Anthony C. Janetos and Jake Brunner (World Resources Institute, Washington, DC)
There is general agreement about the issues for which the twin technologies of remote sensing and geographic information systems are well suited. Applications of these technologies for understanding the extent and location of natural resources and human infrastructure, and the change in these resources over time are among the best examples. Remote sensing and GIS are currently being used in some countries to augment more traditional methods of determining forest and agricultural areas, understanding patterns of geomorphology, and tracking changes in urban areas, the location of concessions, etc.
Remote sensing developed primarily from a scientific basis, rather than a strict resource management basis, and scientific applications of these technologies remain a vibrant area of progress. The international scientific community has made enormous strides in the past decade in producing a variety of global and regional datasets on the distribution of land-cover types, forest resources, the identification of human settlements, fire occurrence, etc. In addition, there is beginning to be scientific capacity for using remote sensing to parameterize models of ecosystem processes, such as net primary productivity, that are of both fundamental scientific interest and of practical importance for resource management.
Government and private agencies which have responsibilities for emergency response have also found value in the use of remote sensing and GIS. The rapid notification of ephemeral, but potentially catastrophic environmental events such as large wildfires and severe storms, can be extremely useful in terms of warning the general populace about impending danger, and also can assist relief agencies in targeting areas for special attention when disaster strikes.
In this paper, we examine some of the policy, pricing, and institutional issues that have hampered the use of remote sensing and GIS, and draw from our experience some recommendations for moving forward. In particular, we use the US experience with the Landsat system, and how it has evolved over time, as a model for the potential role of government, universities, and the private sector in the further dissemination of these technologies.
One way to think about the roles that policies and institutions might play to be successful in enhancing the use of remote sensing and GIS for managing natural resources is to define the characteristics of a successful system. This is relatively easy to do, even in the absence of specifying all the technical details of implementation.
The most important management goal for effective natural resource management is to build in feedback loops about the extent and condition of the natural resources themselves to those actors who are responsible for managing the resources. If the management agent in question is a government ministry, or a private landowner, the principle remains the same. Accurate information about the status, extent, and condition of the resource itself is enormously valuable. The information’s value increases if it is comparable to information about the same resource in other places, e.g. if data about forests in one region are comparable to data about forests in another region, and if it is comparable over time, e.g. measurements of forest extent in 1980 are directly comparable to measurements of forest extent today.
In general, we can identify five characteristics of a successful system for using remote sensing and GIS data for natural resource management:
The system should be rapid. This may seem obvious, but has been difficult to achieve in many places for technical reasons. Standard forest inventory practices in many countries, for example in the US, require a decade to do a complete national inventory. Some portions of that inventory, in particular the determination of forest area in particular regions, can be accomplished much more rapidly with the use of remotely sensed information. Changes in year-to-year status of the resource cannot be reliably detected and understood unless the monitoring methods can keep up with the actual management practices on the ground. Even the use of remote sensing has proven problematic in this respect because of the difficulty in acquiring reasonable data quickly enough, and then the subsequent difficulties in analyzing data rapidly enough. However, this is an area in which the use of remotely sensed data has enormous potential for some, but not all aspects of natural resource management.
The system should be spatially informative. Monitoring systems that provide single national numbers for resources, such as the total amount of forest cover, can be useful for some scientific applications. However, they are of extremely limited utility for resource management. Practical resource management requires information about specific places, with reasonable resolution. Obviously, the use of remotely sensed data and GIS have important advantages in this regard. However, it is clear that resource management typically requires fairly high spatial resolution, on order of a few tens of meters. Of current remote sensing systems, Landsat and SPOT are those that are typically the most useful for management applications. These satellite systems also have the advantage of having had years of operations and large communities of users familiar with the potential uses of the data. While synthetic aperture radar systems have great potential for monitoring some aspects of ecosystems, many of the analytical algorithms are still highly experimental.
The system should have clear and transparent methods. There is enormous public benefit for resource management agencies, and even private landowners, from having clear and transparent monitoring methods. Transparency enables comparison of results from different regions, and can have the effect of enhancing the public support for the agencies providing information. Achieving transparency of methods is all the more important because of the technical nature of using remote sensing imagery; it is unrealistic to think that all research institutes, or all natural resource decision makers, are going to understand the details of all the steps in the analysis. However, it is not unrealistic for them to understand the broad design of a monitoring system that can identify land-cover of different types, show that the remote sensing data can be validated in a sufficient number of places with in situ observations, and then be used to monitor change over time.
The system should have replicable analyses. Quantifying changes over time are a crucial element of any resource management system. The analyses that are done with remote sensing data must be replicable, so that a consistently produced time series of data can be produced as easily as possible. It is only through the analysis of time series data that the actual performance of natural resource management policies and practices can be understood and changed if need be.
The system should be inexpensive. The differences between the initial capital costs of initiating a monitoring system using remote sensing and GIS technologies and the ongoing operational costs of maintaining it are important to differentiate. There are very few nations that are capable of bearing the cost of developing, launching and operating remote sensing systems. However, the costs of acquiring and interpreting data can be relatively modest, and these are the costs to consider for new national efforts.
There are a number of basic reasons for the continued difficulty in using remote sensing and GIS technologies in basic resource management. In this section of the paper, we identify the main challenges, and how the experience with the Landsat system provides useful models.
Science. The use of digital remotely sensed data still does not have much more than a 30 year history, even in the US. The first Landsat mission was launched in 1972, and even though there was scientific research with aircraft instruments and meteorological satellites before that, there has still been only a few "generations" of scientists and resource managers with direct experience with these measurements. So in many ways, the basic physics of the instrumentation, the basic biology, ecology, chemistry, and geology of the Earth’s surface have all required lots of investment within the research community to develop tools and applications for broader use. However, much of this investment has occurred. There is always a need for more and better science, to be sure. But much of the scientific understanding that is needed for the application of these technologies to natural resource management now has a fairly strong foundation. While we should expect that the science will continue to progress, there is certainly now enough understanding of the basics to feel comfortable with extending the practical applications of the technologies into management arenas. The Landsat 7 mission emphasizes within its own science team the development of new methods for land-cover change research and resource management.
Acquisition of Data. In the early stages of the Landsat program, the main challenges to the actual acquisition of the data were technological. The sensor design was novel, communications technology with ground receiving stations was untested, and the experimental nature of the mission did not result in a scientifically grounded acquisition policy. Rather, there was a notion that the primary mission of the system was to map the surface of the Earth in an exploratory mode. With the decision by the US to commercialize the Landsat system in 1980, the problems of data acquisition quickly became economic. Essentially the business model became, pay for the data and it will be acquired. One of the problems with this model is that it assumes that the data are of immediate value. In fact, the value of these data for natural resource management increases over time as a time series accumulates. The potential uses of the data also accumulate as scientific knowledge advances. It is difficult for the operators of a satellite system to incorporate these features into a purely commercial model. These constraints have changed dramatically with the Landsat 7 system, which returns Landsat to the public domain for the first time in nearly 20 years. In particular, the acquisition policy has been designed with a fundamentally scientific basis of understanding the changes in the surface cover of the Earth at seasonal to annual time scales. The acquisition policy and design of the Landsat 7 system are meant to provide for a refreshed global archive of data held in the US, supplemented with holdings in the international ground receiving stations. The global archive is meant to be refreshed seasonally. The acquisition strategy is further optimized to ensure that difficult-to-obtain scenes, such as those over often cloud-covered tropical forest, are given priority in the queue. These characteristics of the Landsat 7 acquisition strategy are immediately commensurate with the needs of the natural resource management community.
Information Policy. The primary issues about information policy are who owns the information, who controls access to it, and what are the rules governing its use. In the US, there are several operating philosophies that have governed the use of remotely sensed data. One is a general scientific philosophy, common throughout the world. Information and analyses should be published, and data shared broadly. This enables peers to review each other’s work for quality, other research groups to reproduce results and gain confidence in them, and improvements to occur in a largely open forum. The second philosophy is the appropriate role of the government. In the US, there is a strong tradition that the government should avoid activities that compete with the private sector in markets in which a viable private sector exists. There is also an accompanying corollary that the government should take steps to promote the development of a viable private sector where one does not exist. The role of the university community in this respect is both interesting and important, since they operate mid-way between government institutions and corporate institutions. The third operative philosophy in the US system is that information collected at public expense should in general be made available to the public at no additional cost, or at a small marginal cost (sufficient to cover expenses). The experience within the US is that adherence to these policies as general operating principles greatly expands the use of remotely sensed data, providing other obstacles aren’t too great. The changeover of Landsat 7 from a privately operated system to a publicly operated system has resulted in policies that do not discriminate between for-profit and not-for-profit users. They also allow the free copying and redistribution of data that have already been purchased by any party. In fact, since the US government only produces Level 1R data, its data policy creates space for other institutions that use Landsat data, both within and outside of government agencies to tailor products for their particular applications. In particular, the policies of basically unrestricted access and redistribution rights have had the effect of creating additional demand for data. There is no evidence of saturation of demand over the past five years or so.
Technological Hurdles. The availability of the technology to access, read, process, and interpret Landsat imagery is still somewhat limited, but this hurdle is certainly lower than even ten years ago. Within the US, Landsat imagery is now processed on commonly available workstation technology. Several US universities have developed simple processing software for interpretation that runs on PC’s, and is available free over the Internet. Even more sophisticated software is relatively common commercially. We have not yet reached the day when Landsat data are easily delivered to every laboratory that wants them and has an Internet connection, but from a strictly technological perspective, we probably aren’t too far off. Even though the specifics can’t be predicted, the computer and information industry is currently so dynamic that it is safe to predict that these hurdles will continue to decline.
Access and Interpretability. These are still problems, but they are declining. One problem is simply knowing what is available in archival holdings around the world. The commercialization of the US Landsat system in 1980 had the undesirable effect of making it more and more difficult for any user to know what the holdings were in different receiving stations around the world. The current policy of holding a global archive in the US, but also of enhancing the knowledge of international holdings through shared metadata catalogues has the potential for spurring international demand. There is an enormous gap to be overcome in the area of interpretability for natural resource management, however. The simple classifications that the scientific community has used for global datasets, or that have been used for scientific studies such as the Landsat Pathfinder project are useful for their own purposes; they are generally not adequate for resource management. Nations will require some standardization of terminology and classification for their own purposes, and to ensure that national stakeholders are able to communicate with one another. However, at the same time, it is important to remember that all classifications are useful for the specific purposes for which they were originally intended and may not be for others. Therefore, there is a premium on establishing a set of standards for archiving and documenting the original data on which any classification system depends, for potential future use. Within the US, the federal government has proposed standards for the development and archiving of all geographic data, and has created an open process among its own agencies, and with other scientific and commercial institutions to promote voluntary adoption of those standards. In addition, within the federal land management agencies, there are important national efforts to standardize the national classification systems that are used to identify biological resources, so as to harmonize often conflicting policy demands.
Price. This is of course one of the main hurdles, and has three components. One is the capital cost of designing, developing, and launching satellite systems. These capital costs are large and risky. They are obviously beyond the capability of most nations, and beyond the capabilities of even very large corporations (because of the risk) to bear. Operating costs of satellite systems, while still in the millions of US dollars per year range, are smaller in comparison to the development costs, and have relatively smaller risks. The size and nature of the costs of development and operations, however, are such that for most remote sensing systems they will and should remain in the public domain. They are likely to remain primarily in the hands of the technologically advanced nations with a strong industrial base for some time to come, although that does not mean that they will be exclusively the province of the US, western Europe and Russia, as the experience of the India and soon Brazil and China demonstrates.
The cost of the data themselves are the third component, and the one most relevant to this workshop. The return of Landsat 7 to the public sector has resulted in an enormous decrease in prices for data; from over $4000 (US) for a single level 1 Landsat scene to about $600, and a hope that prices could decline further in the future. And yet neither the commercialized system nor the current system is attempting to amortize the costs of designing, building, and launching the satellite and sensor. Why then, are the costs so different? One perspective is, of course, that the commercial business was in the business of making a profit, and the US government is not. This is true, but seems insufficient to explain the entire difference, since the Landsat 7 ground system is attempting to be self-sufficient with respect to its yearly operational costs.
We believe that there is a more fundamental policy difference, which is reflected in the underlying business model. If you believe that you are in the business of selling data, and the sale of the bits must be used to cover other costs, then the typical pattern is to charge as much as possible for each sale, attempt to restrict the redistribution of the data so that you can sell it again should another customer want it, and generally set the prices fairly high. This is clearly the model that drove the enormous increase in prices that followed the commercialization of the Landsat system in 1980. However, there is another model. If you believe that what you are really selling is the information that the data enable the user to acquire, then one tends to invest in methods and technologies for extracting information, sell those, and reduce the costs and constraints on the data themselves. This tends to enhance the tendency of data users to return to the provider for future purchases, enhance their ability to make larger purchases, and enhance the number of users who are capable of purchasing data. The general direction of US policy with respect to Landsat data is clearly in the latter direction, although there is still progress to be made. There is no reason in principle why the data from a publicly funded remote sensing system in which such large investments have been made should not be made nearly freely available, in much the same way that many meteorological data are made available.
Remote sensing and GIS technologies by themselves cannot hope to provide everything that is necessary for natural resource management systems. Some information will always be required that can only be collected through in situ observation, measurements, and through tracking economic activity. However, the main hurdles of the past in using remote sensing, science, data acquisition, information policy, technological capacity, and price, are changing rapidly. In nearly all cases, these changes are in a direction that would enhance the use of these technologies. Thus, there is a clear opportunity for progress and expanded use of these technologies in natural resource management.
What then, might governments do to take advantage of these opportunities? We believe that there are five main lessons from the Landsat experience that provide some guidance.
Take immediate advantage of the dramatic drop in the price of data from the Landsat system, and ensure that there is complete coverage of Vietnam for the year 2000, and on nearly annual time steps after that. The cost of simply purchasing up-to-date coverage for the entire country (about 30 scenes) should on the order of $20,000, a sum for which sufficient financing should be achievable.
Promote a data policy that would make the raw data, once purchased, freely and openly available for all capable institutions. Ensure that different research and resource management ministries and institutes can acquire copies of the national data set as easily as possible. Collaborations among research and resource management institutions can reduce the overall cost of ensuring nationally important analyses to be done by reducing overlap and unnecessary competition for limited resources.
Promote standards for documentation, archiving, distribution of information, geographic control, and accuracy of analysis. These standards could be developed in collaborative processes with research and resource management institutions. They need not be mandatory, but should be widely known so that all institutes have a reasonable idea for what is expected of them.
Promote the development and use of semi-automated methods of classification to ensure adequate accuracy and precision for natural resource management. As these methods continue to be improved, they have the potential to lower costs substantially.
Promote information sharing, validation procedures, and the public release of information to ensure that there are broadly recognized gains in knowledge, but also in accountability and transparency of the institutions doing remote sensing analysis. In the long run, the public sharing of information allows the development of a strong base of public support for the increased use of these technologies by public institutions.