A review and evaluation of alternatives for updating U.S. Geological Survey land use and land cover maps

H. William Menard, Director

Land use and land cover maps and data must be kept current for effective planning and management of the Nation's land resources. In order to meet the increasing demands for up-to-date land use and land cover information, a primary goal of the U.S. Geological Survey's national land use and land cover mapping program is to. update existing maps and data in a timely and uniform manner. Plans are now underway to begin the selective updating of this land use and land cover information as part of an ongoing program of nationwide land use and land cover mapping. Requisite to establishing an update-mapping program, however, is a careful evaluation of the various alternatives for updating land use and land cover maps.

This circular has been written to satisfy three purposes. The first is to present a review of the considerations that are important in planning a program to update nationwide land use and land cover maps. The second is to evaluate the options within each of the considerations for their applicability to the requirements of the U.S. Geological Survey's land use and land cover mapping program and to offer recommendations concerning those options thought to best satisfy these requirements. The third purpose is to stimulate thinking about additional approaches to updating land use and land cover maps. We hope that users of land use and land cover maps and data will consider their require-ments and de.sires for updated land use and land cover data, and we invite their input. An addressed, postpaid response card which can be detached for use in replying is included in the back of this circular.

ILLUSTRATIONS

v

Since 1974, the U.S. Geological Survey (USGS) has been engaged in nationwide baseline mapping of land use and land cover and associated data at a scale of 1 :250,000. As 1 :100,000-scale base maps have become available, they have been used for special applications and for mapping certain areas. These scales are appropriate for mapping land use and land cover data on a nationwide basis within a practical time frame, and with an acceptable degree of standardization, accuracy, and level of detail. Land use and land cover has been mapped using remotely sensei data at Level II of the classification systen presented by Anderson and others (1976) (see table 1 of this report) and using the compilation specifications documented by Loelkes (1977). As part of this program, the land use and land cover maps and as~sociated data are digitized and the information is s~tored in digital format in a computerized geographic information system, as described by Mitchell and others (1977). The system has the capability to input, store, manipulate, and retrieve the data and makes possible the generation of digital data base tapes and a wide array of graphic and statistical output products.

An essential requisite to better use of the Nation's land is current information on land use and land cover and on the rates and trends of changes-hence, the need for a program of map update. Before e~mbarking on such a program, howeve·r, a systematic appvoach to monitoring and mapping changes in land use and land cover must be developed. A program to update the USGS land use and land cover maps must portray an appropriate level of information, relate to and build upon the existing USGS land use and land cover digital and statistical data base, incorporate the elements of timeliness, cost-effectiveness, and standardization, and meet the varying needs of da.ta users at regional, Sta.te, and other agency levels. An evaluation of the various considerations in the updating of land use and land cover maps is requisite to the establishment of a nationwide p~ogram of map update that will best achieve these objectives.

The major considerations in planning a nationwide program for updating USGS land use and land cover maps are the following:

• How often should maps be updated?
• What remotely sensed source rna te,rials should be used for detecting and compiling changes in land use and land cover?
• What base ma.ps should be used for presenting data on land use and land cove·r changes?
• What maps or portions of a map should be updated?
• What methods should be used for identifying and mapping changes?
• What procedures and what formats should be used for updating maps?

Options within each of these categories must be evaluated in terms of advantages and limitations in the task of map update in order to recommend the optimal alternatives for implementation in an ope·rational natic nwide program of land use and land cover map update.

Since land use and land cover changes occur at highly variable rate's fro.m place to place, there is a need to establish diffe·ring intervals for monitoring and mapping changes. Frequency of update, in the ideal sense, is not only a factor of the dynamics of land use and land cover change but also is related to the type of change, the relative importance of the change, and to the loca,tion, concentration, and amount of change. In reality, howeve·r, such factors as priorities, schedule capacities, cooperative agreements, map scales and data level may make a difference in the decisionmaking process involving frequency of update determinations. In addition, one must rememb~r that the ability to update is also dependert upon the availability of suitable remotely sensed source materials.

Before attempting to establish a frequency for ma.p update, an understanding of the nature of land use and land cover cl' ange is required. Changes take place at variable rates in different loc~tions and at different times within the same area. Anderson (1977) cites some examples of the dynamics of land use and land cover change. The revers,ion of Cropland and Pasture to Fore~st Land in New England is occurring slowly while in parts of the South Cropland and Pas,ture a~e converting to planted pine (Forest Land) at a Inuch faster rate. In central Florida during the 1950's, citrus acreage expanded rapidly and changes in land fro·m other uses to the Orchards, Groves, and Vineyards category were numerous and frequent. By the 1960's, the citrus grove expansion had 'S·tabilized and land use and land cover changes involving this category had decreased significantly.

Residential subdivision development, highway construction, and land use ~and land cover changes involving mostothe1r Urban or Built-Up Land uses take place in very shor4:. periods of time. Although the land area involved may be considerably less than for nonurban categories, the number and frequency of the changes and their far-re,aching impact on the landscape make changes in these "dynamic" urban categorie·s importanrt t9 identify as frequently as possible. Other categories of land use and land cover may undergo little or no change over long periods of time. In some areas, there is a pronounced absence of economic or environmental conditions for change. For example, in many parts of the arid Southwest having extensive areas of Barren Land and desert vegetation, very little change occurs. Areas characterized by tundra or perennial snow and ice cover are relatively unchanging. These areas are characterized by certain types of land use and land cove·r categories and can be termed as "stable" land environments because of their resistance to change.

Finally, regional variations based on historical patterns are likely to affect land use and land cover change. The kind and the magnitude of the changes experienced in an area in the past may not be the same as those likely to occur in the future. For example, urbanization is slowing down in the Northeast and Midwest parts of the Nation, while the Southeast and Southwest are experiencing considerable growth in urban land use and land cover categories. Thus, even though the land use and land cover categories involved in change have the same dynamics, the rate of change varies with region and time.

There are other considerations affecting the frequency of map update in addition to and apart from those already discussed. These considerations include obsolete specifications and classification system compilations, errors, map scale, classification level, or cooperative arrangements with certain users. These considerations are, however, incidental determinants of update frequency since they would most likely affect update only on a one-time basis.

Determining the frequency for updating land use and land cover maps can be ap,proached from two distinct viewpoints, each yielding a different approach for frequency dete·rmination and a different inte·rval of update. In the first, update frequency is determined according to elapsed time. In this approach, a n1ap is updated at a predetermined time inte·rval regardless of the a,mount and type of cl' anges that have actually taken p·lace.

The second approach is based on crite:~·ia of change, where the amounts and the types of change dictate the frequency of update. Each land use and land cover map is updated according to certain predetermined land use and land cover change criteria concerning amourt and type of change, regardless of ·the time elapsed. This approach could result in the update of maps with the same or similar general land use and land cover features at different frequencies and al8o in subsequent updates for the same m•ap at varying intervals.

The approach chosen for frequency determination at this stage-either based on elapsed time or on change criteria-will dictate the procedures to be followed. If map update frequency is based on elapsed time, specific time intervals need to be established for mz.p update. This could be accomplished by stratifying the land use and land cover maps according to a ranking of Level II land use and land cover ca,tegories (see table 1), into are·as of high and low national priorities, areas of high and low change potential (dynamic versus stable land environments), by regions, by age of compilation source material, and by other considerations concerning the nature of land use and land cover change mentioned previously. Each of these groups of maps wo lld be as'Signed a preferred update interval. Three other time factors can also be considered in establishing update intervals for selected maps under the elapsed time approach. One is to tie-in update frequency with the census for urban areas. This census, once taken decennially, is planned to be updated every 5 years beginning in 1985. A second consideration is

to plan updates that coincide with the acquisi-

tion and availability of aerial photographs. A third factor, budget and manpower constraints, would establish a minimum practical thne interval for map update. Within the USGS nationwide mapping program, 5 years from the date of the compilation source ma.terial has been established as the minimum time for the scheduling of land use and land cover map update. Given the scale and the level of land use and land cover detail mapped under this p·rogram, a minimum 5-year update cycle is compatible with :the expres~sed goal to maintain current land use and land cover information.

A procedural advantage to the time interval approach to map update frequency is that no routine review or inspection of candidate land use and land coveT maps is required prior to updating. Since the need for map update is based solely on the age of the land use and land cover data, a map need not be reviewed to satisfy any other criteria or conditions of land use and land cover change for update selection.

The second approach to the determination of map update frequency, in which frequency is dict,ated by specified change criteria, requires a diffe,rent procedure. Since the amount and type of land use change determines the need for update, each land use and land cover map must be reviewed periodically with more current remotely sensed source materials to identify the changes. Thus an additional stepphotoinspection-is required to determine update frequency.

The photoinspection process establishes a cycle for the periodic review of maps to determine the need for update. This cycle designates a minimum or maximum time inte,rval for the photoinspection of land use and land cover maps. A cycle could range anywhere from 1 year to 5 years or more. The same strategy developed for selecting maps for update based on elapsed time eriteria can be used to target maps for photoinspection and to determine likely routine time frequencies for photoinspection with available remotely sensed source materials. On this basis, photoinspection is done at constant predetermined time intervals.

The frequency of photoinspection would not automatically correspond to the frequency of actual map update. Due to the nature of land use and land coveT change, not all the maps would meet the change crite,ria for update selection at the given photoinspection time. Thus, although photoinspection is conducted routinely on a prescheduled frequency, map update is performed only as often as necessary.

If a map update program is to be effective, it must incorporate the element of time together with an understanding of land use and land cover dyna.mics in the deternination of update frequency. It is recommended, therefore, that updarte frequency be st-uctured on the basis of need as dete,rmined by criteria of land use and land cover chang~ and that elapsed time be considered an imp'lrtant, supportive input to update planning. The overall approach to establishing a frequer~y for land use and land cover map update sh0uld encompass two s~tages :

1. A photoinspection procedure should be established to periodically revi o.w land use and land cover maps in orde,r to identify the changes that justify updating the maps. Routine photoinspection for seleeted areas should be conducted on predetermined cycles based on elapsed time.
2. A map update procedure should be established in which maps are selected for update p~rimarily on the oo sis of land use and land cover change, where the amounts and types of chang?.'S identified in ~the photoinspection phas~~ determine the need for map update. The frequency of actual map update canne~ be predetermined since it is dependent on the outcome of photoinspecHon. In establishing a routine cycle of photoin-

A land use and land cover map update program should be designed to make maximum use of remotely sensed data. Present remotesensing technology offers a wide selection of sensors and capabilities. From past experience in land use and land cover mapping, some general guidelines have been established by Loelkes (1977) for the selection of remotely sensed source materials. These guidelines are the following:

1. The format of the remotely sensed data must be such that it is or can be rectified, if necessary, to a tilt-corrected format.
2. The scale should be no larger than 1 :60,000.
3. Source photographs for the winter season (leaf-off condition) should be used, if available.
4. There must be no snow cover.
5. The photographs should contain no more than 10-percent cloud cover.

In addition to these guidelines, two additional criteria should be considered. The resolution of the remotely sensed source should be sufficient for interpretation of Level II land use and land cover categories accurately and consistently. Comparability with the past or future land use and land cover data sho1ld be provided in that the level and class of information extracted from the update source is the same as that contained in the earlier map data base.

These photographs range in scale from 1:20,000 to 1:40,000 and a.re taken from aircraft at altitudes of 10,000 to 40,000 fL They have the following advantages for lari use and land cover map update: They can be rectified to a tilt-corrected forma.t for mapping; the resolution is sufficient for interpreting Level II categories; and they often provide more frequent covera,ge for some areas than is provided by higher altitude sources. Large-scale, low-~altitude aerial photographs p·1S·sess, however, several limitations for map update. Since land use and land cover map update is of a nationwide scope, the land area involved preeludes extensive use of large-scale remotely sensed da.ta due to the lack of complete coverage needed over some large areas. In addition, the cost,s associated with low-altitude aerial photograph acquisition and data utilization and integration over large areas are exc~ssive. Finally, the scale and resolution of larg·~-scale photographs far exceed that required for efficient Level II land use and land cover mapping on a nationwide basis, the-reby lessening the comparability of updated data with the existing land use and land cover data base developed from high-altitude source materials. There is a tendency when using larger scale photographs to see and therefore extract more detailed l~and use and land cove'r information. Generally, the structure of the mapped Level II categories de,rived from large-scale sources is somewhat different f~om the same Level II categories derived from small-scale sources. With the low-altitude, large-scale photographs, the .classification is derived f~om an aggregation or building up of more detailed information; with high-altitude, small-scale photographs, more general information is broken down into its component parts in case's where this information conforms to the classification system and mapping specifications in USGS Open-File Report 77-555 (Loelkes, 1977).

Small-scale, high-altitude photographs have nominal scales between 1:80,000 and 1:120,000 and are obtained from aircraft flying at altitudes of 40,000 to 60,000 ft. They are principally NASA high-altitude RB-57 and U-2 photographs and photographs acquired for USGS topographic mapping. Such photographs are usually black-and-white panchromatic or colorinfrared. They have been used successfully in compiling nearly all past land use and land cover data.

Small-scale, high-altitude photographs have the following advantages for map upda,te: They can be rectified to a tnt-corrected format for mapping; the re,solution and scale are sufficient for interpreting and mapping Level II ca,tegories accurately and efficiently for large areas at 1 :250,000- and 1 :100,000-scales; and they P'rovide data comparability with the existing land use and land cover data base. The limitations are that the available or existing source may not be the optimal source for land use and land cover mapping based on the guidelines mentioned previously and the coverage is often nonrepetitive or infrequent thereby limiting the selection of areas that can be updated.

Much research has been conducted on the use of Landsat imagery for land use and land cover mapping and change detection. However, to date, the potential value of Landsat MSS imagery for consis~tent meF.suring and mapping of spatial and temporal changes in land use and land cover has not been fully demonstrated. Clearly, Landsat MSS imagery presents several advantages over conventional aerial photographs. It affords syn0ptic coverage of land use and land cover as frequently as every 18 days. In addition, c1verage for nearly all parts of the country is readily available at relatively low cost and fills in the gaps left by conventional aerial photographs in providing the necessary source for lr.nd use and land cover update maps where and when needed. Its repetitive capability makes possible the routine monitoring of areas for changes in land use and land cover. Landsat MSS imagery can also be used to "target" specific areas, having extensive Level I-type land use and land cover change, which may then be updated using higher resolution sources.

The outstanding advantages of frequency of coverage, repetition, availability, and low cost make Landsat MSS imagery a uniquely appealing source for land use and land cover map update. However, it does possess certain formidable limitations for opera,tional use. Studies undertaken in southern Arizona (Place, 1974) and elsewhere (Fitzpatrick-Lins, 1978) using Landsat MSS imagery have showr that Landsat imagery is not wholly approprir.te for Level II land use and land cover mapping. Several USGS Level II land use and land cover categories cannot be identified and mapped from the Landsat imagery alone. Particularly, the resolution of the MSS imagery does not permit accurate and consistent separation of those land use and land cove~r categories most likely to undergo temporal changes. As a result, Landsat imagery would be least suitable as a source of land use and land cover change information in those very areas where most of the changes are occurring. The inability of Lardsat to replicate the same Level II categories contained in the baseline land use and land cover maps, as well as from one Landsat scene to the next, limits its use as source material for detailed map update since comparability with the existing baseline data cannot be realistically achieved.

A somewhat more promising application of Landsat data for land use and land cover mapping and change detection is offered by computer-aided analysis of Landsat MSS digital data. Ongoing research and development activities conducted by the USGS (Milazzo and others, 1977) suggest the potential for using Landsat digital data for land use and land cover change detection in an operational map update program. Lands~at digital data afford the same currency of information as the MSS imagery. In addition, computer processing makes possible the examination and classification of large amounts of data quickly. Since the data are already in a digital format, they form the basis of a geographic information system capable of providing simultaneous area measurement tabulations. Thus, the need to convert a graphic land use and land cover map into digital format to derive area statistics is eliminated.

In computer processing of Landsat digital data for land use and land cover change detection, changes are derived by comparing the relative spectral responses occurring over time. Georectified and reformatted tapes from two time periods are overlaid in the computer and the spectral differences between the two scenes print out as possible areas of change. These areas are then grouped in a specialized classification and related to corresponding USGS Level I and (or) Level II categories. In this technique, the mapping of change is accomplished by comparing two data sets, not two separate maps or interpretations. In theory, this technique seems quite workable for change detection. However, in practice there are several significant problems which, given the present technology, may or may not be resolved.

One problem lies in converting spectral differences into valid land use and land cover category changes. It is unrealistic to expect a specialized land use and land cover classification developed from computer-classified Landsat data to match category for category or to merge with the USGS Level II classification system which is designed for human compilerr using other remotely sensed sources. Due to the incompatibility of these two systems, the detection of either "from" or "to" changes in each of the USGS Level II categories cannot be accomplished. Hence, a direct update of l2nd use and land cover maps derived from other remotely sensed sources with Landsat digital data is not feasible.

Another formidable problem in using Landsat digital data for change detection and map update is that such a map of "changes" is likely· to contain conside~rable noise, or false changes-areas tha,t appear different spectrally, but their use or cover category has not changed. The identification of false changes is due to a number of factors including misregistration of pixels in the Landsat scenes being compared; comparison of da,ta from different seasons, particularly involving crop rotation cycles, where the rotation causes the same agricultural activity to be erroneously identified as a land use and land cover category change; identification of differences in soil moisture content or other environmenta,l anomalies that result in different spectral signatures for the same land use or land cover category ; and misclassification of land use and lanc. cover due to poorly grouped spectral classes. One way to reduce the occurrences of false change identification is to raise the threshold level. Differences as small as one picture element (pixel), 1.1 acres, can be identified from the Landsat digital data. By raising the threshold, one can sample la,rger groupings of pixels, for example 8-pixel by 8-pixel windows, in which only changes of approximately 64 acres or greater are mapped. Although this approach significantly reduces the amount of "salt and pepper" type false changes, it also results in a loss of sensitivity to many valid land use ard land cover changes, which often occur ir small areas. There is also a loss of comparability with the Level II classification system in which land use and land cover is mapped at 10- and 40-acre minimums. Therefore only a few gross Level I category changes such as Forest to Agriculture are identified.

The newly acquired NASA Landsat RBV imagery appears to be a promising source for land use and land cover mapping and update roughly equivalent to the small-scale, high-altitude aerial photographs used in past mapping. Although no thorough evaluation has been conducted, the imagery viewed thus far suggests promise for use as either a photographic source for direct updating of maps in cases where the resolution permits reliable discrimination of discrete Level II land use and land cover categories or for targeting those areas where enough Level I changes have occurred to justify update using higher resolution materials.

The level, amount, and accuracy of land use and land cover data that may be obtained from remotely sensed data depend on the altitude and resolution of the sensor and, to some extent, on the scale of the derived product. Anderson and others (1976) recommend the use of high-altitude source data from altitudes of 40,000 ft or more, with scales of 1:80,000 or smaller for the extraction of Level II land use and land cover information. In keeping with this recommendation, the principal remotely sensed sources used for the mapping of Level II land use and land cover data have been NASA high-altitude U-2 and RB-57 photographs and photographs acquired for USGS topographic base mapping. Small-scale, high-altitude photographs have been used successfully for compiling nearly all past land use and land cover information and are capable of meeting the seven recommended guidelines for remotely sensed da.ta selection previously enumerwted. It is expected that small-scale, high-altitude photographs will continue to be the mainstay of future land use and land cover map compilations, and it is recommended that they continue to be one of the primary photographic sources for updating land use and land cover maps.

In addition to the use of small-scale photographs for updating maps, it is also reasonable to consider using large-scale photographs in a limited and complementary way when more suitable smaller scale sources do not exist or when a more detailed land use and land cover classification is required for special applications.

The use of Landsat digital data in updating land use and land cover maps relies on the development of new and improved d~,gital data analysis techniques and on continuei improvements in computer technology. At present, Landsat digital data are most appr')priate for monitoring gross land use and land cover changes and for assisting in the urdate selection process for manually compiled land use and land cover maps. The data do not c':mstitute a primary source for such an updatE., but may rather function in a complementary role.

The 1: 250,000-scale topographic map possesses the following general advantager as a base in updating land use and land cover maps: It has proven to be a satisfactory base and scale for the presentation of Level II lard use and land cover data in a standardized, nationwide program of land use and land cover nTapping; it is available nationwide-maps are ccmplete for all 50 States; the scale is adequate for mapping Level II land use polygons at 10- and 40-acre minimum sizes; most baseline land urP. and land cover and associated data mapping has been keyed to the 1: 250,000-scale base map; and most baseline land use and land cover· maps and associated data have been digitized using the 1: 250,000-scale map base control.

The 1: 250,000-scale topographic map has certain limitations in its use as a base for updating maps. The positional accuracy does not meet National Map Accuracy Standards (NMAS) in most cases, even though maps are revised periodically; bases for some areas, and hence planimetric data, are often out-of-date; and the scale does not satisfy the needs of many land use and land cover data users.

The use of the 1: 100,000 intermediate-scale topographic quadrangle map as a base for updating land use and land cover has the following advantages: It has good positional accuracyit meets NMAS; it provides relatively up-todate planimetric base data; the scale is wellsuited to Level II land use and land cover mapping and data presentation at 10- and 40-acre minimum mapping units; it permits easy readability of Level II polygons and identifiers, particularly in congested areas; and it allows the flexibility of mapping polygons at lower minimum 111apping sizes since 10- and 40-acre mapping units are well above the threshold of clarity and readability at this scale.

There are disadvantages to the use of the 1: 100,000-scale 111aps: Almost 2,000 maps are required to cover the conterminous United States at 1: 100,000-scale, as compared to 473 for the 1: 250,000-scale series; there is incomplete coverage nationwide-bases are generally not available for those areas where updates may be needed most ; land use and land cover compilations on the 1: 100,000-scale base offer the same information to the user as does the 1 : 250,000-scale map when the same classification level and the same minimum mapping units are used; and, it is more costly to compile and produce the 1: 100,000-scale map than the 1: 250,000-scale map. In addition, a mixture of baseline land use and land cover maps at 1: 250,000-scale and updated land use and land cover maps at 1: 100,000-scale creates confusion for users and makes cross-sheet boundary land use and land cover data utilization difficult, which diminishes the range of usefulness of the map product for many State and regional applica,tions. In the geographic information system, comparability between the merged graphic and statistical data generated from land use and land cover polygons digitized using 1: 250,000-scale base control and those :rl)lygons digitized using 1: 100,000-.scale base control may not be achieved due to positional accuracy differences between the two base-s.

The advantages and limitations de-scr~bed for the 1: 100,000-scale quadrangle series F.pply to the 1: 100,000-scale and 1: 50,000-scale county map series as well. In addition, use of this base for land use and land cover map upcate deviates even further from the goal of standardized map products in the national mapping program. This map series can function more appropriately as a supplement to the standard land use and land cover map series and is quite appropriate for special applications.

Most land use and land cover maps tllat have been digitized thus far have been concentrated in those areas mapped on the 1: 250,000-scale base. These are usually the older maps and are therefore the ones more likely to be among the first selected for upda,te. For these maps, revised 1: 250,000-scale topographic bas~ maps have become available, or new 1:100,001 intermediate-scale topographic quadrangle maps have been or will soon be completed. In both cases, the use of a map base in the updat-e phase that is different from that used in the initial mapping of land use and land cove1r will impact the statistical data base already developed from the digitized map. Misregistration of land use and land cover polygons ~ould occur due. to differences in positional accuracy. Hence, the integrity of the statis,tical data being built upon and added to the data base may be affected adversely. This is especially so in the case of the 1: 100,000-scale serie~s map versus the 1:250,- 000-scale map.

An additional input to the selection of a map base for land use and land cover update involves regional considerations. In some parts of the Nation, the land use and land cove'r patterns are less complex than in other parts. For example, many areas in the Central United States, as well as many parts of the West and Southwest, have extensive acreage in a few stable land use and land cover ca.te·gories such as Rangeland and Barren Land. In such areas where rapid or numerous land conversions are unlikely, or overall map content and de·tail are minimal, the 1: 250,000-scale base remains adequate for the presentation of land use and land cover information. In other parts of the Nation, where land use and land cover patterns are more dense and complex and where changes in land use and land cover occur with greater number and frequency, it is desirable to map land use and land cover on the 1: 100,000-scale bases. Given these considerations, it seems reasonable to continue the use of two bases for nationwide land use and land cover mapping.

The following app·roach to the selection of a map base for first-round land use and land cover map update is suggested:

• The 1: 250,000-.scale bases should be used for land use and land cover map update in those areas where 1: 100,000-scale bases are not available and should continue to be used for map update in those a.reas where, given the nature of the re·gional land use and land cover patterns, it is desirable to continue the pre,sentation of updated map data at that scale, even if 1: 100,000-scale bases are available.
• Update of existing 1: 250,000-scale land use and land cover maps should be on the 1: 100,000-scale base's if such bases are available at the time of update and if the regional location and nature of the land use and land cover patterns and expected changes make it advantageous to convert to the larger scale base.
• For those land use and land cover maps already compiled at 1: 100,000-scale, update should be continued on the 1: 100,000-scale bases. The approach to map update outlined here

t1ries to balance the blend of old and new. It attempts to maintain the much needed and used digital and statistical data base already developed for the 1: 250,000-scale map while allowing the gradual incorporation of and eventual conversion to the 1: 100,000-scale base on a regional basis for land use and land cover mapping.

Once a map has been selected for updating, a determination of the exact are~.l extent of the required update must be made. The maps, or parts of maps, should be chosen where, upon inspection, the amount, concentration, and type of changes are sufficient to justif~r an update. The map may be updated by either the full map sheet or by partial m·ap sheet.

Full map-shee·t update involves updating an entire map (at any scale) at one time. This approach has the advantage of providing continuity with the forme·r baseline map, in that a completely updated version of the old map is released in the same form. This ler.ves the user with no question as to what partF of the map have been updated. Also, all of the land use and land cover data contained on the map are from the same source and time. The disadvantages, however, in the full map update 2.pproach are significant for small-scale maps such as the 1: 250,000-scale map. Suitable photographic coverage for the entire map is usually difficult and costly to procure. Even so, updating the full map sheet is appropriate in cases where changes in specifications or classification system necessitate complete recompilation and revision of the map or where the changes are in fact extensive enough to cover most of the map area.

In this approach, only those parts of a map are updated where a ·Significant amount of change has occurred, is likely to be found, or where user interest or needs are greates·t. These map pa.rts may be defined by geographic coordinates, quarter-map sections, SMEA's, political boundaries such as those for counties or States, designated regional, environmental, or administrative areas, seleeted land use and land cover categories, or other criteria. This approach has the advantage of providing the user with an updated map product quickly, ef!ciently, and inexpensively. However, a limitation exists in that the release of a full map for which only specific parts have been updated creates the need for additional map legend explanations, such as a coverage diagram, to tell the user what parts of the map have been updated and to what date. Another disadvantage is that the user must obtain supplemental material for identifying the exact boundaries of the updated areas on the map. The need for supplemental data could be increased by successive rounds of map update when update is undertaken for different parts of the map at different times. This would ultimate,ly decrease the comparability and usability of the data.

Another disadvantage of partial map update is that it creates innumerable internal edgejoining difficulties be1tween the polygons at the boundaries of the updated area and those (either extensions of the same polygon or different land use and land cove~r polygons) outside the updated area. To effec,t a proper internal joining, the limits of the update area must become the arbitrary or artificial boundaries between the new and the old adjoining land use and land cover polygons. If this joining procedure is followed through successive rounds of partial map sheet update, a se,ries of artificial boundaries may be set up on the map. These boundaries could crea~te a visible pattern of land use and land cover polygons that follow the outline of each updated area rthereby segmenting the map on this basis. This adversely affects the attempt to portray land use and land cover polygons by natural boundary delineation.

One of the more difficult decisions concerning the update of land use and land cover maps deals with recommending the areal extent of map update. Because both 1:250,000- and 1: 100,000-scale maps have been used as bases for initial data compilation in the USGS land use and land cover mapping program, a different decision concerning what maps or portions of a map should be updated is required in each case. The following recommendations for the areal extent of map update are proposed:

• For existing 1: 250,000-scale land use and land cover maps that are to be updated at the same scale, update should be done for the complete map sheet.
• For existing 1: 250,000-scale land use and land cover maps that are to be c1nverted to the 1: 100,000-scale bases for map update, update should be done on 1: 100,000- scale bases covering either the complete 1 :250,000-scale map sheet or selected quadrants.
• For 1: 100,000-scale land use and land cover maps and aH other maps to be upiated on 1: 100,000-sC'ale bases, update sl,ould be done for the complete map sheet.

Thus, whatever the map scale for the presentation of updated land use and land <'OVer information, it is generally recommended that the complete map be updated at one tim~. Since land us~ and land cover data are compHed and formatted by full map sheet, it should likewise be updated on that same basis. This ~uproach maintains continuity in mapping design and eliminates the hazards associated with overcustomizing the map product to the point where it serves a decreasing number of users.

Change detection is a separate and distinct step in the land use and land cover map update proces,s. It deals with the detection and identification of temporal changes in land use and land cover from a comparison of remotely sensed and (or) map source materials from two or more time periods and the mapping of those changes in accordance with established land use classification standards and map compHation specifications. There are four basic me,thods for detecting land use and land cover changes : original source material-to-new source material, original map-to-new source mate,rial, original map and source material-to-new 'Source material, and original map-to-new map comparisons.

In the original source material-to-new source material change detection technique, changes are de,rived from a comparison of the original compilation photograph'S with the new photograph'S. In thi8 approach, one compiler determines the land U8e and land cover c~ tegorie:s from the old photographs and makes the change asses,sments from the new photographs at the same time. This technique has certain advantages in that the compile1r can see the actual land use and l~and cover fe,atures as they appeared at both times, and e·rrors in the original compilation, eithe·r of commission or omission, can be identified if they are involved in change. This technique thus allows for a limited correction of the data base.

There are at least five limitations of the original source-to-new source p~rocedure for detecting changes in land use and land cover. Firs,t, the costs of storing or reacquiring all of the original compilation mate~rials are high. Second, in an original source-to-new source comparison, the compiler is looking at only the areas involved in change on the two sets of photographs. Since no reference is made to the land use and land cover information on the original map, the compiler is likely to miss original compilation mapping errors that are not involved in change. The third limitation of of an original source-to-new source change detection procedure is that it tends rbo add more land use and land cover detail in an update because the compiler intentionaHy focuses on small, discrete land use and land cove1r categories in order to search out every possible change that might meet minimum mapping unit size1s. Thi8 results in a tendency to rearrange the more gene~ralized existing boundaries of the land use and land cover polygons in order ·to accommodate many small polygons of change.

A fourth limi~tation of the original source material-to-new source ma.te·rial method i8 that it generates an undue numbe·r of inte·rpretation and delineation "disagreements" between the original data and the data interp.reted during update. The disagreements usually involve indefinite categorie'S and boundaries which must be subjectively interpreted and delineated. The individual conducting the update, who has access to both sets of sources, may not make the same decisions as were made by the original compiler. The· end result may be an update fraught with unneceiSsary data base correetions. Finally, the detection of valid land use and land ·cover changes by direct visual comparisons between two or more remotely sensed sources acquired in diffe,rent years is cumbersome, time consuming, and difficult because the sources are frequenMy of different film types, film quality, seasons of year, and scales. The first three variables may be selectively controlled. Scale differences can be minimized, but this usually requires costly rescaling of the photographs by either optical-mechanical methods or by labora,tory processing.

In the original map-to-ne·w sour~e material procedure, land use and land cover changes are identified by comparing the original compilation with the new photographic sc~1rce materials. In this technique, a transparency is made of the old land use and land cover oompilation. The transparency could be a matte-p•)sitive base with land use and land cover polygons and identifie·riS shown in blue-line or a negative scribecoat guide of the old land use and land cove~r data. This transparency, which serves a1s the compilation base for the mapping of land use and land cover changes, is then overlain on the new photographic base film which has been rectified and (or) scaled, either photomechanieally or photo-optically, to the ~scale of the compilation base. Using either the scaled base film or the new source in i,ts original form as viewing film, each delineated polygon on the base map is compared with the coJ·responding land use and land cover feature, appea,ring on the new source photographs. Any changes in land use and land cover between the photograph and the map are recorded on the compilation base ; only the actual areas of change are mapped. The ba,se photographs are used to help locate and position the new land use and land cover polygons on t:he compilation base. Each polygon of land use and land cover change can be labeled by a four-digit "from-to" code (fig.

1) which identifies the former (fro'll) Level II

land use and land cove~r category (as mapped in the old compilation) and the new (to) Level II land use and land cover category (appearing on the new remotely sensed source) . This is the least co8tly, the easiest, and the mo<;'t expedient method of detecting land use and land cover change and maintains an overall ac~uracy level oommensurate with the es,tablishei standards for USGS land use and land cove-r mapping.

c;::?-4-z-12

p~76-/i

OS cJ [}J~2-7b

However, in the original map-to-new source material approach to change detection, there is an inherent danger that errors in the original compilation may go undetected. Figure 2 shows examples of five types of errors that can occur. These errors are more · easily identified and corrected if the old compilation source material

A~ v:-21~13

and the original map are compared with the new remotely sensed source. The differences in the final maps of iand use and land cover change that result from the original map-tonew source and the original m31p a11 d source-tonew source change detection methods are also illustrated in figure 2.

Perhaps the mos.t thorough and comprehensive technique for detecting land use and land cover change is provided by the original map and source material-to-new source materia,} method. It combines the elements of ,the two change detection techniques described previously. In this procedure, the land use and land cover map is scaled to the new remotely sensed source material which is then compared simultaneously with both the original map and the original source materials. This p·rocedure has the following advantages: The oompile·r is able to see the actual land use and land cover features as they appeared at both times as well as

to see how they were originally mapped ; the

original land use and land cover map can serve as a base control for measuring and mapping all land use and land cove:r changes ; the same general level of detail in land use and land cover mapping used in the original compilation is retained in the update; and all errors in the original compilation can be identified thus permitting complete correction of the data base.

There are also two limitations as,sociated with the original map and source material-tonew source matedal change detection procedure. The eosts to store or reacquire the original compilation sources are high, and the detection and assessment of land use and land cover change by direct compa.rison between two or more photographs can be a difficult and timeconsuming procedure owing to differences in film type and quality, season of the year, or scale.

A final change detection method is the original map-to-new map procedure. In this technique, a new map is compHed using new source material. Polygons of land use and land oove~ change can be derived from a comparison of the original map with the new map. This procedure possesses one advantage in that the original compilation source material is not needed.

The,re are also sevocal disadvantages to using this method for land use and land cover change detection. First, after the new land use and land cover map is produced, an extra step is required to isolate the polygons of land use and land cover change. Second, some of the land use and land cover changes identified may be a result of purely subjeetive differences in cate·gory interpretations and boundary delineations between the two compilations and do not necessarily reflect valid sequentia,l changes. Third, the time and cost to manuaHy compile a new land use and land cove~r map are excessive and seldom warranted. A final disadvantag1~ to this method is that the remote'ly sensed source materials used to compile the new map may not be as good as the source materials usei in the original compilation (In this case, the original map is replaced by one that is indeed m0·re current but does not achieve the same quality or accuracy level.) .

Of the four change detection methois, it is recommended tha.t maps be updated whenever possible by using the original mar-to-new source material method. However, since the change's mapped by the original ma.p and source material-to-new source material approach a.re likely to be more correct, it is strongly recommended that the old compilation photcgraphic source material if still readily available be used to supplement the basic original mar-to-new source material method for detecting cl·anges.

The final considerations in a land use and land cover map update program deal with map update procedures and product formats. Procedures for updating land use and land cover data include complete or partial map recompilation, the incorporation of detected changes in an updated original map, the mapping of detected changes on a separate map overlay, or the separate generation of statistical data.

If ~the map recompilation procedur1~ is selected, all or parts of a map are rec'1mpiled from new remotely sensed source materials. In a complete reoompilation approach, r. whole new land use and land cover map is produced from new remotely sensed source materials. This is the most drasUc of the alte·rnative~s-a "start from scratch" approach. The new map has the same product format as the original. In this approach, all data are changed or corrected as necessary, and no attempt is made to keep previously mapped land use and land cover bounda.ries or categorie·s.

A partial recompilation is a less drasrtic approach to map update than the complete recompilation alternative. In this approach, only s·pecified portions of the original map, or selected map content, are corrected or updated by recompilation. The result is a complete land use and land cover map in the same format as the original compilation that combines parts of the original with the revised or new compilation. Again, the partial recompilation approach could be used most practically when parts of the original compilation need extensive reworking, for example, where Forest Land or Forest/ Wetland delineations are significantly inaccurate due to poor quality sources or where inappropriate seasonal coverage was used in the original mapping phase. Any parts of the old map, however, known to be of acceptable quality and accuracy and mee~ting >Specifications are retained.

In an update of the original compilation, only the changes that have occurred over a specified time period are delineated and these changes (new polygons) are incorporated with the original compilation. In this approach, the original compilation is pre,served as the base. The previous land use and land cover boundaries and categories are held as mapped, and only appropriate corrections, additions, and deletions are made. In this procedure, a separate ove1rl~y regis.tered to the original map is used to recotd only the polygons of change. The land use and land cover polygons on the old map that correspond to these areas of change are deleted on the original base. The ove·rlay is then composited with the original compilation, either photographically or digitally by compute·r. The result is an updated map in the same format a.s the old compilation.

In this approaeh, the changes in land use and land cover over a specific time period are isolated and mapped on a sepa.rate overlay keyed to the original compHation. A sin ,gle map or ove,rlay is produced, ei,ther by photographic laboratory process.ing or by computer, showing only the polygons of change. Eacr polygon is labeled by a four-digit change code which identifie~s both the prior Level II land use and land cover category (from the original map) and the new land use and land cover carte·gory (from the new photographic source) . An example of a prototype map showing land use and land cover changes for a part of the Washington, D.C., metropolitan area is shown in figure 3.

In this approach, a new map product is not provided. The land use and land cover change·s between the original map and the new remotely sensed sources are recorded on an ove·rlay in manusc.ript forma,t only, solely for the purpose of digitizing. Once the new data are digitized, updated statistics are released in computer printout format, which provide information concerning the prior and new lane, use or land cover and the areas of change. The original land use and land cover map remains unchanged; only the statistics reflect the changes.

It is recommended that the upda·ted original procedure be used whenever pos.rible to generate a map product showing comt:·ete land use and land cover information. In the updated original approach, changes are incorporated into the original compilation ·to create a new map showing the current land use and land cover information at the time of update. Except for a written explanation in the legend of the map identifying the upda1ted area (s) (if it is less than the full map sheet), nc provision is

made to distinguish between the old land use and land cover polygons and the new ones.

Another product, updated s,tatistics, should accompany the complete land use and land cover map. The statistics reflect the amount and type of change and its location in terms of selected data associated by drainage basin, county, census tract, or othe·r associated map data.

A separate overlay showing the "from-to" polygons of land use and land cover change for a given period also should be provided. either in the form of a manually-compiled cr computer-derived map, for those users whose applications require the specific polygon data on change. It is recommended that the change overlay be made a part of the standard set of associated maps and data which supplement the land use and land cover maps. Since a change overlay can be produced quickly and with ve,ry little additional preparation and expense, it probably will be the first product of the map update program.

It is recommended that map recompilation be unde1rtaken only in cases where there were significant changes in the classification system and (or) mapping specifica,tions, or othe-r situations where it is infeasible or imp,ractical to build upon the established land use and land cover map base.

In the review and evaluation of various considerations in accomplishing map update within the framework of the present USGS land use and land cover mapping program, recommendations for updating land use and land cover maps in accordance with the goals of the USGS land use and land cover mapping program have been offered. In summary, the recommendations are as follows:

• Initiate periodic inspection of land use and land cover maps using available remotely sensed source materials to identify changes, with inspection scheduled at a 5-year minimum interval and actual frequency of map update dependent on the amount and type of land use and land cove~r change.
• Continue to use small-scale, high-altitude photographs as the principal remotely sensed source material for updating land use and land cover maps.
• Use the 1:250,000-scale base maps for updating land use and land cover maps of those areas where either 1: 100,000-scale bases are not available or for those regions where it is desirable to continue the presenta,tion of map data at 1: 250,000-scale. Use the 1: 100,000-seale base maps for updating those maps already compiled at that scale or for updating existing 1 :250,000-scale land use and land cover maps for which 1: 100,000-scale bases are available and the regional location or the nature of the land use and land cover data make conversion to the large,r scale base desirable.
• For existing 1:250,000- and 1 : ] 00,000-scale land use and land cove1r maps which are to be updated at their respective scales, update the complete map sheet. For exis,Ung 1: 250,000-scale land use and land cover maps that are to be converted to the 1: 100,000-scale bases, update should be by whole 1: 100,000-scale map she.Bts covering either the full 1: 250,000-scale map area or selected quadrants.
• Use the original map-to-new source material method as the principal land use and land cover change detection techniClue. Supplement this method by using the original compilation remotely sensed Pource material where needed and when available.
• Update maps using the updated original approach in which only the outd~.ted portions of the original map are replaced with new data. Present data in a complete land use and land cover map format. Provide updated statistics on land use and land cover. Provide a separate overlay showing only the polygons of land use and land cover change as an optional associated map supplement. Conduct recompilations only when necessary. These recommendations do not necessarily

represent the only approaches to updating land use and land cover map's. They are suggested as a foundation for the development of a strategy for updating maps in the nationwide land use and land cover mapping program. We hope that the recommendations will stimulate users to consider and exp,ress tl'eir requirements for updated land use anc land cover information.

.;;, WE :nED YOUR HELP ~

r----- --- --- --- -- -- -- -- --- -- ---

Where this page came from

This page was imported from U.S. Geological Survey. Published by the U.S. Geological Survey and, as a work of the United States government, in the public domain.

Nobody has written it yet — it is the source material at a new address, which is why search engines are asked to skip it and why no one earns from it. It is up for grabs: take it on, and it is yours to rewrite and to earn from.