Detection of surface heterogeneity in eddy covariance data

Size: px

Start display at page:

Download "Detection of surface heterogeneity in eddy covariance data"

Nicholas Holt
5 years ago
Views:

Hauf 1 1 Institute for Meteorology and Climatology, Leibniz University

1 Detection of surface heterogeneity in eddy covariance data R. Deckert 1,2 and T. Hauf 1 1 Institute for Meteorology and Climatology, Leibniz University Hannover, Germany 2 now at: German Aerospace Centre, Institute of Atmospheric Physics, Oberpfaffenhofen, Germany EGU General Assembly, 23 April 2012

2 Introduction Eddy covariance method Fast sampling of w and transported quantity c Vertical turbulent flux via covariance: w ' c ' Usually half-hour interval Here: w ' ' and w ' T ' Slide 2

3 Introduction Eddy covariance method Fast sampling of w and transported quantity c Vertical turbulent flux via covariance: w ' c ' Usually half-hour interval Here: w ' ' and w ' T ' Slide 3

4 Introduction Eddy covariance method Fast sampling of w and transported quantity c Vertical turbulent flux via covariance: w ' c ' Usually half-hour interval Here: w ' ' and w ' T ' Interpretation in terms of surface exchange Quality issues (ecosystem research: Mahrt 2010; Vickers et al. 2010; Vesala et al. 2008; Foken et al. 2005;...) Here: focus on stationarity Slide 4

5 Introduction Instationarity 1 Atmospheric trends / variability Slide 5

6 Introduction Instationarity 1 Atmospheric trends / variability 2 Surface heterogeneity + wind-direction fluctuations grass mast (2.5m) ete concr Slide 6

7 Introduction Instationarity 1 Atmospheric trends / variability 2 Surface heterogeneity + wind-direction fluctuations source area dir d win n o i t ec Slide 7

8 Introduction Instationarity 1 Atmospheric trends / variability 2 Surface heterogeneity + wind-direction fluctuations Slide 8

9 Introduction Instationarity 1 Atmospheric trends / variability 2 Surface heterogeneity + wind-direction fluctuations (water resources, agriculture, roads,...) Slide 9

10 Introduction Instationarity 1 Atmospheric trends / variability 2 Surface heterogeneity + wind-direction fluctuations Presenting a data-based detection scheme Slide 10

11 Method Fetch ~ location of source area fetch fetc h Slide 11

12 Method Association of w and c with fetch Slide 12

13 Method Association of w and c with fetch Median fetch defines two bins 50% 50% Slide 13

14 Method Association of w and c with fetch Median fetch defines two bins Separate calculation: and 50% 50% Slide 14

15 Method Statistical inference: Slide 15

16 Method Statistical inference: Autocorrelation keep individual blocks together Repeat a lot of times: 1 Random permutation of color 2 Binning and flux calculation blocks Slide 16

17 Campaign Airport in northern Germany trees building grass c tarma ete r c n o c mast (2.5 m) N Slide 17

18 Campaign Airport in northern Germany Acceptance sector 40 half-hour data sets Spring, daytime, dry conditions trees building grass c tarma ete r c n o c mast (2.5 m) N Slide 18

19 Results Hypothesis test H0: H1: two sided Slide 19

20 Results Hypothesis test H0: H1: two sided Pooled inference Weighted least squares (generalized) / weighted bootstrap Assumptions ok p= p=0.1 Slide 20

21 Results p= p=0.1 Slide 21

22 Results Low test power? p= p=0.1 Slide 22

23 Results Low test power? p= No physical effect? p=0.1 Slide 23

24 Results Decomposition: Low test power? p= No physical effect? p=0.1 Slide 24

25 Results Roughness contrast physically plausible p<< Evapo-transpiration contrast physically plausible p< Slide 25

26 Results Equal sign Roughness contrast physically plausible p<< Evapo-transpiration contrast physically plausible p< Slide 26

27 Results But: scatter from correlation coefficient Roughness contrast physically plausible Evapo-transpiration contrast physically plausible p=0.008 p<< p< Slide 27

28 Results Roughness contrast physically plausible p<< Temperature contrast physically plausible p< Slide 28

29 Results Opposite sign Roughness contrast physically plausible p<< Temperature contrast physically plausible p< Slide 29

30 Results In addition: scatter from correlation coefficient Roughness contrast physically plausible Temperature contrast physically plausible p= p<< p< Slide 30

31 Summary Eddy covariance method Quality issue instationarity Here: surface heterogeneity + wind-direction fluctuations Detection scheme Fetch-based binning Statistical inference - Uses random permutation - Accounts for auto-correlation Test: paved surface grassland Results agree with physical setting Individual / pooled tests match But: correlation coefficients Slide 31

32 Slide 32

33 Slide 33

34 Slide 34

A scheme for assessment of quality and uncertainty for long-term eddy-covariance measurements

A scheme for assessment of quality and uncertainty for long-term eddy-covariance measurements Clemens Drüe 3, M. Mauder 1, M. Cuntz 2, A. Graf 4, C. Rebmann 2, H. P. Schmid 1, M. Schmidt 4, and R. Steinbrecher