Weakening temperature control on the interannual variations of spring carbon uptake across northern lands

Similar documents
Supplementary Figures

Assimilation of satellite fapar data within the ORCHIDEE biosphere model and its impacts on land surface carbon and energy fluxes

Drylands face potential threat under 2 C global warming target

Supplement of Vegetation greenness and land carbon-flux anomalies associated with climate variations: a focus on the year 2015

Supplementary Information Dynamical proxies of North Atlantic predictability and extremes

A revival of Indian summer monsoon rainfall since 2002

Supporting Information for

Human influence on terrestrial precipitation trends revealed by dynamical

Precipitation patterns alter growth of temperate vegetation

Persistent shift of the Arctic polar vortex towards the Eurasian continent in recent decades

The Northern Hemisphere Sea ice Trends: Regional Features and the Late 1990s Change. Renguang Wu

Lightning as a major driver of recent large fire years in North American boreal forests

Large divergence of satellite and Earth system model estimates of global terrestrial CO 2 fertilization

Velocity of change in vegetation productivity over northern high latitudes

Annex I to Target Area Assessments

particular regional weather extremes

Characteristics of Global Precipitable Water Revealed by COSMIC Measurements

A 3DVAR Land Data Assimilation Scheme: Part 2, Test with ECMWF ERA-40

Display and analysis of weather data from NCDC using ArcGIS

Development of a Tropical Ecological Forecasting Strategy for ENSO Based on the ACME Modeling Framework

Water cycle changes during the past 50 years over the Tibetan Plateau: review and synthesis

SUPPLEMENTARY FIGURES. Figure S1) Monthly mean detrended N 2 O residuals from NOAA/CCGG and NOAA/CATS networks at Barrow, Alaska.

The importance of including variability in climate

SNOWGLACE. Yvan Orsolini 1,2 and Jee-Hoon Jeong 3. Impact of snow initialisation on subseasonal-to-seasonal forecasts

Coupled assimilation of in situ flux measurements and satellite fapar time series within the ORCHIDEE biosphere model: constraints and potentials

Increasing frequency of extremely severe cyclonic storms over the Arabian Sea

Modeling the Arctic Climate System

Vegetation effects on mean daily maximum and minimum surface air temperatures over China

The role of sea-ice in extended range prediction of atmosphere and ocean

SEASONAL AND ANNUAL TRENDS OF AUSTRALIAN MINIMUM/MAXIMUM DAILY TEMPERATURES DURING

CHAPTER 2 DATA AND METHODS. Errors using inadequate data are much less than those using no data at all. Charles Babbage, circa 1850

Top-of-atmosphere radiative forcing affected by brown carbon in the upper troposphere

Assessment of Snow Cover Vulnerability over the Qinghai-Tibetan Plateau

No pause in the increase of hot temperature extremes

Malawi. General Climate. UNDP Climate Change Country Profiles. C. McSweeney 1, M. New 1,2 and G. Lizcano 1

CLIMATE CHANGE AND REGIONAL HYDROLOGY ACROSS THE NORTHEAST US: Evidence of Changes, Model Projections, and Remote Sensing Approaches

ROBUST ASSESSMENT OF THE EXPANSION AND RETREAT OF MEDITERRANEAN CLIMATE IN THE 21 st CENTURY.

The regional distribution characteristics of aerosol optical depth over the Tibetan Plateau

Using Arctic Ocean Color Data in ocean-sea ice-biogeochemistry seasonal forecasting systems

Regional dry-season climate changes due to three decades of Amazonian deforestation

NOTES AND CORRESPONDENCE. Seasonal Variation of the Diurnal Cycle of Rainfall in Southern Contiguous China

The importance of long-term Arctic weather station data for setting the research stage for climate change studies

2013 ATLANTIC HURRICANE SEASON OUTLOOK. June RMS Cat Response

Climate Dataset: Aitik Closure Project. November 28 th & 29 th, 2018

ENSO and April SAT in MSA. This link is critical for our regression analysis where ENSO and

peak half-hourly Tasmania

SUPPLEMENTARY INFORMATION

SUPPLEMENTARY INFORMATION

SUPPORTING INFORMATION. Ecological restoration and its effects on the

Greening of Arctic: Knowledge and Uncertainties

P1.34 MULTISEASONALVALIDATION OF GOES-BASED INSOLATION ESTIMATES. Jason A. Otkin*, Martha C. Anderson*, and John R. Mecikalski #

Training: Climate Change Scenarios for PEI. Training Session April Neil Comer Research Climatologist

SUPPLEMENTARY INFORMATION

Karonga Climate Profile: Full Technical Version

Terrestrial Snow Cover: Properties, Trends, and Feedbacks. Chris Derksen Climate Research Division, ECCC

ENSO: Recent Evolution, Current Status and Predictions. Update prepared by: Climate Prediction Center / NCEP 30 October 2017

Extreme, transient Moisture Transport in the high-latitude North Atlantic sector and Impacts on Sea-ice concentration:

8.1 CHANGES IN CHARACTERISTICS OF UNITED STATES SNOWFALL OVER THE LAST HALF OF THE TWENTIETH CENTURY

Second-Order Draft Chapter 10 IPCC WG1 Fourth Assessment Report

Radiative Climatology of the North Slope of Alaska and the Adjacent Arctic Ocean

Norwegian Earth System Model (NorESM)

peak half-hourly New South Wales

ADVANCES IN EARTH SCIENCE

Grassland Phenology in Different Eco-Geographic Regions over the Tibetan Plateau Jiahua Zhang, Qing Chang, Fengmei Yao

ENSO Cycle: Recent Evolution, Current Status and Predictions. Update prepared by Climate Prediction Center / NCEP 5 August 2013

Stormiest winter on record for Ireland and UK

ENSO Cycle: Recent Evolution, Current Status and Predictions. Update prepared by Climate Prediction Center / NCEP 25 February 2013

SPECIAL PROJECT PROGRESS REPORT

Observed Climate Variability and Change: Evidence and Issues Related to Uncertainty

Climate Feedbacks from ERBE Data

Mozambique. General Climate. UNDP Climate Change Country Profiles. C. McSweeney 1, M. New 1,2 and G. Lizcano 1

Preliminary Research on Grassland Fineclassification

Analysis on Temperature Variation over the Past 55 Years in Guyuan City, China

I T A T I O N H B I T B T V A O C J K M R S A T M O S P H E R E

Dust Storm: An Extreme Climate Event in China

Climate Models and Snow: Projections and Predictions, Decades to Days

Twenty-first-century projections of North Atlantic tropical storms from CMIP5 models

We greatly appreciate the thoughtful comments from the reviewers. According to the reviewer s comments, we revised the original manuscript.

Weather and climate outlooks for crop estimates

The increase of snowfall in Northeast China after the mid 1980s

Climate variability rather than overstocking causes recent large scale cover changes of Tibetan pastures

Global Warming is a Fact of Life

ENSO Cycle: Recent Evolution, Current Status and Predictions. Update prepared by Climate Prediction Center / NCEP 11 November 2013

Short Communication Shifting of frozen ground boundary in response to temperature variations at northern China and Mongolia,

Resuspension and atmospheric transport of radionuclides due to wildfires near the Chernobyl Nuclear Power Plant in 2015: An impact assessment

From El Nino to Atlantic Nino: pathways as seen in the QuikScat winds

Predicting climate extreme events in a user-driven context

Decrease of light rain events in summer associated with a warming environment in China during

Assimilating terrestrial remote sensing data into carbon models: Some issues

Interannual variation of MODIS NDVI in Lake Taihu and its relation to climate in submerged macrophyte region

Arctic sea ice response to atmospheric forcings with varying levels of anthropogenic warming and climate variability

Probabilistic Decision-Making and Weather Assessment

Regional climate change in Tibet: past and future

Canadian Prairie Snow Cover Variability

Northern New England Climate: Past, Present, and Future. Basic Concepts

The Formation of Precipitation Anomaly Patterns during the Developing and Decaying Phases of ENSO

Fine-scale climate projections for Utah from statistical downscaling of global climate models

ENSO Cycle: Recent Evolution, Current Status and Predictions. Update prepared by Climate Prediction Center / NCEP 15 July 2013

Land Data Assimilation at NCEP NLDAS Project Overview, ECMWF HEPEX 2004

Global warming Summary evidence

Transcription:

In the format provided by the authors and unedited. SUPPLEMENTARY INFORMATION DOI: 10.1038/NCLIMATE3277 Weakening temperature control on the interannual variations of spring carbon uptake across northern lands Shilong Piao 1,2,3 *, Zhuo Liu 2, Tao Wang 1,3, Shushi Peng 2, Philippe Ciais 4, Mengtian Huang 2, Anders Ahlstrom 5, John F. Burkhart 6, Frédéric Chevallier 4, Ivan A. Janssens 7, Su-Jong Jeong 8, Xin Lin 4, Jiafu Mao 9, John Miller 10,11, Anwar Mohammat 12, Ranga B. Myneni 13, Josep Peñuelas 14,15, Xiaoying Shi 9, Andreas Stohl 16, Yitong Yao 2, Zaichun Zhu 2 and Pieter P. Tans 10 Ongoing spring warming allows the growing season to begin (hereafter referred to as Barrow) atmospheric measurement station 1 Key Laboratory of Alpine Ecology and Biodiversity, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100085, China. 2 Sino-French Institute for Earth System Science, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China. 3 Center for Excellence in Tibetan Earth Science, Chinese Academy of Sciences, Beijing 100085, China. 4 Laboratoire des Sciences du Climat et de l Environnement, CEA CNRS UVSQ, Gif-sur-Yvette 91191, France. 5 School of Earth, Energy and Environmental Sciences, Stanford University, Stanford, California 94305-2210, USA. 6 Department of Geosciences, University of Oslo, PO Box 1047 Blindem, 0316 Oslo, Norway. 7 Department of Biology, University of Antwerp, Universiteitsplein 1, 2610 Wilrijk, Belgium. 8 School of Environmental Science and Engineering, South University of Science and Technology of China, Shenzhen 518055, China. 9 Climate Change Science Institute and Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA. 10 National Oceanic and Atmospheric Administration Earth Systems Research Laboratory (NOAA/ESRL), 325 Broadway, Boulder, Colorado 80305, USA. 11 Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder 80309, USA. 12 Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, Xinjiang, China. 13 Department of Earth and Environment, Boston University, 675 Commonwealth Avenue, Boston, Massachusetts 02215, USA. 14 CREAF, Cerdanyola del Valles, Barcelona 08193, Catalonia, Spain. 15 CSIC, Global Ecology Unit CREAF-CEAB-CSIC-UAB, Cerdanyola del Valles, Barcelona 08193, Catalonia, Spain. 16 NILU Norwegian Institute for Air Research, PO Box 100, 2027 Kjeller, Norway. *e-mail: slpiao@pku.edu.cn NATURE CLIMATE CHANGE www.nature.com/natureclimatechange 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

Figure S1 A schematic to describe the terms for characterizing spring carbon uptake. We use a smoothed detrended annual cycle of CO2 (the black solid line) for the year 1979 at Barrow. The horizontal dashed line represents the detrended mean CO2 concentration. The two vertical dashed lines indicate the start and end of the spring carbon uptake period (May-June). Spring zero crossing date (SZC) is defined as the day of the year when CO2 crosses down its annual mean level (marked in blue). Spring carbon capture (SCC) is calculated as the seasonal magnitude of the observed CO2 decrease between the first week of May and the last week of June (marked in red).

Figure S2 Temporal evolution of observed spring zero crossing date (SZC), spring carbon capture (SCC) and the average spring (March-June) temperature over the vegetated region north of 50 o N (ST). a, SZC, b, SCC and, c, ST, from 1979 to 2012.

Figure S3 The partial correlation coefficient of a, SZC (RSZC) and b, SCC (RSCC) with preseason temperature using different preseason periods. The preseason is defined as the period before 30 June with different start months varying from November to June. All variables are detrended before the partial correlation analysis. ** indicates statistically significant at the 5% level and * statistically significant at the 10% level.

Figure S4 Same as Figure 2, but showing the frequency distributions of the partial correlation coefficients of observed SZC (RSZC) and SCC (RSCC) at Barrow with spring (March-June) cloudiness (a, b) and precipitation (c, d) during the first 17 years (1979-1995) and the last 17 years (1996-2012).

Figure S5 Frequency distributions of the temperature sensitivity of (a) SZC (γszc) and (b) SCC (γscc) at Barrow during the first 17 years (1979-1995; blue) and during the second, more recent 17 years (1996-2012; red). The temperature sensitivity of SZC (SCC) is calculated as the slope of temperature in a multiple regression of SZC (SCC) against temperature, cloud cover and precipitation during March-June over the vegetated lands north of 50 o N. The frequency distributions of temperature sensitivity are calculated by randomly selecting 14 years during 1979-1995 and 1996-2012. All variables are detrended for each study period before multiple linear regression analysis. Abbreviations of transport simulations are defined in Table 1.

Figure S6 Mean spring footprint for Barrow derived from two different approaches during three time periods. In the left panel, the footprint was derived from the adjoint code of the LMDZ model. In the right panel, the footprint was derived from the Lagrangian particle dispersion model FLEXPART. Note that the FLEXPART simulations are only available from 1985 to 2009.

Figure S7 Frequency distributions of the partial correlation coefficient of observed SZC (RSZC) and SCC (RSCC) at Barrow with spring (March-June) temperature during the first period (1979-1995) and during the second, more recent period (1996-2012). Here climate variables (temperature, precipitation and cloud cover) were computed as the spatial average weighted by the sensitivities (flux sensitivity from LMDZ and potential emission sensitivity from FLEXPART) over the vegetated land area within the mean spring footprint. In a-d, we used the mean spring footprint during the whole study period (1979-2012 for LMDZ and 1985-2009 for FLEXPART, see Fig. S6 c and f). In e-h, we used the mean spring footprint during the two time periods for the first 17 years and the last 17 years (Fig. S6 a, b, d and e).

Figure S8 Same as Figure 2, but for frequency distributions of the partial correlation coefficient of observed SZC (RSZC) and SCC (RSCC) at Barrow with spring (March-June) temperature based on three different climate data sets during the first period (1979-1995) and during the second, more recent period (1996-2012). The climate data sets include Climate Research Unit (CRU, a and b), WATCH Forcing Data methodology applied to ERA-Interim data (WFDEI, c and d) and Climatic Research Unit-National Centers for Environmental Prediction (CRU-NCEP, e and f). The partial correlation coefficient between SZC (SCC) and ST is calculated by statistically controlling for interannual variation in precipitation and radiation (here approximated by cloudiness for CRU) during the period from March to June.

Figure S9 Same as Figure 2, but for frequency distributions of the partial correlation coefficient of observed SZC (RSZC) and SCC (RSCC) at Barrow with preseason temperature during the first period (1979-1995) and during the second, more recent period (1996-2012). We calculate frequency distributions of RSZC and RSCC through randomly selecting 14 years during both 1979-1995 and 1996-2012. For each randomly selected period, the preseason is defined as the period (with 1 month steps) before June for which the negative correlation between SZC and temperature (positive correlation for SCC) was highest (see Methods). Here SZC and SCC were calculated from daily atmospheric CO2 concentration records derived from surface in situ continuous measurements based on three outlier rejection criteria (2.5, 3.0 and 5.0 standard deviations) and two FWHM averaging filters (1.5 and 1.0 month). Frequency distributions of (a) RSZC and (b) RSCC based on the outlier rejection criterion of 5.0 standard deviations and the 1.5 month FWHM averaging filter. Frequency distributions of (c) RSZC and (d) RSCC based on the outlier rejection criterion of 5.0 standard deviations and the 1.0 month FWHM averaging filter. Frequency distributions of RSZC (e) and RSCC (f) based on the outlier rejection criterion of 3.0 standard deviations and the 1.0 month FWHM averaging filter. Frequency distributions of (g) RSZC and (h) RSCC based on the outlier rejection criterion of 2.5 standard deviations and the 1.0 month FWHM averaging filter.

Figure S10 Same as Figure 2, but for frequency distributions of partial correlation coefficient of observed SZC (RSZC) and SCC (RSCC) at Barrow with preseason temperature based on weekly atmospheric CO2 concentration records during the first period (1979-1995) and during the second, more recent period (1996-2012). Here the weekly atmospheric CO2 concentration records were derived from (a, b) surface in situ continuous measurements of SZC and SCC respectively, using the 1.5 month FWHM averaging filter; and (c, d) surface flask samples of SZC and SCC respectively, using the 1.5 month FWHM averaging filter. We calculate frequency distributions of RSZC and RSCC through randomly selecting 14 years during both 1979-1995 and 1996-2012. For each randomly selected period, the preseason is defined as the period (with 1 month steps) before June for which the negative correlation between SZC and temperature (positive correlation for SCC) was highest (see Methods).

Figure S11 Frequency distributions of P value for the partial correlation coefficients of observed SZC (PSZC) and SCC (PSCC) at Barrow with spring (March-June) temperature considering the co-variation in snow water equivalent and previous winter temperature during the first period (1979-1995) and during the second, more recent period (1996-2012). Frequency distributions of P value of the partial correlation coefficient of (a) SZC (PSZC) and (b) SCC (PSCC) with spring temperature after statistically controlling for spring precipitation, spring cloud cover and maximum snow water equivalent from November to June. Frequency distributions of P value of the partial correlation coefficient of (c) SZC (PSZC) and (d) SCC (PSCC) with spring temperature after statistically controlling for spring precipitation, spring cloud cover and winter (November to February) temperature. We calculate frequency distributions of PSZC and PSCC through randomly selecting 14 years during both 1979-1995 and 1996-2012. Note that snow water equivalent data is only available from September 1979. Thus the maximum snow water equivalent from November 1978 to June 1979 was replaced by one value (null) when calculating P value of partial correlation coefficient. Statistically significant P values are marked by the dotted line (magenta: P < 0.05 and brown: P < 0.1). The positive and negative sign indicate the corresponding positive and negative partial correlation coefficients, respectively.

Figure S12 Frequency distributions of P value for the partial correlation coefficients of observed SZC (PSZC) and SCC (PSCC) at Cold Bay (CBA) and Ocean Station M (STM) with preseason temperature over the vegetated lands north of 50 N during the first period (1979-1995) and during the second, more recent period (1996-2012). We calculated frequency distributions of PSZC and PSCC using the 1.5 month FWHM averaging filter for (a, b) CBA and (c, d) STM, respectively. The CO2 data at CBA and STM stations are based on surface flasks sampled on a weekly basis. We calculate frequency distributions of PSZC and PSCC through randomly selecting 14 years during 1979-1995 and 1996-2012. For each randomly selected period, the preseason is defined as the period (with 1 month steps) before June for which the negative correlation between SZC and temperature (positive correlation for SCC) was highest (see Methods). Note that for STM, weekly CO2 records are only available from 1981 to 2009. Thus all missing values were replaced by one value (null) when calculating P value of partial correlation coefficient.

Figure S13 Changes in the partial correlation coefficients of a) SZC (RSZC) and (b) SCC (RSCC) at Barrow with preseason (March-June) temperature over the vegetated lands north of 50 o N during 1979-2012 after applying 15-year moving windows. The partial correlation coefficient RSZC (RSCC) is computed as the correlation between the residuals calculated after regressing SZC (SCC) on precipitation and cloud cover and those after regressing ST on precipitation and cloud cover. Year on the horizontal axis is the central year of the 15-year moving window (e.g., 1986 indicates a moving window from 1979-1993). Solid circles indicate statistically significant partial correlation (P < 0.05), solid squares indicate statistically marginally significant partial correlation (P < 0.1), and hollow circles indicate insignificant partial correlation (P > 0.1). All variables are detrended for each study period before partial correlation analysis. Abbreviations of transport simulations are defined in Table 1.

Figure S14 Anomalies of observed and transport model simulated (a) SZC and (b) SCC at Barrow from 1979 to 2012. Abbreviations of transport simulations are as defined in Table 1. TFTT: simulation with transient global NEE and transient transport; CFTT: simulation with global NEE of year 1979 but transient transport (indicating the effect of wind change on SZC and SCC variability); TFCT: the difference between TFTT and CFTT (indicating the effect of global land carbon flux change on SZC and SCC variability); TFCT-B: the difference between TFTT-B and CFTT (indicating the effect of boreal land carbon flux change on SZC and SCC variability); TFCT-TE: the difference between TFTT-TE and CFTT (indicating the effect of land carbon flux change over temperate regions defined as 30-50 o N on SZC and SCC variability). The coefficient of determination (R 2 ) between observed and transport model simulated SZC (SCC) is given. R 2 = 0.11 and R 2 = 0.08 correspond to the 0.05 and 0.1 significance levels, respectively. All variables are detrended.

Figure S15 Frequency distributions of the partial correlation coefficient of (a) SZC (RSZC) and (b) SCC (RSCC) with March-June temperature during the first 17 years (1979-1995) and the second more recent 17 years (1996-2012). Here SZC and SCC were calculated from the difference of the transport simulation CFTT-Ocean and CFTT (see Methods), by which the impact of ocean flux on RSZC and RSCC can be estimated. Frequency distributions of RSZC and RSCC were calculated as for Fig. 2 in the main text. Statistically significant partial correlation coefficients are indicated as the dashed lines (magenta: P < 0.05 and brown: P < 0.1).

Figure S16 Detrended anomalies of net ecosystem productivity (NEP), net primary productivity (NPP) and heterotrophic respiration (HR) in boreal regions derived from ORCHIDEE simulations (a) S3 and (b) S1. In S3, atmospheric CO2 and all historical climate factors were changed. In S1, only historical temperature was changed. The coefficient of determination (R 2 ) between NEP and NPP/HR is given.

Figure S17 The consistency in the direction of change in the partial correlation coefficient of spring NPP and NDVI with temperature ( R) between NPP and NDVI shown in Fig. 3. The direction of R is shown on the horizontal axis, with the first symbol for NDVI and the second for NPP under different scenarios (CO2+climate/ Only T/ Only T during dormancy period). (+ +) and (- -) indicate a consistent direction of R between NDVI and NPP, whereas (+ -) and (- +) indicate an opposing direction of R. The percentage of all grids over the boreal vegetated area is given on the vertical axis.

Figure S18 Difference in (a) extreme hot days and (b) frost days between 1996-2012 and 1979-1995. The extreme hot days were calculated as the sum of days when daily temperature exceeded the 90th percentile of the temperature during 1979-2012 from 1 March to 30 June. The frost days were calculated as the sum of days when daily minimum air temperature was below 0 from the start of the growing season (SOS) to the summer solstice. Here we determined SOS by taking the ensemble mean of the results from four SOS estimation methods (HANTS-Maximum, Polyfit-Maximum, double logistic and piecewise logistic, see Methods) applied to satellite NDVI data. The dots indicate regions with statistically significant differences (P < 0.05).

Figure S19 Spatial distribution of (a and b) the mean green-up onset dates and (c and d) trends in vegetation green-up dates during 1979-2012. The ORCHIDEE model-derived results are shown in a and c, and the satellite-derived results are shown in b and d. The modelled spring green-up date was estimated based on the seasonal cycle of simulated leaf area index (LAI) following the approach developed by ref 31. The observed spring green-up date was obtained by taking the ensemble mean of the four estimation methods (HANTS-Maximum, Polyfit-Maximum, double logistic and piecewise logistic, see Methods) applied to satellite NDVI data. In the ORCHIDEE simulation, atmospheric CO2 and all historical climate factors are changed. Note that the satellite data are only available from 1982 to 2011.

Figure S20 Same as Figure 2, but with no variables detrended.

Figure S21 The standard deviation (sd) of (a) SZC, (b) SCC, (c) spring temperature (ST), andthe partial correlation coefficient of (d) SZC and (e) SCC with ST during the first 17 years (1979-1995), the second 17 years (1996-2012) and the first 17 years excluding year 1990. All variables are detrended before the sd calculation and partial correlation analysis. The uncertainty is given using 500 bootstrap estimates.