H31L-01 INVITED
Comparison of a Global Landslide Event Inventory to a Satellite-based Landslide Algorithm
A global, satellite-based landslide algorithm has been developed using surface information and multi-satellite rainfall data. The technique integrates surface parameters such as slope, land cover, soils, and elevation with satellite precipitation data to obtain an estimate of areas susceptible to landslides in near-real time. This research compares the predictions from the global landslide algorithm run retrospectively for individual years with global landslide inventories to assess both the relative skill of the technique and the value of currently available landslide information on a global scale. Results indicate that the general pattern of landslide activity (number of total events, geographic distribution, etc.) can be reproduced, but finer-scale distributions and individual events are difficult to match between the forecast and the event inventory. Preliminary results indicate that three-fourths of the landslide events correspond to locations with high susceptibility values based on the satellite-based Landslide Susceptibility Index map. Probability of Detection and False Alarm Rate statistics are presented for the global database, with results varying based on the size of area used for event validation. Results are also shown to be a function of population density with more densely populated areas having higher scores, as expected. This global algorithm represents the first phase in identifying landslide hazards at this scale. With adjustment, the algorithm shows great promise in approaching landslide hazard assessment globally and providing information for the research community to address landslide issues in a broader context. The evaluation also provides insight into the necessary considerations and potential adaptations to the algorithm for improved landslide hazard forecasting on a global scale and the need for international efforts for developing accurate landslide inventories.
H31L-02
The use of remote sensing for landslide hazard management in the aftermath of the Kashmir earthquake
The ML=7.6 Kashmir earthquake of 8th October 2005 triggered large numbers of landslides, with the distribution being concentrated on the hanging wall block in the vicinity of the fault rupture. These landslides were responsible for an estimated 25,500 deaths, but also caused serious problems for the distribution of aid in the immediate aftermath of the earthquake event. In the longer term period after the earthquake, slope instability represents a very major problem for the authorities in Pakistan, primarily because many slopes appear to have been left in a quasi-stable state. As the area affected by landslides is so large, remote sensing represents the only reasonable way to assess the occurrence of landslides and the threats posed by future events. In this study, satellite imagery has been used for two purposes. First, several epochs of imagery have been used to examine the Hattian Bala landslide, which is a 68 million m3 rock avalanche that destroyed three villages and killed around 1000 people. The landslide deposit has blocked two valleys to a depth of about 130 m. Until the completion of a spillway in 2007, an outburst flood threatened a major settlement 3 km downstream. The images have allowed analysis of the landslide deposit itself and of the lakes, permitting estimates of the magnitude of a potential outburst flood. Interestingly, the satellite images also revealed clusters of landslides in the source area of the landslide prior to the earthquake. Furthermore, the images have allowed identification of a creeping landslide on the slopes above the lake, which represents a serious ongoing threat. In the second application, mapping has been undertaken of the landslide distribution in general. This has demonstrated that, contrary to other reports, many of the landslides now observed in the landscape in fact predate the earthquake. For example, in the valley containing the Hattian Bala landslide, 94 landslides can now be mapped. However, 76% of these are observable on imagery collected before the earthquake, meaning that just 24% were first time failures triggered by the event. The data have allowed the construction of landslide magnitude-frequency distribution plots, providing the basis for the assessment of the contribution of seismically triggered events as compared to background rates of activity. The study uses a combination of remote sensing tools, including Landsat 5, Landsat 7, Quickbird and TOPSAT. The latter is a new micro-satellite system that has been designed and built by a consortium led by Qinetiq Ltd to provide low cost, high resolution images of the Earth. TOPSAT has four spectral bands, consisting of a pan band with a spatial resolution of 2.5 m and RGB bands with a spatial resolution of 5.0 m. The image size is 17 km x 17 km. A key feature of TOPSAT is the potential for the satellite to deliver data directly to a mobile ground station immediately after an has been obtained. As such it is potentially a useful tool for the determination of the impact of natural disasters such as floods, landslides and earthquakes. A brief analysis is presented of the strengths and weaknesses of TOPSAT in this application when compared to the other systems.
H31L-03
An InSAR-based survey of surface subsidence hazard in Iran
Land-surface subsidence due to over-extraction of groundwater has been long recognized as a potential problem in many areas that have undergone extensive groundwater development. In Iran, decades of unrestrained groundwater extraction for domestic, agricultural, and industrial use have resulted in a precipitous depletion of this valuable resource. Here we use satellite radar interferometry (InSAR) and show that the decline in groundwater levels is associated with land-surface deformation on local and regional scales. We have utilized Envisat ASAR data and created interferograms imaging 13 plain aquifers and valleys in Iran. We infer 6 major sources of groundwater-induced deformation showing up to 10 – 50 cm/yr of subsidence. Temporal analysis of InSAR data show no evidence of major seasonal rebound due to seasonally fluctuating groundwater levels, possibly implying that inelastic compaction is the dominant deformation process at water reservoirs.
H31L-04
Radar Interferometry Measures Underground Salt Leaching of Anthropogenic Origin in a Saharan Oilfield
Remote sensing by satellite radar interferometry can identify and monitor subsidence caused by underground disturbances of anthropogenic origin, such as oil or water pumping or mine collapse. Here, we report on a case of ground subsidence at an oil field in Algeria (Ouargla area) following a well accident. In 1986, a 250 m-wide ground collapse appeared at the abandoned OKN32 oil-well near the town of Ouargla in the Sahara desert. It is hypothesized that a break in the well casing allowed for deep artesian water to reach a subsurface salt layer, which led to salt dissolution, the formation of an underground cavity, and intense pollution of a shallow aquifer used by the local population for irrigation and drinking. We have 22 ERS images spanning from 1992 to 2000 to investigate the temporal and spatial evolution of the ground subsidence associated with this process. In particular, we aim at quantifying the size of the underground cavity and whether or not it is still continuing to grow. Although no information was available on the location of the accident (or on any of its characteristics), we found a circular subsidence area clearly correlated with the OKN32 oil field. Its diameter spreads from 900 m in 1993 to about 1300 m in 1996, with a total subsidence of up to 20 cm over that time period. The subsidence rate decreases gradually from 1993 to 1996 and slows to a stop around that date, suggesting that the process has healed at depth. Preliminary calculations indicate that a volume of the order of 80,000 cubic meters of salt have been leached and transported to the shallow aquifer.
H31L-05 INVITED
Integrating earth observations and model results provides earlier Famine Early Warning
Remote sensing allows us to detect slowly evolving natural hazards such as agricultural drought. Famine early warning systems transform this data into actionable policy information, enabling humanitarian organizations to respond in a timely and appropriate manner. These life saving responses are increasingly important. In 2006, 1 out of 8 people did not have enough to eat, 22 million more people became undernourished, and 22 countries provided 6.5 billion dollars in food aid. The motivation is strong, therefore, to increase the effectiveness of every dollar of food aid provided, ensuring that the assistance arrives sufficiently early to ward off human and economic catastrophe. Properly interpreted remote sensing information reduces the influence of politics in determining the amount and location of aid delivered. In this talk we will review three recent contributions that earth observations have provided to famine early warning: trend identification, increasingly accurate forecasts of food security conditions, and enhanced integration of biophysical and socio-economic data.
H31L-06
Relative Spectral Mixture Analysis for monitoring natural hazards that impact vegetation cover: the importance of the nonphotosynthetic fraction in understanding landscape response to drought, fire, and hurricane damage
Remote sensing provides a unique ability to monitor natural hazards that impact vegetation hydrologically. Here, the use of a new multitemporal remote sensing technique that employs free, coarse multispectral remote sensing data is demonstrated in monitoring short- and long-term drought, fire occurrence and recovery, and damage to hurricane-related mangrove ecosystems and subsequent recovery of these systems. The new technique, relative spectral mixture analysis (RSMA), provides information about the nonphotosynthetic fraction (nonphotosynthetic vegetation plus litter) of ground cover in addition to the green vegetation fraction. In some cases, RSMA even provides an improved ability to monitor changes in the green fraction compared to traditional vegetation indices or standard remote sensing products. In arid and semiarid regions, the nonphotosynthetic fraction can vary on an annual basis significantly more than the green fraction and is thus perfectly suited for monitoring drought in these regions. Mortality of evergreen trees due to long-term drought also shows up strongly in the nonphotosynthetic fraction as green vegetation is replaced by dry needles and bare trunks. The response of the nonphotosynthetic fraction to fire is significantly different from that of drought because of the combustion of nonphotosynthetic material. Finally, damage to mangrove ecosystems from hurricane damage, and their subsequent recovery, is readily observable in both the green and nonphotosynthetic fractions as estimated by RSMA.
H31L-07
Mapping Pollution Plumes in Areas Impacted by Hurricane Katrina With Imaging Spectroscopy
New Orleans endured flooding on a massive scale subsequent to Hurricane Katrina in August of 2005. Contaminant plumes were noticeable in satellite images of the city in the days following flooding. Many of these plumes were caused by oil, gasoline, and diesel that leaked from inundated vehicles, gas stations, and refineries. News reports also suggested that the flood waters were contaminated with sewage from breached pipes. Effluent plumes such as these pose a potential health hazard to humans and wildlife in the aftermath of hurricanes and potentially from other catastrophic events (e.g., earthquakes, shipping accidents, chemical spills, and terrorist attacks). While the extent of effluent plumes can be gauged with synthetic aperture radar and broad- band visible-infrared images (Rykhus, 2005) (e.g., Radarsat and Landsat ETM+) the composition of the plumes could not be determined. These instruments lack the spectral resolution necessary to do chemical identification. Imaging spectroscopy may help solve this problem. Over 60 flight lines of NASA Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) data were collected over New Orleans, the Mississippi Delta, and the Gulf Coast from one to two weeks after Katrina while the contaminated water was being pumped out of flooded areas. These data provide a unique opportunity to test if imaging spectrometer data can be used to identify the chemistry of these flood-related plumes. Many chemicals have unique spectral signatures in the ultraviolet to near-infrared range (0.2 - 2.5 microns) that can be used as fingerprints for their identification. We are particularly interested in detecting thin films of oil, gasoline, diesel, and raw sewage suspended on or in water. If these materials can be successfully differentiated in the lab then we will use spectral-shape matching algorithms to look for their spectral signatures in the AVIRIS data collected over New Orleans and other areas impacted by Katrina. If imaging spectroscopy can be used to identify plume composition on a regional scale than this information would help emergency personnel prioritize evacuations, help government agencies formulate cleanup strategies, and help ecologists assess the potential damage to wetlands and wildlife. This work could be the start of a new application of hyperspectral data for world-wide monitoring of spills from space-based imaging spectrometers. AVIRIS data used to test our method were corrected for solar flux, atmospheric absorptions, and scattering using the Atmospheric CORrection Now (ACORN) radiative transfer algorithm and residual artifacts were removed using ground spectra of a concrete runway at the Gulfport Airport in Mississippi. The resulting apparent reflectance data were mapped for spectral signatures of pollution plumes and results will be presented.
H31L-08
Comparison of Change Detection Techniques for Assessing Hurricane Katrina-Induced Damage to Forests
This study compared performance of four change detection algorithms with six vegetation indices derived from Landsat Thematic Mapper (TM) imagery taken pre\- and post\-Hurricane Katrina. The overall goal of the study was to select an optimal remote sensing approach for identifying disturbed forests by the hurricane in the Lower Pearl River Valley, USA. The algorithms included univariate image differencing (UID), selective principal component analysis (selective PCA), change vector analysis (CVA), and post-classification comparison (PCC). The indices consisted of near-infrared to red ratios (RVI), normalized difference vegetation index (NDVI), Tasseled Cap index of greenness (TCG), brightness (TCB) and wetness (TCW), and soil-adjusted vegetation index (SAVI). In addition to the satellite imagery, "ground truth" data of forest damage were also collected through field investigation and interpretation of post\-Katrina aerial photos. Disturbed forests were identified by classifying the composite and the continuous change imagery with the supervised classification method. Results showed that the change detection techniques largely affected the results with an overall detection accuracy varying between 59% and 86% and with a Kappa Statistics ranging from 0.15 to 0.72. Detected areas of disturbed forests were noticeable in two groups: 186,625 \- 264,617 ha and 106,783 \- 124,205 ha. The PCC algorithm along with the composite image contributed the highest accuracy and lowest errors (0.5%) in estimating disturbed forested land areas. Both UID and CVA performed similarly, but caution should be taken when using selective PCA in detecting hurricane disturbance to forests. Among the six indices, TCW outperformed the other indices owing to its maximum sensitivity to the forest modification. This study suggests that compared with the detection algorithms, proper selection of vegetation indices is more critical for obtaining a satisfactory result.