Security: perspectives of remote sensing - WIT Press€¦ · Security: perspectives of remote sensing O. M. Bucci, ... 3 Optical remote sensing Optical remote sensing has a long heritage
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Security: perspectives of remote sensing
O. M. Bucci, P. A. Brivio, G. Fornaro, R. Lanari & A. Rampini Istituto per il Rilevamento Elettromagnetico dell’Ambiente (IREA-CNR), Consiglio Nazionale delle Ricerche, Italy
Abstract This work aims to describe the perspective on the use of remote sensing in the contest of security, particularly with respect to civil protection issues. Product samples of optical and microwave remote sensing systems are included to give indications about potential applications.
1 Introduction
Earth Observation (EO) satellites were in origin mainly on a military basis and focused on meteorological forecasting and defense issues. However, in the last decades a large number of commercial and science-oriented satellites have been launched to allow measuring parameters associated to physical, chemical and biological processes of our planet. Images of the Earth from space collected by these instruments are nowadays provided on a regular basis to allow detection and monitoring of ground areas. In parallel, population and thus pressure of human activity on the planet are increasing at an unprecedented rate, thus raising the social demand of protection against natural as well as human threats.
Earth monitoring and security play thus a central role in the EU research programs: see for instance the program for “enhancement of the European industrial potential in the field of Security research”, “Global Monitoring for Environment and Security (GMES)” and “Improvement of Risk Management”. As a matter of fact, remote sensing is a powerful technology, at relatively low cost, for the generation of geospatial information, on different scales, with many applications to the risk assessment: vulnerability, prevention and protection as well as response and recovery.
This fact is pushing the research in the remote sensing field towards the development and application of new techniques for civil security issues, not necessarily limited to the protection against natural risks.
With regard to active (microwave) and passive (optical/infrared) remote sensing technologies, the development of new sensors with high resolutions and a huge number of bandwidths, as well as different operative frequencies and modes, and the availability of new technologies for data storage and processing give a larger offer to the application of such systems within the safety contest. Peculiarities and potentialities of optical and microwave remote sensing interest areas such as urban monitoring, crisis management, air quality monitoring, civil protection and rescue supports, fire detection, coastal control, water resources management and flood prevention day-night all-weather monitoring and support to security policies [1,2,3,4]. Pros and cons of current airborne and satellite technologies with respect to their use in security related issues regard essentially coverage, accessibility, surface penetration, operational flexibility and complementariness, resolutions, acquisition continuity and imaging properties.
The paper is organized in two main sections, one addressing remote sensing systems operating at microwave frequencies, in particular Synthetic Aperture Radars (SAR) and the other optical multispectral sensors.
2 Microwave remote sensing
Microwave remote sensing is mainly carried out with SAR systems [5]. Despite working at wavelengths that are almost 10.000 larger than those of optical systems, SAR allows to overcome the corresponding spatial resolution decrease by synthesizing a large aperture exploiting the movement of the small aperture antenna. The price paid for this improvement is the need of a rather huge processing of the received raw data. Nevertheless, based on this somehow “tricky mechanism”, achievable ground resolutions for civil systems from space orbiting at 800Km altitude, such as ASAR/ENVISAT, RADARSAT, are of the order of few meters with very limited antennas size (10m). Future planned systems, by using spot-mode illumination capability (COSMO/SKYMED, SAR Lupe, etc.) will reach resolution as large as the order of one meter or even few tens of centimetres: today such resolution may be achieved only with the use of airborne systems.
Beside the fact that SAR is equipped with an own transmitters, thus allowing day-night operative capabilities, and that it provides complementary information about the observed scene, i.e., the electromagnetic properties of the observed scenes, microwave remote sensing is characterized by the useful property to have reduced interactions with water particles (clouds and fog) thus allowing all-weather monitoring capability. All these features make SAR a unique remote sensing system in the monitoring of environment and in the security area.
Advantages related to cloud penetration are shown in fig.1, that presents a comparison between optical and radar images over the area of the Kliuchevshoi (Russia) volcano acquired during the NASA SIR-C/XSAR mission in 1994.
(a) (b) Figure 1: (a) Optical image of the Kliuchevshoi (Russia) volcano acquired
during the SIR-C/X-SAR mission (copyright JPL). (b) False colour radar image of the Kliuchevshoi (Russia) volcano acquired during the SIR-C/X-SAR mission (copyright JPL).
An additional interesting property of radar systems is associated to the phase
stability of the onboard transmitter, which allows providing a coherent radiation. This fact enables precise phase measurements and thus allows interferometric applications to be implemented; as a consequence, target distances can be accessed with accuracies that compares to the radar wavelength. Based on this fact, SAR interferometry [6] can be implemented by beating images collected by two different antennas that image the scene from slightly different angles. The two antennas can be operated simultaneously onboard a single platform (single pass interferometry) or synthesized by successive passes of a platform with a single antenna (repeat pass interferometry).
Similarly to the principle which allows human eyes pair discriminating the depth of the objects, SAR interferometry allows evaluating the height distribution of the targets with very high accuracy. The Shuttle Radar Topography Mission (SRTM) in 2001 carried out a single pass interferometry data collection, which allowed generating Digital Elevation Model (DEM) with a posting of 30mx30m and 10m height resolution according to DTED-2 standards. Fig.2 shows a DEM from the SRTM mission at X-Band processed by the DLR (Germany).
Differential SAR Interferometry (DInSAR) represents a relatively recent advance in the active microwave remote sensing field that enabled direct application of SAR systems to the risk management area. This technique exploits the knowledge of an external DEM to single out the topographic contribution from repeat pass interferometric data and to measure targets range differences associated to possible ground slow deformations occurring in between the two different acquisition time. In other words, although the radiation makes almost 28.5 million of oscillation in its round trip to from the satellite to a ground radar
target, a small movement of the target on the centimetre scale, may generate a fraction of an additional oscillation that can be reliably detected by the system As a result, deformation maps with an accuracy on the centimetre scale can be generated. Use of multiple, i.e. more than two passes, acquired routinely with subsequent passes (35days temporal separation for ASAR/ENVISAT) allow us to monitor ground targets with a high coverage and extremely high precision (millimetre/year) at a relatively low costs with respect to classical geodetic techniques, such as levelling, or GPS.
Figure 2: DEM from the SRTM mission at X-Band processed by the DLR (Germany).
Figure 3: Mean deformation velocity measured by ERS 1 and 2 satellites over
Application of such technique to environmental monitoring and hazard interest, volcanoes, tectonic movement and earthquakes, slope instabilities, urban areas, water and oil extraction, etc. Fig.3 shows two examples of application of the DInSAR technique to the monitoring of volcanoes. In this example we show how subsidence can be detected and monitored for different phenomena such as subsidence of the Campi Flegrei caldera (left), the Vomero area, and some spots around the volcano. Campi Flegrei and Vesuvio are of volcanology origin whereas the subsidence in the Vomero area is strongly correlated in space and in time with subway excavations.
In conclusion, multitemporal DInSAR certainly represent a first example of how multidimensional information, SAR images acquired at different time, views, frequency and modes can be combined to enable new remote sensing techniques. Research in progress interests issues of superresolution imaging, very accurate DEM reconstruction and 3D focusing to increase imaging capability and therefore target identification and monitoring. Additional strategic lines regard airborne systems where, in addition to continuous-on-demand monitoring, the system allows high flexibility and very high resolution (few centimeters), and potential use of low frequency radar for surface penetration. Such lines may have a very high impact on border and coastal control as well as in after-shocks actions in terms of damage assessment, transportation, search and rescue operations.
3 Optical remote sensing
Optical remote sensing has a long heritage of application in the domain of monitoring of environment and land cover changes, which constitutes the basis for management of resources and risks, and in general for civil security issues, thank to a suite of sensors covering a wide range of spatial, spectral and temporal resolution. In the initial stages of optical satellite remote sensing the data available was from NOAA-AVHRR with 1 Km. Resolution and Landsat Multi Spectral Scanner (MSS) with 80 m resolution. From the 1980s, Landsat Thematic Mapper (TM) imageries with 30 m resolution and SPOT with 20 m resolution became the primary sources of data for monitoring land cover and land cover changes at local scale. In the recent years, with the advent of the new generation science-oriented and commercial satellites, new possibilities are provided in the disaster and security management. The application fields range from damage assessment in the case of floods, earthquake, fire, landslides, oil spill, to civil crisis management, such as refugee camp monitoring and humanitarian intervention planning, and to vulnerability and hazard assessing and risk prevention and mitigation. In the framework of the use if space-based information for disaster and natural resource management, a clear example can be related to water resources.
In the recent years the increased demand for water resources has led to a conflict between human needs and the needs to sustain freshwater ecosystems. On seasonal to annual timescales, the accumulation and melting of snow dominates the hydrological cycle of many alpine and high latitude drainage basins. Over 50% of Eurasia and North America can be seasonally covered by snow [7]. It was estimated that 50% of the annual runoff from a representative basin in Sweden occurs during the spring snowmelt flood in May [8]. In Norway about 50% of the annual precipitation falls as snow, and furthermore Norway produces nearly 100% of its electricity from hydropower [9]. In the Italian Alps most of the reservoirs are characterized by a nivo-pluvial regime and receive about 40-70% of their annual contribution during April-July. About 45 000 billion kWh of hydropower is being generated per annum, which is approximately 25 % of the total energy production in Italy. The prediction of water availability from the snowmelt may be accomplished through the application of hydrological models that use either the temperature index method either the energy balance method. Energy balance methods can account for most of the physical processes involved in the snowmelt, but the input data required by these approaches are rarely available. The degree-day method is a temperature index approach that computes the daily snowmelt depth by multiplying the number of degree-days, i.e. a temperature difference between the daily average air temperature and a base temperature (usually 0°C) by the degree-day factor [10]. An advantage of the degree-day method is that it is easy to use operationally because a limited amount of data is required for the forecast, usually precipitation and temperature.
4.1 Snow Melt Runoff model
In the GLASNOWMAP project (developed in the framework of the ESA DUP 2 programme) the Snow Melt Runoff model (SRM) [11] has been used for calculating daily stream flow. In the SRM the snowmelt contribution to runoff is calculated with the degree-day method using time series of snow cover maps which are derived from satellite data. Satellite remote sensing offers the advantage of providing low-cost, repetitive, multispectral, synoptic and uniform observations over large areas: these spatially distributed observations are in principle more directly linked to snowmelt because this process is extremely variable, temporally and spatially, mainly in areas where the topography is complex, as in mountain range. The basin is subdivided into elevation zones, and the remote observations are used to determine the portion of each elevation zone where the snow cover remains and degree-day approach should be applied to melt the existing snow pack. The Snow Melt Runoff model has been used in more than 20 countries of different continents. As example we recall here the experiments conducted in the European alpine areas: in Switzerland [12], in Austria [13] and in Italy [14].
Remote sensing satellite images are processed to produce snow cover maps and their seasonal variations to derive the snow cover depletion curves (SDC). SDCs represent the day-by-day decreasing of snow cover extent during the melting season. Mean value of daily temperature derived from meteorological station measurements are corrected for the temperature lapse rate in order to represent the temperature of the mean elevation for each zone. Precipitation recorded at meteorological stations includes both snow and rain contribution, whose contributions are separated by means of a critical temperature. Model parameters are fixed during the calibration phase. For real time forecast the input variables requested by SRM model are: precipitation, temperature and snow depletion curve. Temperature and precipitation values must be forecasted for the hydrological basin, concerning with the snow depletion curve the model uses the SDC defined during the calibration. Whenever a new satellite image is acquired and processed, the derived snow cover extension allows to update the percentage values of snow cover for each elevation zone that are used in the model for the next period prediction (days until next image acquisition). This update is realised through the so-called modified depletion curves (MDC).
4.2 Application to Dora Baltea River basin
Study area concerned with the upper part of the Dora Baltea River basin, located in North-Western Italian Alps (Valle d’Aosta). The area of the basin is 1290 km2 and the elevation range is between 619 m. and 4729 m. a.s.l., with an average elevation of 2300 m. Six meteorological stations measuring data on air temperature and precipitation were available: Villeneuve, Morgex, Rhemes Notre Dame, La Thuile, Cogne and Valsavarenche. The highest location of meteorological stations is Valsavarenche (1951 m.) that is lower than the average elevation of the basin. The watershed was subdivided into three elevation zones defined on the basis of the location of meteorological stations in the catchment. Hydrological measurements, i.e. water level, is collected at Aymavilles gauging station (619 m.), and were provided by Regione Valle d’Aosta (Dipartimento Territorio e Ambiente) together with meteorological data. NOAA-16 AVHRR imagery was utilized to map the snow cover extent in the basin. Seven cloud free images were selected to represent the 2002 snow melting season (April-July): two acquisitions in April, one in May and June, and three in July. Best AVHRR image was georeferenced to a reference map, and image to image registration for applied for the other images: RMSE resulted to be less than 0.5 pixel. In fig.4 four of the 7 NOAA AVHRR images used for the model calibration qualitatively show the entity of the snow area reduction during the 2002 snow melting season.
Figure 4: A sample of the set of NOAA AVHRR images used for the model
calibration. Snow cover maps were obtained from satellite images by means of a classification process. Classification accuracy was analyzed through the confusion matrix method: overall accuracy values ranged between 80% to 95% and kappa coefficient of agreement ranged between 0.76 and 0.90. Seasonal variation of snow cover extension for each elevation zone is described by snow depletion curves (SDC) obtained trough the interpolation of values of snow cover percentage in time. Evaluation of the accuracy of the daily river discharges simulated by the model was based on the comparison with measured discharges. Quantitative evaluation was realized computing a set of parameters according to the suggestion of World Meteorological Organization during an inter-comparison study of different hydrological models for snowmelt runoff [15]. Statistical analysis through linear regression between measured and simulated discharges gave a correlation coefficient r=0.89. The standard Nash-Sutcliff coefficient, that quantifies the success of the prediction capabilities of the model, resulted to be R2=0.71 and volumetric difference of –6.0%. Volume underestimation occurs mainly in two short time intervals, at the beginning of April and at the end of July. During the main snowmelt period, May and June, the performance of the model improves: R2 increases from 0.71 to 0.81 and volumetric error reduces to +4.2%.
The forecasting phase has been performed by using ENVISAT AATSR images in the 2003 snowmelting season. As expected results from AVHRR and AATSR are e consistent due the similar spatial and spectral resolution of these sensors. The snow cover maps were extracted from the classification of AATSR images.
The percentage of snow-cover area combined with temperature values was used to update Modified Depletion Curves used in the simulation of the river discharge forecast. The plot of the three days forecasted values is reported in fig.5.
Figure 5: Distribution of measured and forecasted discharges with ENVISAT
AATSR data for Dora Baltea river basin during the 2003 snowmelt season (15 April – 20 June 2003).
Quantitative evaluation was realized computing the set of parameters as the correlation coefficient r, Nash-Sutcliff coefficient R2 and volume deviation (Table I) The quality of the runoff forecast decreases slightly from the one to the three day forecast, in particular when considering the volume deviation that pass from –7 % to –18 %.
Table 1: Qualitative evaluation of discharge forecasting for the period 15 April – 20 June 2003.
Parameters Day 1 forecast Day 2 forecast Day 3 forecast
[1] Gamba P., and F. Dell’Acqua, 2004. Monitoring Urban Areas for Environment and Security through Remote Sensing. ISPRS-04, Istanbul, Turkey.
[2] R. A. Williamson, 2002. Remote Sensing and Transportation Security. ISPRS-02 Commission I Symposium, Denver, CO, USA
[3] J. Lein, and M. Kovacic, 2003 “Developing A Satellite-Based Vulnerability Assessment for Homeland Security Applications”, Papers of the Applied Geography Conferences, 26, pp.133-143
[4] I. Shepherd, and H. J. Lotz-Iwen, 2001 “Application of remote sensing for security needs of the wider European Union – the GMOSS Network of Excellence”, Fifth Pan-European Conference, Netherlands
[5] G. Franceschetti and R. Lanari, 1999, Synthetic Aperture Radar Processing, CRC-PRESS, Boca Raton, FL.
[6] R. Bamler, P. Hartl, Synthetic aperture radar interferometry, Inverse Problems, vol. 14, pp. R1-R54, 1998
[7] Robinson D., K. Dewey and R. Heim, 1993. Global snow cover monitoring: an update. Bull. American Meteorological Society, 74: 1689-1696.
[8] Bergstrom, S. 1979. Spring flood forecasting by conceptual models in Sweden. Proceedings, Workshop on Modelling Snow Cover Runoff. US Army Cold Regions Research and Engineering Laboratory, Hanover, New Hampshire. pp. 397-405.
[9] Winther J. G. and D. K. Hall, 1999. Satellite-derived snow coverage related to hydropower production in Norway: present and future. Int. Journal of Remote Sensing, 20 (15): 2991-3008.
[10] Westertrom G., 1982. Estimating snow cover runoff by the degree-day approach. Vannet Norden, 3: 47-53.
[11] Martinec J., A. Rango and E. Maior, 1983. The Snowmelt Runoff Model (SRM) User's Manual. NASA RP-1100. Greenbelt, Maryland, p. 110.
[12] Seidel K., U. Burkart, R. Baumann and J. Martinec, 1989. Snow cover monitoring by satellites and real time runoff forecasts, IGARSS' 89, Vancouver, R. C., Canada. IEEE, pp.558-561.
[13] Nagler T. and H. Rott, 1999. SAR snow cover retrieval for snowmelt runoff modelling. Earth Observation Quarterly, ESA n.62, 29-31.
[14] Swamy A. N. and P. A. Brivio, 1996. Hydrological modelling of snowmelt in the Italian Alps using visible and infrared remote sensing. Int. Journal of Remote Sensing, 17 (16): 3169-3188.
[15] World Meteorological Organization, 1986. Intercomparison of models for snowmelt runoff modelling. Operational Hydrology Report No. 23. WMO, Geneva, Switzerland.