consensus answer: photosynthesis and/or transpiration decline in dry seasons: Climate and/or ecosystem models Dickenson & Henderson-Sellars (1988) Nobre et al. (1991) Tian et al. (1998) [TEM] Botta et al. (2002) [ IBIS] Werth & Avissar (2002) [ GISS GCM ] Lee et al. (2005) [ NCAR CLM] • LBA-Eco produced a suite of evidence suggesting a different picture: Amazonian ecosystems are not water-limited (at least over seasonal timescales) but are driven by available energy and sunlight Results partly anticipated by Potter et al. (1998) modeling study (CASA model)

Measurements across the basin 9 8 A. Manaus, km 34 7 6 5 precip (mm mo-1) 4 11 B. Santarém (km 67) 12 10 10 9 8 8 6 C. Caxiuana 400 900 300 800 200 700 100 600 0 J FMAM J J A SO ND J FMAMJ J A SON D Restrepo-Coupe, in prep. (and Araujo et al. (2002) Manaus) J FMAMJ J A SOND See poster PAR ( mol m-2 s-1) GPP (g. Cm-2 d-1) Eddy Flux towers measuring photosynthesis (GPP)…

Measurements across the basin Remote Sensing (MODIS EVI) Eddy Flux towers… GPP (g. Cm-2 d-1) 9 8 A. Manaus, km 34 7 6 5 precip (mm mo-1) 4 11 B. Santarém (km 67) 12 10 10 9 8 8 6 C. Caxiuana 400 900 300 800 200 700 100 600 0 J FMAM J J A SO ND J FMAMJ J A SOND Huete et al. (2006)

Measurements across the basin Remote Sensing (MODIS EVI) Eddy Flux towers… GPP (g. Cm-2 d-1) 9 8 A. Manaus, km 34 7 6 5 precip (mm mo-1) 4 11 B. Santarém (km 67) 12 10 10 9 8 8 6 C. Caxiuana 400 900 300 800 200 700 100 600 0 J FMAM J J A SO ND J FMAMJ J A SOND Huete et al. (2006) Also see parallel results in LAI seasonality (Myneni et al. , 2007)

The seasonality of forest metabolism: is it linked to the future of the forest under climate change? Forest? . . . or Savanna?

Model-simulated responses of Amazon forest to drought: (U. K. Hadley Center model) Long-term drought (Climate change) Changes in broadleaf tree-cover By 2080: Widespread loss of Amazon forest (Betts et al. 2004) cover (fraction)

Model-simulated responses of Amazon forest to drought: (U. K. Hadley Center model) Long-term drought (Climate change) Short-term drought (e. g. El Nino) Hadley model-predicted GPP & precip in central Amazonia in years relative to El Nino drought Changes in broadleaf tree-cover Forest Photosynthesis (Mg C ha-1 yr-1) El Nino Drought By 2080: Widespread loss of Amazon forest (Jones et al. , 2001) (Betts et al. 2004) Years: -3 -2 -1 0 1 2 3 4 cover (fraction)

Model-simulated responses of Amazon forest to drought: (U. K. Hadley Center model) This prediction is testable! Long-term drought (Climate change) Short-term drought (e. g. El Nino) Hadley model-predicted GPP & precip in central Amazonia in years relative to El Nino drought Changes in broadleaf tree-cover Forest Photosynthesis (Mg C ha-1 yr-1) El Nino Drought By 2080: Widespread loss of Amazon forest (Jones et al. , 2001) (Betts et al. 2004) Years: -3 -2 -1 0 1 2 3 4 cover (fraction)

Observed response to 2005 Amazon drought precipitation anomaly Units: number of standard deviations in 2005 from the longterm mean for the July/Aug/Sept (JAS) quarter. I. e. , for each pixel: Saleska, Didan, Huete, Rocha (2007), Science

Observed response to 2005 Amazon drought precipitation anomaly vegetation “greenness” anomaly Units: number of standard deviations in 2005 from the longterm mean for the July/Aug/Sept (JAS) quarter. I. e. , for each pixel: Saleska, Didan, Huete, Rocha (2007), Science

Observed response to 2005 Amazon drought precipitation anomaly vegetation “greenness” anomaly Short term drought, contrary to model predictions, does not cause photosynthetic slow-down: forests may be adapted to drought, to take advantage of extra sunlight

2. Oceans: How do changes in climate affect ocean productivity?

Climate change will alter ocean phytoplankton Stratified Oceans (low latitude) • Perpetual growing season • Nutrient impoverished surface layer • Inverse relationship b/w temperature and phytoplankton chlorophyll Surface warming low nutrient high light mixed layer restricted vertical exchange low light high nutrient deep layer enhanced nutrient stress decreases growth rates and biomass, shallower mixing increases growth irradiance – all of which decrease chlorophyll levels

Climate change will alter ocean phytoplankton Stratified Oceans (low latitude) • Perpetual growing season • Nutrient impoverished surface layer • Inverse relationship b/w temperature and phytoplankton chlorophyll Surface warming low nutrient high light mixed layer restricted vertical exchange low light high nutrient deep layer enhanced nutrient stress decreases growth rates and biomass, shallower mixing increases growth irradiance – all of which decrease chlorophyll levels Seasonal Seas (high latitude) • Variable growing season • Light seasonally limiting • Nutrients seasonally limiting • Positive relationship b/w temperature and chlorophyll Surface warming nutrient charged low light mixed layer enhanced vertical exchange low light high nutrient deep layer enhanced stratification increases growing season, chlorophyll increases with improved growth rates

Model-based predictions Primary Productivity change Details • Six different coupled climate models • Ocean biological responses to climate warming from industrial revolution to 2050 • Expansion of low production permanently stratified ocean by 4% (N) to 9. 4% (S) • Significant shifts in community composition (Pg C deg-1 y-1) low-lat decreases (stratified ocean)

Model-based predictions Primary Productivity change Details • Six different coupled climate models • Ocean biological responses to climate warming from industrial revolution to 2050 • Expansion of low production permanently stratified ocean by 4% (N) to 9. 4% (S) • Significant shifts in community composition (Pg C deg-1 y-1) High-lat increases low-lat decreases (stratified ocean)

Satellite-based (Sea. Wi. FS) observations Stratified Oceans: 1997 - 2007 El Nino warmth Δ chlorophyll Δ stratification ‘ 98 ‘ 00 ‘ 02 ‘ 04 ‘ 06 Year Stratification anomaly Decrease • Chlorophyll and temperature are inversely related - i. e. , chlorophyll decreases as temperature increases • Temperature-effect not direct • Temperature related to stratification • Stratification influences nutrients & light, which directly effect phytoplankton Increase Behrenfeld et al. (2006) This Region Chlorophyll anomaly (Tg C month-1) Δ chlorophyll Δ SST Temp. anomaly (o. C) Increase I

Satellite-based (Sea. Wi. FS) observations High Latitudes: 1997 - 2007 Δ chlorophyll Δ SST Decrease • Chlorophyll changes in high-latitude north larger than the south • Clear relationships between chlorophyll and temperature High-latitude North Δ chlorophyll Δ SST • In both high latitude regions, overall pattern is decreasing chlorophyll with increasing temperature – this is the opposite of what models predict High-latitude South Decrease Increase Behrenfeld et al. (2006) These Regions Temp. anomaly (o. C) Chlorophyll anomaly (Tg C month-1) Increase

3. Summary, Outstanding Science Questions, and Research Needs

Summary & Outstanding Science Questions 1. In Northern high latitude terrestrial systems: -- 1980 s: earlier springs/more vegetation activity -- 1990 -2000 s: differential response; drought reduces vegetation activity

Summary & Outstanding Science Questions 1. In Northern high latitude terrestrial systems: -- 1980 s: earlier springs/more vegetation activity -- 1990 -2000 s: differential response; drought reduces vegetation activity 2. Questions: -- what caused the slow-down in atmospheric CO 2 after Pinatubo?

Summary & Outstanding Science Questions 1. In Northern high latitude terrestrial systems: -- 1980 s: earlier springs/more vegetation activity -- 1990 -2000 s: differential response; drought reduces vegetation activity 2. Questions: -- what caused the slow-down in atmospheric CO 2 after Pinatubo? 3. 2. In tropical Amazon forests: -- seasonality of ecosystem metabolism driven by available sunlight -- 2005 drought suggests Amazon forests are resilient

Summary & Outstanding Science Questions 1. In Northern high latitude terrestrial systems: -- 1980 s: earlier springs/more vegetation activity -- 1990 -2000 s: differential response; drought reduces vegetation activity 2. Questions: -- what caused the slow-down in atmospheric CO 2 after Pinatubo? 3. 2. In tropical Amazon forests: -- seasonality of ecosystem metabolism driven by available sunlight -- 2005 drought suggests Amazon forests are resilient 4. Question: what are the limits of forest tolerance of drought?

Summary & Outstanding Science Questions 1. In Northern high latitude terrestrial systems: -- 1980 s: earlier springs/more vegetation activity -- 1990 -2000 s: differential response; drought reduces vegetation activity 2. Questions: -- what caused the slow-down in atmospheric CO 2 after Pinatubo? 3. 2. In tropical Amazon forests: -- seasonality of ecosystem metabolism driven by available sunlight -- 2005 drought suggests Amazon forests are resilient 4. Question: what are the limits of forest tolerance of drought? 5. 3. Ocean: declines in productivity (chlorophyl) with increasing temperature, in both low latitude (stratified) and high latitude (seasonal) seas. 6. Question: Why ?

Future Research with a Comprehensive Earth Observation system Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Long-term ground observation network (e. g. Flux. Net plus)

Future Research with a Comprehensive Earth Observation system • Equally long records of satellite and surface for understanding trends Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Long-term ground observation network (e. g. Flux. Net plus)

Example: Does NDVI detect changes in vegetation phenology – or in snow cover? (discussed by Shabonov, et al. , 2002; and Dye & Tucker, 2003)

Example: Does NDVI detect changes in vegetation phenology – or in snow cover? (discussed by Shabonov, et al. , 2002; and Dye & Tucker, 2003) Flux-defined growing season NDVI Daytime maximum uptake mol CO 2 m-2 s-1 Canadian Boreal Forest NDVI Snow-cover defined season Mac. Millan & Goulden (in press)

Example: Does NDVI detect changes in vegetation phenology – or in snow cover? (discussed by Shabonov, et al. , 2002; and Dye & Tucker, 2003) Flux-defined growing season NDVI Daytime maximum uptake mol CO 2 m-2 s-1 Canadian Boreal Forest NDVI Snow-cover defined season Cautious interpretation needed for understanding springtime NDVI increases Emerging long-term (~decadal) flux datasets will help Mac. Millan & Goulden (in press)

Future Research with a Comprehensive Earth Observation system • Equally long records of satellite and surface for understanding trends Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Long-term ground observation network (e. g. Flux. Net plus)

Future Research with a Comprehensive Earth Observation system • Equally long records of satellite and surface for understanding trends • Better understanding of what satellite vs. surface observations measure Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Long-term ground observation network (e. g. Flux. Net plus)

Example: What satellite index best compares to eddy flux -derived GPP? Mae. Klong Tropical Forest, Thailand 2500 0. 6 2000 1500 0. 4 Tower GPP MODIS indices 0. 8 3000 1000 0. 2 Dec Nov Oct Sep Month Aug Jul Jun May Apr Mar Feb 500 Jan -1 -1 GPP (kg C ha mo ) 3500 Huete et al. , (in press) Multiple site tower flux and remote sensing comparisons of tropical forest dynamics in monsoon Asia, Ag. For. Met.

Example: What satellite index best compares to eddy flux -derived GPP? Mae. Klong Tropical Forest, Thailand 2500 0. 6 2000 1500 0. 4 Tower GPP MODIS GPP 1000 (R 2=0. 07) MODIS indices 0. 8 3000 0. 2 Dec Nov Oct Sep Month Aug Jul Jun May Apr Mar Feb 500 Jan -1 -1 GPP (kg C ha mo ) 3500 Huete et al. , (in press) Multiple site tower flux and remote sensing comparisons of tropical forest dynamics in monsoon Asia, Ag. For. Met.

Example: What satellite index best compares to eddy flux -derived GPP? Mae. Klong Tropical Forest, Thailand 2500 0. 6 2000 1500 0. 4 Tower GPP (R 2=0. 07) MODIS FPAR (R 2=0. 01) MODIS GPP 1000 MODIS indices 0. 8 3000 0. 2 Dec Nov Oct Sep Month Aug Jul Jun May Apr Mar Feb 500 Jan -1 -1 GPP (kg C ha mo ) 3500 Huete et al. , (in press) Multiple site tower flux and remote sensing comparisons of tropical forest dynamics in monsoon Asia, Ag. For. Met.

Example: What satellite index best compares to eddy flux -derived GPP? Mae. Klong Tropical Forest, Thailand 2500 0. 6 2000 1500 0. 4 Tower GPP (R 2=0. 01) (R 2=0. 88) Sep Month Aug Jun May Apr Mar Feb Jul MODIS EVI 500 0. 2 Dec MODIS FPAR Nov MODIS GPP Oct 1000 (R 2=0. 07) MODIS indices 0. 8 3000 Jan -1 -1 GPP (kg C ha mo ) 3500 Huete et al. , (in press) Multiple site tower flux and remote sensing comparisons of tropical forest dynamics in monsoon Asia, Ag. For. Met.

Example: What satellite index best compares to eddy flux -derived GPP? GPP (kg. C/ha/mo) South East Asia Tropical forests: Monthly tower GPP vs MODIS EVI across three sites Regression lines from other studies: n A zo a m rd a v t r Ha res Fo

Future Research with a Comprehensive Earth Observation system • Equally long records of satellite and surface for understanding trends • Better understanding of what satellite vs. surface observations measure Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Long-term ground observation network (e. g. Flux. Net plus)

Future Research with a Comprehensive Earth Observation system • Equally long records of satellite and surface for understanding trends • Better understanding of what satellite vs. surface observations measure Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Hyperspectral (high resolution, direct biophysical observations) Hyperspectral Long-term ground observation network (e. g. Flux. Net plus) Hyperspectral

Future Research with a Comprehensive Earth Observation system • Equally long records of satellite and surface for understanding trends • Better understanding of what satellite vs. surface observations measure • Long-term inter-comparability of datasets for trend analysis: MODIS vs. AVHRR (looking backwards) and MODIS vs. VIIRS (looking forward) Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Hyperspectral (high resolution, direct biophysical observations) Hyperspectral Long-term ground observation network (e. g. Flux. Net plus) Hyperspectral

MEASURES project Vegetation Phenology and Vegetation Index Products from Multiple Long Term Satellite Data Records An improved global time series in support of CC&E science Kamel Didan (PI), Jeff Czapla, Mark Friedl, Alfredo Huete, Calli Jenkerson, Willem van Leeuwen, Thomas Maiersperger, Tomoaki Miura, Xiaoyang Zhang http: //phenology. arizona. edu [summer ’ 08] See poster

Future Research with a Comprehensive Earth Observation system • Equally long records of satellite and surface for understanding trends • Better understanding of what satellite vs. surface observations measure • Long-term inter-comparability of datasets for trend analysis: MODIS vs. AVHRR (looking backwards) and MODIS vs. VIIRS (looking forward) Moderate-Resolution remote sensing (AVHRR, MODIS, VIIRS) for comprehensive spatial and temporal coverage Hyperspectral (high resolution, direct biophysical observations) Hyperspectral Long-term ground observation network (e. g. Flux. Net plus) Hyperspectral

Thanks!

Nemani et al. (2003) Increase in CO 2 anomaly

Does NAO set the pace for the biosphere and for growing season CO 2 drawdown? (Joellen Russell & Mike Wallace, 2004) Data series 1980 -2000 (Climate from NCEP, CO 2 from NOAA, NDVI from AVHRR)

Does NAO set the pace for the biosphere and for growing season CO 2 drawdown? The North Atlantic Oscillation (NAO) (Joellen Russell & Mike Wallace, 2004) (Jan-Mar Sea Level Pressure field) Negative anomaly Positive anomaly Regression: Sea-Level Pressure field with growing-season CO 2 drawdown High CO 2 draw-down is associated with high NAO index Data series 1980 -2000 (Climate from NCEP, CO 2 from NOAA, NDVI from AVHRR)

Does NAO set the pace for the biosphere and for growing season CO 2 drawdown? The North Atlantic Oscillation (NAO) (Joellen Russell & Mike Wallace, 2004) Regression: air temperature field vs. NAO Index (Jan-Mar Sea Level Pressure field) Negative anomaly Positive anomaly (when NAO is high, temperature anomalies look like this… Regression: Sea-Level Pressure field with growing-season CO 2 drawdown High CO 2 draw-down is associated with high NAO index Regression: growing season NDVI vs. NAO Index … and NDVI anomalies look like this: Data series 1980 -2000 (Climate from NCEP, CO 2 from NOAA, NDVI from AVHRR)