Presentation is loading. Please wait.

Presentation is loading. Please wait.

On the estimation of land surface temperature from AMSR-E measurements Jean-Luc Moncet P. Liang, J. Galantowicz, G. Uymin, A. Lipton, C. Prigent*, B. Hornbuckle**

Similar presentations


Presentation on theme: "On the estimation of land surface temperature from AMSR-E measurements Jean-Luc Moncet P. Liang, J. Galantowicz, G. Uymin, A. Lipton, C. Prigent*, B. Hornbuckle**"— Presentation transcript:

1 On the estimation of land surface temperature from AMSR-E measurements Jean-Luc Moncet P. Liang, J. Galantowicz, G. Uymin, A. Lipton, C. Prigent*, B. Hornbuckle** Atmospheric and Environmental Research, Inc. *LERMA **Iowa State University

2 First version of global monthly average surface emissivities (includes std dev and QC flags) available from: http://www.aer.com/scienceResearch/mwrs/emis.html http://www.aer.com/scienceResearch/mwrs/emis.html 89-11GHz, ~40km resolution AMSR-E land surface emissivity atlas A: Instantaneous emissivity estimate at AMSR-E frequencies using MODIS LST (T sfc_MW ~ T sfc_IR ) B: Arid regions (subsurface penetration) – use 1D thermal model. Surface forcing described by 2 term cosine series expansion with mean temperature T 0. Parameters estimated from Aqua/Terra MODIS LST (amplitude and T 0 ) and AMSR-E/SSM/I 89V Tbs (phase). Depth parameter adjusted to match thermal cycle amplitude at each MW frequency. Emissivity simultaneously adjusted to match T 0. C: Vegetated/frequently cloudy: substitute with A estimates from clearer areas with same surface type - 2 - March 2003

3 Approach Use MODIS LST as reference in derivation of surface emissivity – removes biases between AMSR-E and MODIS in the clear-sky Key advantage: good spatial temporal co-location between the 2 instruments Use clear-sky derived MW surface emissivity to perform MW analysis in cloudy conditions - 3 - 11GHz polarization ratio used to monitor changes in physical surface characteristics Daily outliers removed and flagged (studied separately) Amazon Bamba, Mali

4 11V emissivity standard deviations (July 2003) - 4 - Good consistency between MODIS and AMSR measurements results in stable emissivities AMSR-E/MODIS derived product SSMI/ISCCP LST derived product (from Prigent)

5 Seasonal stability - 5 - AMSR-E Database (emissivities more stable in arid and semi arid areas) SSM/I Database

6 Known issues 19 GHz calibration (~2K bias; Meissner and Wentz, 2010) 89 GHz appears too warm (89-37 GHz emissivity larger than with SSM/I) Unexplained latitudinal dependent bias in 22GHz emissivities (?) New calibration work on going at RSS - 6 - Examples of retrieved emissivity spectra over Amazonian forest

7 Current work Goal: assess usefulness of microwave data (in combination with dynamic surface emissivity atlas) for surface/atmosphere characterization (non-precipitating environment) over land Good knowledge of surface emissivity is necessary but not necessarily sufficient for “useful” atmosphere/surface temperature estimation in retrieval applications (model constraints available in assimilation environment) Focus on 2 parameters: surface temperature and PW Applications: climate (spatial/temporal averaging), IR cloud property (e.g. IWP) retrieval (instantaneous estimates) - 7 - IR LST: No estimate provided under overcast conditions LST estimate only representative of clear portion of the grid box Impacted by undetected (thin) clouds/dust Example of MODIS daily LST product

8 MODIS vs. AMSR-E monthly average diurnal surface temperature differences - 8 - 200307 Monthly mean of LST day/night difference Loose QC Strict QC

9 MODIS vs. AMSR-E monthly average diurnal surface temperature differences - 9 - 200307 Monthly mean of LST day/night difference Loose QC

10 Soil temperature over desert Penetration effects give rise to significant emission temperature gradients (even over rocky areas) Retrieval strategy over deserts consists of assuming known surface emissivity (from 1b algorithm) and retrieve subsurface temperature profiles Makes physical sense over horizontally homogeneous surfaces Enough degrees of freedom to account for thermal and emissivity inhomogeneities ? Surface temporal changes monitored by 11 GHz polarization ratio - 10 -

11 89 -19GHz temperature difference maps - 11 - Positive 89-19 GHz temperature differences may be due to impact of calibration errors

12 0.67 0.55 0.47 um Liquid cloud over deserts Retrieval of liquid water over penetrating surfaces may be difficult Could at least detect presence of liquid clouds from microwave signal Impact of clouds on retrieved Teff(89GHz) – Teff(19GHz) due to: Neglecting CLW in retrieval (and NCEP water vapor errors) Impact of clouds on net surface radiation - 12 - Classified as ice clouds by IR algorithm

13 Other issues Observed polarization differences in retrieved Teff (89GHz) may be indicative of errors in specification of atmospheric term Plans to look at water vapor correction Other? 11-12 um

14 Positive day/night emissivity anomaly in the Midwest Systematic positive day/night differences in our AMSR- E/MODIS emissivity product are observed during the summer months in the Midwest Spatial pattern appears to coincide with corn/soybean crop Are these differences real or artifacts of our process/data? - 14 - JulJun Monitoring corn growing season at 11 GHz

15 Comparison of  DN>0 &  DN  0 regions July-August, 2003 10 GHz  DN>0 (Iowa) 10 GHz  DN  0 (Missouri)  (day): 0.94 – 0.96 & e(night)<e(day) usually  (day)  e(night): 0.94 – 0.96  DN: 0 – 0.04 & v-pol.  h-pol.  DN: -0.02 – 0.01 & v-pol.  h-pol. Polarization ratio (TB H /TB V ): no large differences between regions - 15 -

16 Evidence for emissivity reduction by dew on large-leaf crops (corn/soybean) 1.  DN>0 occurs most days in July-August 2.Nighttime dew at AMSR-E overpass time (~0130) is also persistent 3.  DN>0 region daytime emissivities are consistent with nearby  DN  0 regions   (night) occasionally rises to level of  (day) 4.  DN is independent of polarization & there is little day–night polarization ratio difference  i.e., effect is quasi-polarization-neutral (not due to soil moisture) 5.Effect is strongly associated with mature, large-leaf crops (corn & soybeans)  Ground surface is obscured at 10 GHz  Large, dew-covered leaves may induce scatter-darkening (also seen at 1.4 GHz, Hornbuckle et al., 2007) - 16 -

17 Preliminary analysis with 2009 (SMEX09) Iowa dew field measurements* Nighttime AMSR-E overpass times without detected dew 3 automatic dew sensors (mV output) Sensor disagreement suggests light dew amount Ad hoc “no-dew” algorithm: Any of 3 sensors reporting <280 mV *Experiment conducted by Brian Hornbuckle from U. of Iowa Reasonably good agreement between MODIS LST and in situ air temperatures in the clear-sky (night time) Bias = -0.3K Std dev = 0.77K - 17 -

18 Results Preliminary analysis indicate correlation between  DN anomalies and occurrence of dew deposition on corn leaves - 18 - AMSR-E emissivities derived using in situ air temperatures at night (b) Dew (a) No dew (see previous slide) Night time emissivities X: No dew X: Dew

19 Future plans Continue assessing value of AMSR-E derived surface temperatures (NASA/NEWS) Implement water vapor correction over deserts IR surface temperature prediction over deserts Compare with MODIS/validation Refine QC Snow/RFI flags Increase yield (QC too strict in certain areas) Improve surface classification approach Add dew index Planned improvements Open water correction Process Terra/MODIS over penetrating surfaces Plan to regenerate emissivity database only when new AMSR-E L2A product is available - 19 -


Download ppt "On the estimation of land surface temperature from AMSR-E measurements Jean-Luc Moncet P. Liang, J. Galantowicz, G. Uymin, A. Lipton, C. Prigent*, B. Hornbuckle**"

Similar presentations


Ads by Google