Land Albedo and Global Warming Validation for RE- Analysis

Transcription

1 CORE-CLIMAX (COordinating earth observation data validation for RE-analysis for CLIMAte ServiceS) Capacity building workshop I March 19-20, 2014 Challenges of surface albedo validation, Examples of European products: GlobAlbedo (ESA-CCI) and CLARA-A1-SAL (CM-SAF) Dr. Terhikki Manninen, FMI Dr. Aku Riihelä, FMI

2 The definition of surface albedo α = I /I Black-sky albedo: Completely direct illumination White-sky albedo: Completely diffuse illumination The albedo under actual illumination is called the blue-sky albedo. It is related to the black- and white-sky albedo through the fraction of direct illumination d: 2 α blue-sky = α black-sky * d + α white-sky * (1-d)

3 Example of challenging validation: black-sky surface albedo 1. The black-sky broadband surface albedo can t be measured at the surface of the globe, because the atmosphere affects the broadband measurements. 2. The white-sky broadband surface albedo can t be measure from the satellite, because the cloud cover masks the surface. The black-sky surface albedo is estimated from clear sky reflectances of satellite images and the white-sky albedo is integrated from that assuming isotropic illumination. The blue-sky albedo is a composite of the black-sky albedo and white-sky albedo. 3

4 Challenges of validation - Accuracy of the validation data - Spatial coverage of validation data - Temporal coverage of validation data - Seasonal coverage of the validation data - Resolution difference between product and validation data - Representativeness of the validation measurement site - Coverage of mountainous areas - No actual validation data available: - validation of a related parameter - comparison of non-identical parameters with estimation of difference 4

5 Challenges of interpretation of validation - The accuracy statistic is given, but - Does it apply to the whole spatial coverage of the product? - Does it apply to the whole temporal coverage of the product? - Does it apply in all seasons? - Does it apply throughout the day? - Does it describe the accuracy of the methodology used or the end product? - Does it apply to all land cover classes? - Does it apply in rough terrain? 5 - Does it apply to the parameter of interest, if only a related parameter is validated? Check what is mentioned as Known issues on web pages or in validation reports and publications!

6 Accuracy of validation data: pyranometers - Instrumental accuracy defines the minimum error - Outdoor pyranometer measurements are challenging: - Ventilation and heating are needed to guarantee 6 all-year operation - Horizontal orientation is crucial for the hemispherical measurement accuracy Kipp & Zonen CM21&CM22 Pyranometer - The upwards looking pyranometer is sensitive to trash, dust etc. accumulation with wind. Also birds may cause trouble. - Regular calibration and maintenance is essential. (Unmanned stations!) - Location of the pyranometer should guarantee unobstructed hemispherical visibility. Difficulties are caused by forests and rough terrain, because of shadowing.

7 Spatial/temporal coverage: BSRN network Continuous measurements 7

8 Spatial/temporal coverage: SNORTEX campaign Individual flights during within an area of about 50 km x 50 km 8 Jean Louis Roujean, Terhikki Manninen, Timo Sukuvaara, Jouni Peltoniemi, Sanna Kaasalainen, Olivier Hautecoeur, Panu Lahtinen, Aku Riihelä, Niilo Siljamo, Milla Lötjönen, Tuure Karjalainen, Anna Kontu, Hanne Suokanerva, Osmo Aulamo, Juha Lemmetyinen, Juha Suomalainen, Teemu Hakala, Harri Kaartinen, Laura Thölix, Outi Meinander and Juha Karhu, 2010, SNORTEX: Remote sensing measurement of snowmelt in European boreal forest, ileaps Newsletter Issue No. 9, April 2010,

9 Example seasonal coverage problem 9 It is difficult to get a statistically sufficient amount of validation data in - areas of high cloud cover fraction: - rain forests - polar areas - areas with typically high aerosol optical depth: - deserts - India and southeastern Asia (for example) - areas with typically low sun elevation: - polar areas and boreal zone - unmanned measurement sites in areas of high risk of human or animal violence to the instrumentation

10 Validation site representativeness - Pyranometer measurement masts: - Typically located in flat terrain surroundings above grass, agricultural fields or barren land. - Very few masts have pyranometers above forest. - Impossible to have continuously measuring instrumentation in wetlands. - Some measurement stations are located in mountainous areas (for example ASRB in Switzerland), but the usage of the data is complicated because of the topography. Comparing with satellite pixels is difficult due to geolocation inaccuracy (half a pixel). 10

11 Validation site representativeness What are we comparing, exactly? Determining site representativeness is quite important for assessing the validation results! 11 pyranometer

12 Coverage of land use classes problem Various types of wetlands are not suited for construction of masts for continuous albedo measurements. Wetlands vary with season markedly. Northern latitudes are full of wetlands. 12 Polunin, O. and Walters M. (1980). A guide to the Vetetation of Britain and Europe. Oxford University Press.

13 Summary - Validation of surface albedo is especially demanding in cases of: - Low sun elevation - High cloud cover fraction - Areas having typically high aerosol optical depth - Snow covered area, especially in periods of rapid changes - Rough terrain - Very heterogeneous land cover - Wetlands 13

14 Examples of known issues: GlobAlbedo product 14

15 Examples of known issues: ESA CCI GlobAlbedo product 15

16 Examples of known issues: ESA CCI GlobAlbedo product Sources for visually observed artifacts: Problems at polar latitudes to fill in gaps due to low sun angles Problems with snow detection. Missing satellite data (cutoff) at latidues >65ºN and <56ºS 16 The artifacts appear in the dataset in the following forms: Box 1. The tile h18v03 does not accurately follow seasonal albedo cycles due to problems with snow detection. The only year unaffected by this artifact is Box 2. A wedge shaped feature does not follow seasonal albedo cycles. This effect is not always apparent, though when present it is extremely obvious. The effect occurs irregularly throughout all GlobAlbedo dataset years ( ). Boxes 3 and 4. Snow detection related artefacts occur commonly throughout the GlobAlbedo dataset.

17 Examples of known issues: ESA CCI GlobAlbedo product Figure above. Polar Regions demonstrate significant irregularities. With commonly occurring missing data, swath artifacts and prior artifacts present in all dataset years. Example errors for Antarctica. 17

18 Related parameter validation: GlobAlbedo product What does blue-sky albedo validation tell about black-sky albedo accuracy? α blue-sky = d α black-sky + (1 - d ) α white-sky = 0.9 * * 0.23 = 0.8 * * = Blue-sky albedo validation of GlobAlbedo product vs. tower measurements. 18

19 Related parameter validation: GlobAlbedo product What does blue-sky albedo validation tell about black-sky albedo accuracy? α blue-sky = d α black-sky + (1 - d ) α white-sky = = = 19 Fraction of diffuse radiation d vs. sun elevation at Sodankylä in spring on a perfectly clear day. The box indicates angles typically used for surface albedo estimation from satellite images.

20 Seasonal coverage validation: GlobAlbedo product Mast measurements are carried out continuously. Satellite passes the site only at certain times. Changing cloud cover may mask the satellite retrievals to not catch the whole variation range of albedo. Blue-sky albedo validation of GlobAlbedo product vs. tower measurements including snow covered season. 20

21 Example land cover representativity: surface albedo Mast measurements may get higher contribution from open area than the satellite in forested region. During snow covered season this has a large effect. Blue-sky albedo validation of GlobAlbedo product vs. tower measurements including snow covered season. 21

22 Examples of known issues: CLARA-A1-SAL product 22

23 Examples of known issues: CLARA-A1-SAL You should be aware of the constant AOD in CLARA-A1-SAL if using the dataset over low latitudes where aerosol loading is high A post-processing correction algorithm for the data is available (for use with your own AOD data) Manninen et al., 2013: Subsequent atmospheric correction of the CLARA-SAL surface albedo time series , EUMETSAT Meteorological Satellite Conference, proceedings For cryospheric high-latitude studies, the dataset should be directly applicable also for trend studies For a more detailed discussion, please see the CLARA-A1-SAL reference paper (Riihelä et al., 2013, ACP) or the CLARA-A1-SAL documentation, available from the CM SAF website 23 An improved AOD correction in CLARA-A2-SAL is under construction!

24 Examples of known issues: CLARA-A1-SAL Mt. Pinatubo June 15th, 1991 Riihelä et al., ACP 13, ,

25 Example of comparison of different parameters: surface albedo What does comparison of blue-sky albedo and black-sky tell? Below simulated black-sky broadband albedo a0bb vs. simulated blue-sky broadband albedo for diverse land cover types (left) and for grass (right). 25 T. Manninen, A. Riihelä and G. de Leeuw, Atmos. Meas. Tech., 5, , Author(s) CC Attribution 3.0 License.

26 Example of comparison of different parameters: surface albedo The difference of the blue-sky albedo α bb and the corresponding black-sky value α 0bb can be estimated, if the sun zenith angle θ z and direct I dir and diffuse irradiance I diff values are known. (α 0bb T. Manninen, A. Riihelä and G. de Leeuw, Atmos. Meas. Tech., 5, , /amt Author(s) CC Attribution 3.0 License. 26

27 Example of topography correction The Alps and Northern Italy, April 15, 2009 The topography correction removes variation of the snow albedo caused by the slopes of the mountains 27 AVHRR RGB

28 Validation information about European albedo products GlobAlbedo: (Documents) GlobAlbedo_FVR_V1.2, Date16/07/13 CLARA-A1-SAL: (Publication&Documentation) CM SAF Cloud, Albedo, Radiation dataset, AVHRR-based Edition 1, Version 1.2, 11 June Geoland SA: (Global/Products/SA/Documents) VALIDATION REPORT SURFACE ALBEDO VERSION 1 (Notice! Copernicus was previously GMES) 28 - Help desks on the www-pages! The data providers are ready to help.

29 Questions? 29

30 Frequently asked questions Validation instrument accuracy? Global representativity Pointwise validation? Satellite instrument aging effect on product accuracy? Rapid phenomena effect on validation? 30