Track 2) Quantitative remote sensing of eco-environments and natural resources of tropical, arid, semi-arid regions

Remote sensing is the science of acquiring and analyzing information about objects or phenomena from a distance. Satellite remote sensing is the use of satellite-borne sensors to observe, measure, and record the electromagnetic radiation reflected or emitted by the Earth and its environment for subsequent analysis and extraction of information. The era of satellite remote sensing can be traced back in 1950s, when the former Soviet Union launched Sputnik 1 on 4 October 1957, the world's first artificial satellite (a 22-inch diameter sphere that weighed 83 kg). It circled Earth once every 96 minutes and transmitted radio signals that could be received on Earth providing the first space views of our planet's surface and atmosphere. Since then many remote sensing satellites have been launched and they have gotten bigger and bigger.   On 1 March 2002 the European Space Agency (ESA) launched the ENVIronmental SATellite (ENVISAT), which is the largest Earth-Observation spacecraft ever built (8211 kg), for monitoring of the Earth's land, atmosphere, oceans and ice caps.  Recent trends in remote sensing satellites is to move away from big multi-functional Earth-observing satellites, like ENVISAT, to smaller satellites dedicated to one particular observational task.

One of the our recent research activities is the retrieval of surface bio/geophysical parameters by a number of satellite observation and using of the multisource remote sensing data on the monitoring of surface drought.

Drought is a chronic, potential natural disaster, which is seriously challenging human survival by affecting the social stability, agricultural production and sustainable development of resources and eco-environments. Thanks to its close relationship with some biophysical parameters, such as vegetation indexes (VIs), leaf area index (LAI), albedo, temperature (LST), soil moisture, groundwater and evaportranspiration (ET), remote sensing has been providing a promising tool for accurate mapping of drought phenomenon.

Farmland drought is the most important issue in the drought monitoring. The main objective of our resesarch in this field is to establish simple, effective means for satellite monitoring of surface drought with special attention monitoring of farmland drought. On the basis of summarizing recent advances in drought related research work, regarding to the soil moisture and leaf (canopy) water content, which are the key elements related with crop water stress, there have been several drought and water indices developed. The fundamental theory behind the modeling work is that the target pixel location or the distance between the target pixel and coordinate origin in n-dimensional parameters (reflectance) space is determined by surface water condition, roughness, illumination and viewing geometry etc and, therefore, the distance can provide useful information about the water stress.

• Atmospheric correction for TM/ETM+, MODIS data is carried out by 6S, MODTRAN4.0 incorporating with the in-situ atmospheric profiles, and surface biophysical parameters including VIs, LAI, albedo, LST, soil moisture and groundwater are retrieved.1) Retrieval of surface albedo. Land cover of study area is classified to simulate the spectral data not measured by ETM+ and to reduce both the adjacency and angular effects. The whole shortwave region are divided into 13 spectral bands following by the bandwidths of ETM+ data and the bandwidths not included in ETM+ sensor are simulated by 6S using field measured spectral library. Upwelling and downwelling spectral irradiances at the surface are calculated per band to estimate the spectral albedo and weight of each band. Broadband albedos are estimated and validated with field data. 2) Retrieval of land surface temperature (LST). TM/ETM+ data has only one thermal infrared channel. Atmospheric transmittance, path radiance and single scattering albedo etc are calculated with MODTRAN, effective mean atmospheric temperature, ground emissivity are obtained from ground meteorological observations and image data. Then, TM/ETM+ LST is retrieved with the mono-window algorithm. MODIS LST is retrieved with operational algorithms used in NASA MODIS LST products. 3) Using the TM/ETM+ visible, near infrared data, a model of Groundwater Level Distribution using Remote Sensing (GLDRS), which relates satellite sensor spectral radiance with soil moisture and groundwater level, is developed via in-situ measurement and field examination of soil moisture and groundwater. The results are further improved 6S based atmospheric simulations.

Fig. 1 Sketch map of Ningxia test site in China. + labels represent the ground measuring plots (MODIS image on 6 June 2005)

• Soil moisture is one of the most direct indicators of farmland drought. Crop water stress can be effectively evaluated with the combination of soil moisture content and crop water demand in different phenological dates. Regarding soil moisture, paper develops Perpendicular Drought Index (PDI), a perpendicular distance from the line that intersects the coordinate origin in NIR-Red reflectance space, which has been proven to be quite effective for bare soils. Considering the drought status of farmland conditions and soil spectral behavior, there are two drought indices developed, Modified Perpendicular Drought Index (MPDI) and Normalized Perpendicular Drought Index (NPDI), which takes into account the fraction of green vegetation and soil color.
• Albedo includes the whole information from the visible, near infrared (sensitive to vegetation) and shortwave infrared (sensitive to liquid water). Besides, the spatial resolution of visible, near and shortwave infrared are the same for many sensors, and therefore, unlike the obtaining of the LST-NDVI (resampling causes rather uncertainty), construction of albedo-NDVI spectral space does not need to resample the albedo or NDVI data, so that the maintaining of original information is guaranteed. A new index called Vegetation Condition Albedo Index (VCADI) is designed by analyzing drought behavior in albedo-NDVI spectral space.
• N-dimensional parameters space derived from the combination of albedo, VIs, LAI, LST, soil moisture and groundwater provides a robust tool for drought detection. n-Dimensional Drought Index (NDDI), which is a mathematical model describing the distance from a given initialization water condition to the pixel location in n-space, is developed.
• Regarding canopy water content, the most important indicator of crop water stress, aiming at scientific irrigation and agricultural applications, several methods have been developed using TM/ETM+ NIR and SWIR bands, namely Shortwave infrared Perpendicular water Stress Index (SPSI), Vegetation Water Content Index (VWCI) and Vegetation Water Stress Index (VWSI). A quasi-physical model is established then, for vegetation water content estimation in terms of equivalent water thickness (EWT) or fuel moisture content (FMC) from these water indices with the combination of leaf radiative transfer model PROSPECT, canopy readiative transfer model Lillesaeter/SailH and surface-atmosphere radiative transfer model 6S.
• FMC Guidelines for farmland drought monitoring are established based on field measured FMC over water controlled and stressed winter wheat fields. FMC is related to the dry matter and EWT. It is identified through in-situ measurement data and radiative simulations that the crossing points of dry matter and EWT isolines against VWSI may serve as the indicator of water stress. Ergo, crop water stress can be effectively monitored with VWSI derived from satellite data, neither the calculating of the EWT and FMC nor the field data is required.
• Finally, comparison between drought indexes developed is carried out at micro-scale (TM/ETM ＋) and macro-scale (MODIS) applications. The robustness of the indexes are tested over different time and different eco-systems with field observed data.

To conclude, the drought methods are highly concordant with in-situ data, provide robust tools for remote estimation of soil moisture, vegetation water content and crop water stress. 1) NDDI and VCADI may show the greatest precision providing that the surface bio-physical factors are retrieved with high accuracy; 2) PDI, MPDI must be the first choice if a quick and simple method with higher accuracy is of interest; 3) comparing the PDI and MPDI, both shows similar performance for sparsely vegetated surfaces or bare soils, however, MPDI is much better than PDI in the drought assessment of agricultural fields or dense vegetation surfaces; 4) drought indices (SPSI, VWSI) based on vegetation water content indicate the greatest potential for farmland drought.

Fig. 3 Spatial distribution of drought and drought mapping

Fig. 4 VCADI response on rainfall. MODIS derived VCADI data are compared aginast rainfall.