Satellite observations of the microwave emissivity of a semi-arid land surface

https://doi.org/10.1016/S0034-4257(01)00202-4Get rights and content

Abstract

Microwave emissivity is an important parameter for rainfall estimation over land, as well as for atmospheric temperature and humidity retrievals. However, over land surfaces, it varies over a considerable range depending principally on vegetation cover and soil moisture. This study examines the feasibility of estimating emissivity from satellite-based vegetation and moisture indicators for a semiarid region in the African Sahel. Microwave emissivity was calculated from SSM/I observations at 19, 37, and 85 GHz horizontal (H) and vertical (V) polarisation. The technique was validated by comparing the measured emissivity of a sea surface area with the theoretically predicted emissivity. For a dry atmosphere, there was good agreement between theory and measurement. However, the discrepancy was considerably higher in an area where the atmosphere was humid, particularly at 85 GHz. This is attributable to increased uncertainty in atmospheric correction. The land surface emissivity over a 5° square area, which included the Hapex Sahel site, was studied from August to October 1992. The horizontally polarised emissivity eH and polarisation difference measured over dry land areas were found to be well-correlated with Normalised Difference Vegetation Index (NDVI) such that NDVI can be used to estimate pixel eH to within ±0.02. For a wet land surface, there is a general trend for the emissivity to increase with increasing NDVI and for the polarisation difference to decrease. However, the trend is much less well defined than in the dry case. A weak relationship was observed between areal averages of previous day's rainfall (PDR) and emissivity for various vegetation cover classes. A similar relationship was observed with ground-based soil moisture measurements. The results show that emissivity can be estimated with a S.E.<0.015 at 19 GHz from a combination of NDVI and rainfall or soil moisture information.

Introduction

Passive microwave satellite observations have an important role to play in meteorology and climatology. Knowledge of the microwave emissivity of the land surface is essential for the application of microwave data to temperature and humidity sounding (English, 1999), land surface temperature retrievals (Basist, Grody, Peterson, & Williams, 1998), and rainfall estimation over land (Xiang & Smith, 1997).

The research into microwave land surface emissivity described here has been motivated by a need to improve satellite-based rainfall estimates. In semiarid tropical areas such as the African Sahel, real-time rainfall measurements are vital for crop monitoring and drought and famine warning. Ground-based rain gauge networks are usually too sparse and do not provide timely information. On the other hand, satellite images are available in real-time and provide good spatial coverage. Most operational rainfall estimation systems for the tropics are based on thermal infrared (TIR) imagery from Geostationary satellites (e.g., Herman et al., 1997, Thorne et al., 2001). However, these techniques rely on indirect inference of rainfall from cloud top temperatures. Microwave imagery offers the potential for improvement because microwave radiation is strongly influenced by hydrometeors, and thus allows direct observation of rainfall and modelling of rainfall processes. Lower microwave frequencies (less than about 40 GHz) are most sensitive to the rain layer at the cloud base. Enhanced emission due to rain shows up well against the constant low emissivity background of the ocean allowing the retrieval of rainfall using model inversion techniques (Kummerow & Giglio, 1994). Over land, the background signal is both warmer and more variable than that from the ocean making reliable rainfall retrieval considerably more difficult. The variability of the land surface signal is due to variations in microwave emissivity caused by moisture, vegetation, and surface roughness. A major step forward would be the ability to estimate microwave emissivity from remote observations of vegetation and soil moisture. This paper examines the feasibility of making such estimates using satellite data. It follows on from work done using airborne radiometers reported in Morland, Grimes, Dugdale, and Hewison (2000).

The emissivity of a surface at a given wavelength is defined as the ratio of the actual emission of electromagnetic radiation from a surface at that wavelength to the emission, which would be expected from a blackbody at the same temperature.

The relationship between temperature and radiance is approximately linear at microwave wavelengths (the Rayleigh Jeans approximation; Ulaby, Moore, & Fung, 1981). Ignoring atmospheric effects between the surface and the radiometer, To, the brightness temperature of the upwelling surface radiation can be written in terms of Tp, the physical temperature of the surface, Ta, the brightness temperature of the downwelling atmospheric radiation, and e the surface emissivity.To=eTp+(1−e)Tafrom whiche=To−TaTp−Tawhere all variables apart from Tp are a function of frequency ν.

To retrieve surface emissivity from the satellite data, Ta and Tp must be estimated and To as observed by the satellite must be corrected for the effects of the intervening atmosphere. The details of how these corrections were carried out in this study are given in Section 4.

Section snippets

Measurements using ground and aircraft observations

In the frequency range of interest to passive microwave rainfall algorithms (about 20–90 GHz), ground or aircraft-based measurements of surface emissivity have been made by a number of workers (e.g., Calvet et al., 1995, Mätzler, 1994, Morland et al., 2000, Wigneron et al., 1993). The results indicate that soil emissivity decreases with increasing water content and the magnitude of the change is smaller at higher frequencies. The effect of vegetation is relatively small at low frequencies, but

The study area

The area selected for this study was the Hapex Sahel experimental site and surrounding area (0°–5°E and 11°–16°N). Emissivity was calculated for a 92-day period from 1st August to 31st October 1992. This covered the transition from the wet to the dry season, allowing the emissivity to be estimated for different soil moisture conditions. The Hapex Sahel area was chosen because it was the site of an intensive measurement campaign in 1992 and many ground measurements were available. The EPSAT rain

General approach

Emissivity was estimated from the satellite data via Eq. (1). For this calculation, measurements are required of To, Ta, and Tp. Additionally, atmospheric temperature and humidity information is required to correct for atmospheric effects between the radiometer and the surface.

The observed microwave brightness temperature, To, was obtained from SSM/I imagery, the effective atmospheric brightness temperature, Ta, was calculated from the European Centre for Medium Range Weather Forecasting

Vegetation cover and soil moisture information

The main aim of this project is to investigate the dependence of land surface microwave emissivity on vegetation and soil moisture with a view to improving rainfall estimation. The complete spatial coverage of vegetation and soil moisture required for comparison with the emissivity retrievals is only feasible using satellite imagery. In this work, we have used the NDVI to estimate vegetation cover. Unfortunately, no remotely sensed soil moisture measurements are available from a satellite

Dry conditions

For the month of October, when the area was mainly dry, the emissivity data were compared with the appropriate 10-day NDVI composite.

Fig. 5 shows the H polarisation emissivity eH and V–H polarisation difference(eVeH), at 19, 37, and 85 GHz for each pixel plotted against NDVI for 26th October 1992, a dry day. The SSM/I data were acquired during the morning pass. The whole of the 5° square was cloud free, and no rain had fallen since the 24th October 1992 when four stations had reported rainfall.

PDR estimates

In an attempt to establish a more quantitative relationship between soil moisture and emissivity, for each morning pass for which data were available, pixels were binned according to the NDVI classes used in Figs. 7 and 8. The mean and standard deviation of eH were calculated for each NDVI class in each image and compared with the mean PDR for the same set of pixels. The results at 19 GHz for bare soil (0.1≤NDVI≤0.2) are plotted in Fig. 9. Although there is a high degree of scatter, the

Conclusions

A method has been demonstrated for the calculation of microwave emissivity from SSM/I observations, Meteosat TIR data, and the ECMWF analysis of atmospheric temperature and water vapour. The emissivity of the sea surface calculated using this method agreed with theory to within ±0.034 in an area with a dry atmospheric profile. In an area with a moist atmospheric profile, the agreement was poorer, being to within ±0.072 or less apart from the 85-GHz H channel where the discrepancy was as much as

Acknowledgements

The authors gratefully acknowledge the help of Rogerio Bonifacio with the data analysis and interpretation. We would also like to thank the following people and organisations for making data and models available: ECMWF — SSM/I data, atmospheric water, and temperature profile; TAMSAT group (University of Reading) — TIR data; S.J. English (UK Met Office) — microwave atmospheric correction model; K.P. Shine (University of Reading) — infrared atmospheric correction model; G. Dugdale (University of

References (37)

  • B.J Choudhury et al.

    Effect of surface roughness on the microwave emission from soils

    Journal of Geophysical Research

    (1979)
  • Cooper, J. D. (1995). Data from Hapex Sahel experiment on CD-ROM produced by Hapex Sahel information. LERTS, Toulouse,...
  • Dugdale, G. (1995). Data from Hapex Sahel experiment on CD-ROM produced by Hapex Sahel information. LERTS, Toulouse,...
  • S.J English

    Airborne radiometric observations of cloud liquid water emission at 89 and 157 GHz: application to retrieval of liquid water path

    Quarterly Journal of the Royal Meteorological Society

    (1995)
  • S.J English

    Estimation of temperature and humidity profile information from microwave radiances at different surface types

    Journal of Applied Meteorology

    (1999)
  • S.J English et al.

    A fast generic millimeter wavelength emissivity model

    Proceedings of SPIE

    (1998)
  • European Centre for Medium Range Weather Forecasting (1995). MARS — Meteorological Archive and Retrieval System M1.9/2,...
  • M.T Hallikainen et al.

    Microwave dielectric behavior of wet soil: Part I. Empirical models and experimental observations

    IEEE Transactions on Geoscience and Remote Sensing

    (1985)
  • Cited by (0)

    View full text