Previous Page Table of Contents Next Page


CHAPTER 10. SURFACE HYDROLOGY: WATER BODIES, WATER POINTS, DRAINAGE AND WATERSHEDS


This section of the inventory outlines data supporting the representation or analysis of surface hydrological features. The inventory for this section has been broken into three topical subsections of globally consistent data sources covering: surface water bodies (SWB) and water points; surface drainage, rivers, and flow routing database; and watershed delineations and models.

10.1 SURFACE WATER BODY DATABASES OR LIBRARIES

Three classes of surface water bodies have been identified for the inventory. These classes include SWB and wetlands features captured from: cartographic sources; those potentially containing limnological attributes; and lastly, those derived from satellite imagery.

10.1.1 Cartographic surface water body and wetlands databases

Data layers included in the cartographic surface water body and wetlands data topical subsection of the inventory can be broken out based on whether they use points or polygons to represent the spatial SWB features. In total six data layers were identified as relevant to this topical subsection and are summarized in Table 10.1.1.

Data Type/Source

URL

Extent

Scale

Availability1

Notes

SURFACE WATER BODY AND WETLAND POINT D ATA LAYERS

NGA-GNS/GEOnet Gazetteer

http://164.214.2.59/gns/html/index.html

Global

variable
1’- 1"
dms
1:250 000

Public Domain

Database contains ~316 000 named locations of lakes, dams, reservoirs, wells, wetlands, and springs

NGA-DCW Gazetteer and Annotation data layers

DCW available from 1992 original set of four CD-ROMs

Global

1:1 M

Public Domain

Layers contain ~29 000 drainage feature names captured from the original ONC charts, SWB’s can only be differentiated for Anno.

Consolidated DCW and VMap0 point SWB features

DCW from original 1992 set of four CD-ROMs, VMap0 available via USGS-Store

Global

1:1 M

Public Domain

Harmonized layer of dams, lakes, wells, & other SWBs from: DCW-DNPNT, DCW-DSPNT, and VMap0-MP data layers

SURFACE WATER BODY AND WETLAND POINT D ATA LAYERS

UNCS 1:5 million Named SWB data layer

UNCS data portal not currently available

Global

1:5 M

CR, FQ, PD

Four possible classes of SWBs, data not available for comparison

RWDBII-Sv1.1 Named polygonal major, minor, and other SWB data layers

CD available from FA O or WHO

Global

1:3 M

Public Domain

Three layers contain 7 273 SWBs globally, 4 578 are named, but also includes country divisions

NGA-VMap0 Ed.5 Named Inland Water polygonal data layer

http://geoengine.nima.mil or available as a set of CD-ROMs via USGS-Store

Global

1:1 M

Public Domain

Largest scale source of SWBs with three valid classifications; for Africa 2 654 out of 24 389 SWBs named

1 CR=Copyright; PD=Public Domain; FQ=Fair Quotation

Point surface water body and wetland features

Three point SWB data layers are detailed in Table 10.1.1 and the LOE associated with the processing of these data is relatively low. In particular, as the NGA GNS and DCW gazetteers are recommended for processing in total, existing encoding in these databases allow for the identification of named SWB and wetland features fairly easily. Unfortunately, the DCW gazetteer includes SWB features under a more general "drainage" encoding, which also includes named river features. For this reason, the use of the various "levels" of annotation accompanying the DNNET versus the Gazetteer layer of the DCW are the only way to differentiate SWB features correctly. This represents one further argument in favour of considering the 2.5 day LOE associated with processing the DCW annotation into definitive layers and then creating a more robustly encoded DCW Gazetteer for inclusion in any CGDB data library.

The third point SWB data layer recommended for consideration by the UN would be one based on the harmonization of the DCW Drainage Point (DNPNT) and Drainage Supplementary Point (DSPNT) data layers with the VMap0 Miscellaneous Point (MP) data layers. Processing in conjunction with Africa indicate that the DCW still provides a more robust baseline for these layers than the VMap0. Further, the DCW can be used to account for features inexplicably shifted between the two data libraries based on reference back to the source ONC Charts and supplementary data sources of SWB features. The updates included in latter editions of the VMap0 data library can also be accounted for at a fairly minimal LOE.

The issue of whether to conduct such a harmonization may however, come down to the relative LOE required to consolidate the data layers. The LOE for processing the VMap0-MP is straight forward and would require only 0.25 days globally. While the processing and consolidation of the two DCW layers would require one day. The estimated LOE to conduct further consolidation and then harmonization of the DCW layers with the VMap0-MP data layer globally would be estimated at three days. The potential naming of such features based on the DCW Annotation or Gazetteer, while possible, is not recommended as these data layers will already provide spatial referencing capabilities in addition to the provision of base map annotation or map labels.

Polygonal surface water body and wetland features

As summarized in Table 10.1.1 above, three data libraries were identified which contain polygonal representations of SWB features. All three of these sources attempt to provide names for at least some of SWB features. In particular reference to the fifth edition of the VMap0 Inland Water Body (IW) data layer, the NGA has only attempted to provide names for perennial SWB features. The 1:500 000 scale UNCS data library was again not available for direct comparison; however, comparisons between the 1:3 million RWDBII-Sv1.1 and the 1:1 million VMap0 for Africa indicate that the VMap0 and DCW are the only global source data libraries which contain some types of wetland features. Again, the DCW provides a more robust baseline of these features.

In regard to the processing of polygonal SWB data, an issue of critical importance is that features representing linear outlines of the water bodies also be included or processed in order to maintain connectivity with the generally broader linear river/drainage data features of a library. Based on the specifications covering the UNCS QID data effort and then the data layers comprised of linear SWB representations in the RWDBII-Sv1.1, at least some attempt has been made to provide such connectivity in these libraries. However, for the VMap0, additional processing is required to create linear representations of polygonal SWBs as a starting place. This process is further complicated by the need to include non-perennial and inundated land SWB features, and the fact that there is a lack of consistency regarding whether linear river features intersect, i.e. terminate at the SWB outline or pass through the feature. Further, with regard to the issue of connectivity, even with the inclusion of relevant SWB linear outlines, the ends of linear features representing rivers in the VMap0 Water Course (WC) data layer do not in all cases cleanly intersect SWB outlines, i.e. dangles and overshoots were found to be somewhat common.

The consolidation of the feature attribute encoding of the VMap0-IW layer is also recommended to facilitate both the creation of seamless data layers and to support base mapping. Specific to base mapping, it is also recommended that the polygonal DCW land cover, LCPOLY, layer containing the "Undifferentiated Wetlands" feature class dropped from the VMap0, also be considered for processing along side of the DCW Annotation and Gazetteer data layers.

Previous to the release of the VMap0.Ed5, the polygonal SWB features of the DCW constituted a more robust representation of SWBs than those contained in earlier editions of the VMap0. This assessment is based on a number of factors including: distortion of SWB features due to fuzzy creep; SWB features represented in the DCW but missing from the VMap0; features miscoded between the two data libraries; and then features which were shifted "correctively" or for no apparent reason in the VMap0. Combined, these factors argued in the past for the use of the DCW as a baseline for SWB features, which could then be updated based on any new SWB features included in subsequent releases of the VMap0. The harmonization of SWB features from the two data libraries, however, required a rather intensive process based on constant reference to the source ONC Charts. For Africa alone, after the processing of the baseline SWB data layers from each library was completed, their harmonization required an exhaustive 15 day LOE.

Fortunately, although the harmonization of the DCW/VMap0 SWB data layers resulted in a more comprehensive derivative polygonal dataset, a number of considerations now argue against such a harmonization. These considerations include: the likely prohibitive LOE required to create a harmonized SWB layer globally; corrections and coincident fuzzy creep within the VMap0-WC data layer such that linear river features are less likely to connect cleanly with the harmonized SWB outlines using the DCW as a baseline; the re-integration of previously missing SWB features and then further corrections/additions of SWBs in each subsequent release of the VMap0; the segmentation of SWB features representing lakes/reservoirs from "double-lined" river polygons to support naming in the VMap0.Ed5; and lastly, the harmonization of the coincident polygon boundaries instituted for the VMap0.Ed5 library from multiple data layers starting with coastlines and then proceeding with SWBs, built-up urban and peri-urban areas, physiographic features, and then the expanded vegetation layers.

However, as discussed earlier, the issue of the loss of the "Undifferentiated Wetlands" class between the DCW and VMap0 standards bears discussion. Although the VMap0-IW data layer contains two encoding attributes, i.e. the F_CODE field differentiating "Inland Water" from "Land Subject to Inundation" and the HYC field differentiating "perennial/permanent" from "non-perennial/intermittent" SWBs, based on processing of: the DCW; the VMap0.Ed3; and the VMap0.Ed5 for Africa, no combined Perennially Inundated Land features capturing the original DCW "Undifferentiated Wetlands" classification are represented. The inexplicable dropping of this feature class between the DCW and VMap0 libraries can be seen in Figure 10.1.1a below.

Figure 10.1.1a
Comparison of VMap0 and DCW wetlands for the Okavango Delta

As shown in the above figure, even with the addition of the Swamps-Vegetation data layer available for the first time in of the fifth edition of the VMap0, a representative feature classification of swamps, marshes, and bogs (i.e. "Undifferentiated Wetlands"), as depicted on the source ONC Charts, has been lost due to the change in standards between the DCW and VMap0. A further example of this can be seen in Figure 10.1.1b and related LOE processing discussion. In terms of the production of base maps, the loss of the "Undifferentiated Wetlands" class argues strongly in favour of processing the DCW-LCPOLY data layer in conjunction with the: Oil Fields, Mines, Physiographical, and Hydrological inventories presented earlier. It should be noted, however, that the integration of the DCW Wetland and Inundated Lands directly with the VMap0-IW data layer should no longer be considered in any development of a CGDB data library. Rather, what should be considered is the production of a separate feature dataset that can support the substitution of these features for the combined Non-Perennial/Inundated Lands class of the VMap0 for the purposes of base mapping.

LOE estimates for processing SWB data layers inventoried

The LOE associated with any processing of the UNCS 1:500 000 SWB data layer is expected to be negligible as long as the full source data library - rather than AOI subsets downloaded from an IMS server - can be made available.

The estimated LOE for processing and then combining the three SWB data layers of the RWDBII-Sv1.1, including: the removal of country boundaries potentially dividing SWB features; the integration and encoding of relevant SWB island features from two further RWDBII-Sv1.1 data layers, the maintenance of SWB and Island names; and the creation of "named" SWB linear outlines globally is five days.

The LOE for processing the VMap0-IW data layer is a bit more difficult to estimate and would be dependant on whether any corrective editing is to be undertaken; an issue that is exacerbated by the need to maintain any coincident boundaries for features between the coastal, urban, physiographic, and vegetation data layers. The straight forward seamless compilation of SWB polygonal features globally can be estimated at ten days and should include the creation of linear outlines retaining any polygonal name attribute. However, at a minimum, this LOE could be expected to increase by a minimum of three days for each of the six interim continental data subsets likely be created if any gross corrective editing were specified. Figure 10.1.1b depicts just one case where the editing of the VMap0 Inland Water and Water Course data layers might be called for.

Figure 10.1.1b
Example of possible editing required to VMap0 IW and WC data layers

The area covered in Figure 10.1.1b lies just upstream from the confluence of the Congo and Ubangi rivers in central Africa, and required editing to the polygonal SWBs representing the Sangha and Ubangi rivers, and both the polygon and linear features representing the Likouala and Aux Herbes rivers. This figure also provides a further example where the original DCW-LCPOLY "Undifferentiated Wetlands" class have been dropped between the DCW and VMap0 data libraries.

10.1.2 Limnological databases of surface water bodies

Four global databases containing limnological, as well as potentially fisheries, general environmental or anthropogenic factors pertaining to SWBs were identified for the inventory[26].

No samples of the data contained in these databases could be obtained for evaluation in conjunction with the inventory data review. Further, the availability of these data within the public domain and potential access to them in whole or part by the UN could not be established. However, depending on how the data are structured and made available, these databases could be prime examples where the UN could benefit from the international standard on "Information Retrieval: Application Service Definition and Protocol Specification", ISO standard 23950 or ANSI/NISO standard Z39.50. These standards allow for the direct retrieval of information from participating library databases over the Internet and are playing an increasingly larger role in reducing the duplication of data maintenance efforts and providing access to point specific or more general ancillary data and environmental statistics.

To date these standards are contributing to the analysis of terrestrial and aquatic species data by providing seamless access globally to museum samples or other library holdings; see LifeMapper, www.lifemapper.com or FISHBASE, www.fishbase.org for examples. Within the next two years it is anticipated that fairly simple routines will be available which will allow potential users to access spatial data such as an SWB via their GIS, click on a spatial feature and then access the Internet and potentially draw statistical, pictorial, and bibliographic reference data from institutional library archives across the globe relevant to the feature.

Table 10.1.2
Databases containing limnological and/or other ancillary data on SWBs globally

Data Type/Source

URL

Extent

Scale

Availability

Notes

LIMNOLOGICAL D ATABASES OF SURFACE WATER BODIES

Global Lake and Catchment Conservation Database (GLCCD)

Mullard Space Science Laboratory-WCMC- UNEP:
www.cpg.mssl.ucl.ac.uk

Global

Lakes >=100 km2

Public Domain

Data on ~1 400 lakes including: wind effects, freeze/thaw, type, water level/area, anthropogenic activities, & seepage.

WorldLakes & International lake environment committee (ILEC) SWB data servers

www.worldlakes.org
www.ilec.or.jp

Global

n/a
Lat/Long

Assumed to be Public Domain

ILEC contains data on ~500 SWBs in 73 countries, number of World-Lakes holdings n/a, ~212 in Africa

GIS WorldLake, Institute of Limnology, The Russian Academy of Science

No URL, Contact: DSc Sergey Ryanzhin <ryanzhin@peterlink.ru>

Global

n/a
Lat/Long

Unknown
Assumed to be Public Domain

Data cover ~32 600 limnologically studied natural lakes in 149 countries. Access and distribution as public data not ascertained.

FAO Lakes and River Fisheries Database

Contact: Dr. Jose Aguilar- Manjarrez <Jose.AguilarManjarrez@ fao.org>

Global

n/a
Lat/Long


3 294 water body features based on 757 reference publications

The first of the databases listed in Table 10.1.2 is the Global Lake and Catchment Conservation Database. This database was first made available in 1999, and was developed by a consortium of partners including: the Mullard Space Science Laboratory (MSSL), the World Conservation Monitoring Centre (WCMC), and the UNEP Division of Environment Information and Assessment. The database contains information on just over 1 400 natural and artificial water bodies worldwide based on three primary sources. These sources include: the MSSL Global Lakes Database (MGLD), the MSSL Remote Sensing Lakes Database (RSGLD), and the WCMC Lake Conservation Databases. The spatial baseline for the SWB features of this database was likely the DCW, supplemented by satellite monitoring data and published references. On-line access is that these data appear to have been recently archived, and UNEP may need to be approached to provide access.

The second database covered in Table 10.1.2 is actually a Web site associated with the WorldLakes Organization. The baseline for the data available via this site could not be determined and direct access to the database as a whole was not possible. Rather, users can search either by name or regional subsets. The total number of SWBs covered by the organization could also not be determined. However, for Africa, pictorial and statistical data was available for approximately 270 SWBs. The data available through its Web site are either based in part - or at least directly reference - the International Lake Environment Committee (ILEC) database of SWBs. The ILEC database was started in 1988 under the sponsorship of a UNEP project entitled, "Survey of the State of World Lakes". The ILEC database contains physiographic, biological and socio-economic data on some 500 natural and artificial SWBs from 73 countries.

The third database listed in Table 10.1.2 is entitled GIS WorldLake. The GIS WorldLake is being compiled by the Institute of Limnology of the Russian Academy of Science. The database builds on the ILEC database and data from variety of other sources that are difficult to summarize briefly. The database may also include data from the GLCCD database. The GIS WorldLake database contains information on some 32 600 naturally occurring lakes in 149 countries which have been limnologically studied. In addition, the database also holds information on some 4 600 limnologically studied artificial SWBs from 132 countries, and 140 limnologically studied wetlands/swamps from 30 countries. Despite repeated attempts over the past two and a half years, FAO has been unable to either determine the status of this database and whether it lies in the public domain, or to gain access to the data directly. The spatial data are in MapInfo format linked to Excel "tables".

The final limnological database identified for the inventory is the Lakes and River Fisheries Database. This database represents a consolidation of two existing databases detailing the characteristics of tropical SWBs and rivers. These two databases were developed under funding from the Fisheries Management Science Programme of the UK Department for International Development based largely on data contained in a wide range of published scientific literature. The follow-on Lakes and River Fisheries Database effort is funded primarily by FAO, and in addition to restructuring the two previously developed databases into a common structure, the resulting database has been expanded to provide coverage of major temperate SWBs and rivers based on further information gleaned again from published scientific literature.

Although consolidated into a common data structure, due to the differences in the water body and rivers data, the Lakes and River Fisheries Database has been divided into two sets of interlinked tables: one covering lakes, reservoirs and coastal lagoons; and the other, rivers and floodplains. Each of these sets is comprised of five individual tables covering: chemical and biological data, demographic data, fisheries data, hydrology and climate data, and morphological. The tables of each set can be linked based on the use of either an extant name or alpha-numeric attribute encoding fields. Similarly, each of the two tabular sets of data can be linked to a common bibliographic table cataloguing the publication source for each of the entries in the various tables.

Based on the technical specifications for the database, some 3 294 primary SWB or river features from 757 reference publications will be detailed, (MRAG, 1997). The degree to which it will be possible to spatially reference these features could not be determined. However, based on attribute fields enumerated in the specifications, at a minimum latitude and longitude coordinate information will be captured. It is anticipated that the FAO Lakes and River Fisheries Database will be made available sometime in 2005, with distribution subject to a non-commercial fair quotation EULA.

10.1.3 Satellite image derivative databases

Two sources of global satellite image based SWBs were identified for the inventory. The first of these is a fairly high resolution 1.4 ha or 28.5 m derivative of near global SWBs, while the second is a nominal 1 km derivative of land cover including water bodies. Other nominal 1 km resolution based land cover datasets including a "water" classification were identified, notably the International Geosphere Biosphere Programme (IGBP)-USGS Global Land Cover Characteristics Database. However, as these other raster datasets primarily use the SWB features from the DCW/VMap0 for the "masking" of water bodies, they do not represent an independent source or classification of such features.

Table 10.1.3 summarizes these two satellite image based SWB databases[27].

Table 10.1.3
Satellite image derivative surface water body databases (GVM-JRC, 2004)

Data Type/Source

URL

Extent

Scale

Availability

Notes

DERIVATIVE SATELLITE IMAGE BASED SURFACE WATER BODY D ATABASES

Global Land Cover 2000 Project (GLC-2000)

www-gvm.jrc.it/glc2000/defaultGLC2000.htm

Global

30as
~1 km

Public Domain

The GLC-2000 data use daily observations, from 1/11/99 to31/12/2000, of the VEGETATION sensor on the SPOT 4satellite.

EarthSat-NASA GeoCover Landcover LC, Water Bodies

www.geocover.com/gc_lc/ccaupdt_products/with_mmu.html

Near- Global

28.5 m to 1.4 ha

Presently Commercial

Possible source of high resolution water bodies, with FAO efforts to move datainto the public domain

Based on United Nation's ecosystem related international conventions, the Global Land Cover 2000 Project provides a year 2000 global reference database for environmental assessment. The thematic legend of the database has been built up using the FAO/UNEP Land Cover Classification System (LCCS). The use of LCCS has allowed the Project to have a 23 class worldwide consistent legend and much more detailed regional ones. Both the worldwide and regional legends are comparable with regional, national and/or subnational mapping projects, e.g. AFRICOVER, GLCN related mapping activities etc., which are using the FAO/UNEP LCCS classification system. The GLC-2000 is based on the VEGA 2000 dataset including 14 months of daily data acquired globally by the VEGETATION instrument onboard the SPOT 4 satellite. The GLC-2000 effort was sponsored primarily by: the Global Vegetation Monitoring Unit (GVMU) of the European Commission's Joint Research Centre (JRC), FAO, UNEP, the VEGETATION Programme (France, Belgium, Sweden, Italy and the European Commission), and a host of regional cooperators. These data are available in the public domain, free of charge for non-commercial applications.

Figure 10.1.3a
Coverage of the GLC-2000 Database

The GeoCover-LC product contains both raster and vector versions of land cover as well a water bodies specific database. Water Bodies represented in the land cover dataset have a resolution of 1.4 hectares, while those in the water bodies specific database retain the full 28.5 m "resolution" of the base Landsat TM imagery. As GeoCover-LC data were not available for review, Figure 10.1.3b derived from FAO's pending AWRD publication, uses satellite derived SWBs from the FAO-AfriCover Project and VMap0 data from the AWRD for comparison.

Figure 10.1.3b
Example of 28.5 m satellite derivative SWBs based on AfriCover

The centre image in Figure 10.1.3b is based on the GeoCover-Ortho data. As the satellite image data resulting from the NASA SDB contract (including base imagery for: LandSat MSS 80 m scenes covering a base year of circa 1980; LandSat 4/5 TM 28.5 m image scenes for circa 1990; and, LandSat 7 ETM+ 28.5/15 m data for year circa 2000), are now effectively in the public domain, it is possible that the water bodies specific database may also be made available. FAO has been lobbying NASA to also make these data available in the public domain.

10.2 DRAINAGE AND FLOW ROUTING DATABASES OR LIBRARIES

Three classes of river drainage or flow routing data were identified for the inventory. These classes include drainage features captured from: cartographic sources, derived from DEM sources, and then those containing river gauge or run-off data.

10.2.1 Cartographic source databases of river drainage

The data layers included in this topical subsection can be divided based on whether they use point or linear features to represent the spatial features. In all five data layers were identified as relevant to this topical subsection and are summarized in Table 10.2.1.

Table 10.2.1
Cartographic river drainage databases

Data Type/Source

URL

Extent

Scale

Availability

Notes

NAMED RIVER AND STREAM CONFLUENCE/OUTFLOWS AND RELATED DRAINAGE POINT

NGA-GNS/GEOnet Gazetteer

http://164.214.2.59/gns/html/index.html

Global

variable
1’ - 1"
dms
1:250 000

Public Domain

Database contains ~ 550 000 named locations of rivers, streams, canals, rapids, waterfalls, deltas, etc.

NGA-DCW Gazetteer and Annotation data layers

DCW available from 1992 original set of four CD-ROMs

Global

1:1 M

Public Domain

Layers contain ~29 000 drainage feature names captured from the original ONC charts, rivers can only be differentiated for Anno.

LINEAR RIVER DATA LAYERS

UNCS 1:5 million Named river data layer

UNCS data portal currently available not

Global

1:5 M

CR, FQ, PD

Possibly named rivers with six encoding attributes, with purported excellent connectivity

RWDBII-Sv1.1 Major, minor, other river layers

CD available from FA O or WHO

Global

1:3 M

Public Domain

Name attribute dropped for this edition, connectivity rated as good.

NGA-VMap0 Ed.5 Canal and named water course data layers

http://geoengine.nima.mil or available as a set of CD-ROMs via USGS-Store

Global

1:1 M

Public Domain

Largest scale source of rivers with 2 classifications; for Africa 17 040 out of 146 000 rivers are named

CR=Copyright; PD=Public Domain; FQ=Fair Quotation

Point sources of named drainage features

Two point data layers of named drainage features are listed in Table 10.2.1 and the LOE associated with the processing of the data is relatively low. In particular, as the NGA GNS and DCW gazetteers are recommended for processing in total, existing encoding in these databases allow for the rapid identification of named drainage features. However, as with the SWB features in the DCW gazetteer, rivers are included under a general "drainage" classification. Due to this, the various "levels" of the DCW-DNNET annotation again represent the only way to differentiate river name features correctly, supporting again the recommendation that the processing of the DCW annotation into definitive layers and then creating a more robustly encoded gazetteer are called for.

Unlike the DCW based features which are based on map annotation or labels, the river and stream locations in the NGA-GNS Gazetteer represent either the confluence or outflow point of rivers. The outflows can be either oceanic or to an inland SWB or sink. The NGA used both the DCW annotation and GNS databases as the baselines for the naming of river and SWB features in the fifth edition of the VMap0.

Linear river data layers

Table 10.2.1 also contains a summary of the layers from the three data libraries which contain linear river drainage features. With the exception of the 1:3 million RWDBII-Sv1.1, these sources attempt to provide names for the linear features. Based on the data specifications and e-mail communications related to the 1:5 million scale UNCS data library, the connectivity of the river features from this source will purportedly be excellent. In addition to a name attribute, the UNCS river data layer will contain attribute encoding as to: Perenniality; Rank including major, minor, or unranked hierarchy; Type specifying watercourse, canal, or SWB outline; Coincidence with international boundaries; and, then Country encoding. The river data layer of the UNCS 1:5 million data library was the only layer identified which includes SWB outlines as an integral part of the river drainage data layer.

The second potential source of river drainage data listed in Table 10.2.1 is the 1:3 million RWDBII-Sv1.1 data library. The name attribute for rivers was dropped from this release of the RWDBII-Sv1.1 due in part to the incomplete global coverage of this encoding. In both river and SWB features, the Version 1.1 ESRI Shapefile release of the RWDBII-Sv1.1 represents a significant improvement over the earlier World Data Bank II. The organization of the data in the library is however, based on the use of multiple layers, e.g. three for rivers, to create topical subsets. Because this hinders the use the data in the library for both base mapping and analytical purposes, it is recommended that the relevant river, SWB, road, administrative, etc. feature sets comprising each topical subset be integrated into composite layers where possible. Specific to the RWDBII-Sv1.1 rivers data, this would include the integration of SWB outline feature set to create a single layer.

Interestingly, the RWDBII-Sv1.1 is the only source of rivers data identified that also includes a linear feature representing a pseudo mid-line for double lined river polygonal features. Unfortunately, the linear mid-line only intersects to main course river features and does not consider tributaries as well. Assuming that the more extensive LOE for consolidating SWBs and islands has already been undertaken, the LOE for processing the RWDBII-Sv1.1 linear river data features globally and then integrating them with the SWB outlines can be estimated at three days. The overall connectivity of the resulting dataset would be rated as good, but a further LOE would likely be required to ensure connectivity.

The final sources of linear river drainage features represented in Table 10.2.1 are the VMap0.Ed5 Water Course (WC) and Canals data layers. Like the UNCS data layer, an attribute for a name has been added to this data layer for the fifth edition of the VMap0. Additionally, the connectivity between river and SWB features appears to be somewhat improved and some of the missing or dropped linear river features from the original DCW, i.e. rivers coincident with national boundaries and some of those missing from individual tiles, have been reintegrated into the database. However, even taking these improvements into consideration, some gaps still exist.

The sources for the names assigned to the VMap0.Ed5 linear river features were again based on the map annotations captured from the ONC Charts and the NGA-GNS Gazetteer data base. The process used for the encoding of names started with the assignment of annotations to the nearest linear river feature, and then continued using the connectivity between these features to trace "downstream" to any gazetteer location of the same/similar name representing either a river confluence or point of discharge. In cases where ambiguities were identified, no name was encoded. Otherwise, the set of linear features connecting the annotation and gazetteer points were encoded with the same name. Based on a preliminary examination for Africa and Europe, between eight and fifteen percent of the VMap0 linear river features have been encoded with a name.

A very limited number of the linear features in the VMap0.Ed5 Canal data layer have also been named. However, this data layer represents an example where the change from the DCW to VMap0 standard has resulted in the incorrect attribution of feature encoding. In this case, many of the features of the DCW-DNNET drainage layer encoded as canals, are now encoded as rivers in the VMap0-WC data layer.

The LOE associated with the straight forward seamless processing of the VMap0 WC and Canal data layers is estimated at five days globally and would include the integration with surface water body outlines as recommended in Section 10.1.1. However, if basic corrective editing is also to be considered, a minimum LOE of two days would need to be added for each of the six initial continental subsets expected to be produced. Figure 10.1.1b above and Figure 10.2.1 below provide examples where such editing should be considered.

Figure 10.2.1
Examples where editing of the VMap0 drainage layers should be considered

It should be noted that the examples presented in Figure 10.2.1 represent areas where some extensive editing would be required to either correct or complete the VMap0 drainage network layers. These areas are fairly easy to identify based on a visual examination of the ONC chart boundaries, and then the DCW and VMap0 tile boundaries. There are however, many hundreds of other corrective edits that might be required to ensure connectivity between drainage features which are less easy to identify. The correction of these latter errors is not covered under the LOE estimates above.

10.2.2 DEM derivative flow routing databases

Only one global source of drainage networks derived from digital elevation models (DEMs) was identified for the inventory. This is the network derived from the hydrologically filled DEMs created from the nominal 1 km GTopo30 source DEM in conjunction with the USGS/UNEP HYDRO1K (H1k) data effort. This effort was the first of its kind ever attempted, and the outputs of the effort included continentally specific: DEMs filled to remove unrepresentative sinks across the landscape; "river" networks based on flow accumulation models which effectively trace the routing of potential flows across the DEMs; and lastly, polygonal watershed models defining major river systems and various levels of subriver basins. The flow routing data layer resulting from the H1k data effort is summarized in Table 10.2.2.

Table 10.2.2
DEM derivative flow routing databases

Data Type/Source

URL

Extent

Scale

Availability

Notes

DEM DERIVATIVE RIVER NETWORK DATABASES

River Flow Routing Network from HYDRO1k

http://edcdaac.usgs.gov/gtopo30/hydro/

Global

nominal
1:2 M
1:3 M

Public Domain

The only connected river network globally available. Based on flow accumulation from 1 km DEMs.

As listed in this table, the connectivity of the H1k flow routing data layers is excellent and these data effectively represent the only available networked river data globally. The encoding attributes of this database are discussed under Section 10.4 and any LOE associated with processing these data should be undertaken in conjunction with the processing of the H1k Watershed data layer discussed in Section 10.3.3.

10.2.3 River gauge monitoring and runoff databases

Two sources of River Gauge Monitoring and Runoff data were identified for the inventory. These databases are based primarily on data from the Global Runoff Data Centre (GRDC) located in Germany under the primary sponsorship of UNESCO and the World Meteorological Organization (WMO). These databases are summarized in Table 10.2.3.

Table 10.2.3
River gauge monitoring and runoff databases

Data Type/Source

URL

Extent

Scale

Availability

Notes

DEM DERIVATIVE RIVER NETWORK DATABASES

River Gauge Monitoring Data, Global Runoff Data Centre

www.bafg.de/grdc.htm

Global

Point

Public Domain based on specific requests

As of 2001, the GRDC database contains datafor ~2500 daily and ~6 700 monthly station reporting

Simulated Gauge/Runoff Topological Network, Univ. of New Hampshire(UNH)

www.grdc.sr.unh.edu/

Global

Point, 0.5 grid

Public Domain

Observed river discharge data combined via water balance model to simulate composite runoff

Although, the GRDC is currently prohibited from providing access to the entire data archive, the Centre will entertain specific regional requests for access to available daily or monthly river station reporting data. Otherwise, a standard subset providing global coverage for 1 352 stations is available for download. The GRDC data are available in time series subject to the date monitoring was started for each station and the periodicity of reporting.

The UNH River Station and Runoff database listed in Table 10.2.3 provides coverage for some 6 960 river basins globally and includes monthly and annual runoff grids for Observed, Water Balance, and Composite runoff estimates. Although the grid data layers from this source use a 30 minute (0.5°) pixel size, these data are included in the inventory as the only source of potential runoff data identified. In addition to the grid based data, detailed point river station data are available for 660 gauges globally, while more general information is provided for 1 347 stations. A much generalized linear flow routing and grid of the river basins are also included in these data. The documentation of these data is however limited and examples of how the data can best be utilized are not detailed.

10.3 DEM DERIVATIVE WATERSHED AND RIVER BASIN DATABASES

Three global databases of watershed and river basins derived from DEMs were identified for the inventory. For the purposes of this report, the term watershed can be used interchangeably with the terms catchment or drainage basin. Table 10.3 summarizes the three watershed (WS) databases identified for the inventory and categorises them based on whether they are a simple WS delineation or a more complex WS model[28].

Table 10.3
DEM derivative watershed and river basin databases

Data Type/Source

URL

Extent

Scale

Availability

Notes

SIMPLE WATERSHED DELINEATIONS

W atersheds of the World, World Resources Institute

www.wri.org/ watersheds as processed by Rutgers University, USA

Global

~250 000 cell Delineation

Public Domain

Database contains 297 river megabasins globally detailing threats to species diversity.

Global International Water Assessment

GIWA: www.giwa.net and University of Rhode Island, USA

Global

~150 000 cell Delineation

Public Domain

2 955 river basins & large marine ecosystems categorized by ~68 regions globally.

COMPLEX WATERSHED MODELS

HYDRO1k Global W atershed Model

http://edcdaac.usgs.gov/gtopo30/hydro

Global

~4 000 cell Model

Public Domain

The only globally available watershed model. Model has5-6 levels topologically encoded & linked to a river flow network.

Although there is no definitive terminology used with regard to the relative "resolution" of watershed data derived from DEMs, the coarseness or fineness of the boundaries delineated is a function of the pixel size of the base DEM and then the minimum number of pixel cells required for a watershed to be defined during processing. For the purposes of comparison, the resolutions listed in Table 10.3, are based on a relative comparison of the source databases to various WS delineations processed from the GTopo30 DEM for Africa.

Figure 10.3 below, displays a comparison of the above three databases. Due to the resolution of the 6-level H1k WS-Model, the Namibian AOI is used to show the relative level of detail available from this data library. The depiction of the H1k WS-Model in Figure 10.3 also includes a representation of the linear Flow Routing data layer from this library.

A fourth WS-Model is also shown in Figure 10.3. This graphic uses the nominal 5 000 cell 3-level ALCOM-WWF WS-Model of Africa to approximate the level of detail which would be available from a global intermediate resolution 3-level WS-Model based on the specifications described in an FAO report entitled Suggested Hydrological Standards & Base Mapping, (Dooley, 2003). This report was prepared for the FAO Spatial Standards and Norm Task Force in the spring of 2003. Although, the information contained in this report is now slightly dated and requires revision based on the availability of SRTM DEM and "named" or at least partially "connected" linear Rivers/SWB data layers of the VMap0.Ed5, it provides an overview of the issues and complexity involved with the creation of a global 3-level WS-Model that could be consistent with either the VMap0 as a cartographic baseline or a flow routing dataset addressing at least some of the discrepancies of the H1k highlighted in Section 10.3.3.

Figure 10.3
Comparison of globally available or proposed WS delineations and models

10.3.1 World Resources Institute (WRI) Watersheds of the World

As can be seen in Figure 10.3, the WRI Watersheds of the World dataset represents a less complex delineation of river systems at their largest extents, i.e. so called MegaBasins. Due to this, only one level of encoding is possible. Also, many areas of the world are left blank and are not covered by this delineation. For the river and larger megabasins which are delineated, attribute encoding estimating: river fragmentation; the number of dams and RAMSAR sites; available water; the proportion of wetlands and grasslands, arid lands, irrigated and crop lands, forested lands, and peri-urban areas; population density; a name, and others factors are summarized. These data are available in separate tables and an estimated one day LOE would be required to conduct basic topological checks on the spatial data, and to then process, consolidate, and link the tabular attributes to the spatial features globally.

10.3.2 GIWA Terrestrial and Large Marine Ecosystems

The Global International Water Assessment (GIWA) database listed in Table 10.3 and shown in the above figure provides what is effectively a global scale river basin delineation that extends from the continental landmasses to include large marine ecosystems. Although in some few cases, e.g. the Congo, Amazon, and Mississippi Megabasins, major component subriver basins are differentiated, this dataset is rated as a watershed delineation versus a watershed model in the inventory. Given that multiple megabasins and river systems contribute inland flows to the LME Regions - outlined in Figure 10.3 using different colour shading - this rating may undercut the potential utility of these data for at least some types of modelling or analysis. At the time this data effort was reviewed for the inventory, attributes other than a name and regional codes for the sixty-eight GIWA-Regions were not available. Given the use of a regional coding scheme, it could be expected that only a 0.5 day LOE would be required to link or consolidate any future ancillary attribute data to the spatial dataset.

10.3.3 Hydro1k six-level watershed model

The 6-level H1k WS-Model represents both the highest resolution and only globally consistent WS-Model identified for the inventory. The H1k data library uses the Pfafstetter Topological Encoding Scheme (PTES) to ensure a topological linkage is possible between the linear river routing, point nodes representing spill or outflow points, and the polygonal watershed model data layers. SWB features are not included or integrated into the H1k data library, and there is no name attribute associated with any of the layers of the library. With training, the PTES encoding allows for the differentiation of upstream contributing flows and then downstream areas from any point, i.e. an individual watershed, within the model.

In Figure 10.3, colour shading is used to outline at least some of the major river systems within the Namibian AOI based on the PTES level 3 encoding. Depending on the complexity of the river system in question, different levels of the model would be required to similarly define contributing sub-river basins. While outside of the Namibian AOI, a good example of this would consider the Blue Nile River. The Blue Nile river basin can be wholly described in the H1k WS model using the PTES level 2 encoding. Whereas in comparison, at this level the larger the more complex White Nile River basin has already been broken into sub-basins comprised of contributing watersheds with a combined areal extent similar to the Blue Nile. This highlights perhaps the major limitation of the PTES and the use of the H1k WS-Model for base mapping, i.e. that the PTES is not readily in sync with historically name based hydrological classifications and the boundaries of watersheds do not conform to river and SWB data derived from cartographic sources. Another limitation might be that resources were not expended in the effort to correct the WS delineations and flow routing across the landscape based on errors in the source GTopo30 DEM. This can be seen in the inset graphic at the right, for the area surrounding Etosha Pan within the Namibian AOI.

The LOE associated with processing the H1k WS-Model is fairly straight forward and should be accomplished in conjunction with the processing of the linear Flow Routing data layer to ensure the topological linkage between the data layers is maintained. The estimated LOE for globally processing both of these data layers, is seven days, and would include: the projection of the data into a common decimal degrees reference datum; edge matching where required; a review of the topological encoding and overall connectivity; and finally, the hard-coding of a unique downstream identifier and the addition of flow orientation encoding to simplify analytical usage both inside and outside of the GIS environment.

Based on e-mail communications with one of the project leaders of the H1k data effort at the USGS-EDC in early 2004, although proposals have been submitted to NASA and other USG agencies, no sources of funding had been authorized to correct the H1k or to derive higher resolution WS models globally based on the SRTM DEM data discussed in Section 8, (EDC, 2004).

10.3.4 Creation of an intermediate resolution global WS-model consistent with the VMap0

Due in large part to the inability to cross-reference name based hydrological references, its non-conformity with cartographic river and SWB sources, and the lack of corrective editing undertaken for the H1k effort, there is a perceived interest for the development of a global 3-level intermediate resolution 5 000 to 7 500 cell WS-Model. In particular, a WS-Model that would be consistent with source cartographic river and SWB layers which have been “corrected” and harmonized with a DEM to produce representative hydrological layers and surfaces suitable for regional and continentally based analyses.

While the detailing of the tasks involved in the creation and subsequent editing required to create such a WS-Model are beyond the scope of this report, it is important to recognize that the processing of such a model from a source DEM(s) represents the least-cost approach to the creation of a WS-Model. The most time consuming tasks involved with the development of a more robust WS-Model would include: the potential joint harmonization of the source DEM and any cartographic baseline of surface hydrology; the post-process editing of the resulting watershed boundaries; the addition of a name and other encoding attributes to facilitate the analytical use; and then, further editing of - at a minimum - the main stem and branches of the linear hydrological network to ensure basic connectivity. Given these factors, the LOE required to produce a single "continental" WS-Model can be estimated at a minimum of 45 person days. The majority of this LOE would be spent on post DEM editing and ensuring the connectivity of the model.

10.4: CONNECTIVITY AND ENCODING ISSUES RELATED TO DRAINAGE NETWORKS AND WS-MODELS

The issues related to the connectivity and encoding of linear drainage networks and WS-Models are perhaps some of the most complex related to any of the data discussed in the inventory. As these issues potentially affect the representation of such features for the purposes of base mapping and then the analytical use of hydrological data layers, they bear at least a summary overview within this report[29].

10.4.1 Connectivity

In the GIS environment, most commonly a river network is composed of both line and point feature data types, while river catchments and watershed networks are defined using a polygon feature data type. The encoding of such "network" features engenders both stringent requirements and opportunities that go beyond the generic encoding of hydrological features which essentially can be viewed as standing alone. Conceptually the term network implies "connectivity" such that, in the case of hydrological data, the flow or drainage from one point within a network or model can be traced or inferred to any point downstream. In this manner, assuming connectivity, it becomes a relatively simplistic matter to depict all downstream areas that might be impacted by an upstream change in land or some other management practice. By corollary, it also becomes possible to identify all upstream areas which may need to be considered for their potential impact on a particular project design or evaluated based on some set of constraints.

10.4.2 Methods used to encode hydrological networks

There are predominantly two methods employed for the codification of hydrological networks: ranking schemes and topological encoding schemes. Ranking schemes are used as a means of comparing rivers of different size and importance within or between networks (Dunne and Leopold, 1978), while topological schemes are used to establish the linkage within and then between: river networks; broader watershed or river catchment networks; and, feature layers associated with important river confluence's, points of discharge, and/or monitoring stations. Ranking schemes are usually applied to linear "river" layers or flow networks, while topological encoding schemes are usually associated in some manner with the development of GIS databases and the integration of linear river networks and watershed models.

10.4.3 Ranking or classification schemes

As a group, ranking schemes can be referred to as methods for classifying the stream "order" or "magnitude" within, and again between, river networks. It is generally recognized that there are four potential schemes for classifying the rank of streams: the Horton method, the Strahler method, the Shreve method, and the Scheidegger method. Ranking schemes are applicable to both hydrological analysis and for the cartographic representation of river networks on base maps. The two methods most commonly employed are the Strahler Stream Order and the Shrive Stream Magnitude ranking schemes. Assuming a validated "connectivity" between the linear features of a rivers network, the post encoding of the network to conform to either method is a fairly straight forward programming task. This task and subsequent analyses both inside and outside of the GIS environment can be greatly facilitated by the hard coding of a downstream identifier attribute based on a unique numerical ID for each feature in the linear network.

Only one globally available river/flow network data layer was identified for the tabular inventory. This is the linear Flow Routing/Accumulation "rivers" data layer of the H1k data library. This data layer contains feature attribute encoding according to the Strahler stream order ranking scheme. By convention, stream ordering classifications are not commonly applied to watershed networks or models. However, in the cases where a topological encoding scheme has been used to link a river network to a watershed model, the transfer of such attributes can be accomplished between the tabular databases without the need for GIS spatial analysis. For hydrological data, the most widely recognised topological encoding method is the Pfafstetter topological encoding scheme (PTES). The PTES was originally introduced as a part of the H1k effort.

10.4.4 Topological encoding schemes

Coincident with the development of GIS systems which can be used to capture, maintain, and then support the analysis of hydrological data, significant efforts have also been undertaken to develop methods for the systematic codification of such data. The primary focus of these development efforts has been the codification of various standards which can be used to both describe the flow routing amongst the features of an individual river network or watershed model, as well as providing a mechanism for establishing a direct correlation between: linear flow networks, polygonal watershed networks, and node/point source databases supporting flow routing and/or environmental monitoring stations.

In the absence of a standard encoding schema between these feature types, a GIS can be used to both establish and then maintain the topological linkage within and between the various feature data types. However, because the GIS will use unique feature identifiers based on internally generated numeric ID values for both analysis and reporting, results may be difficult to interpret and summarize for tabular reporting outside of the GIS environment.

Table 10.4 provides a comparison of the major positive and negative aspects of potentially adopting a topological encoding standard.

Table 10.4.4
Comparison of benefits and limitations of topological encoding schemes

Benefits of implementing a topological encoding scheme

Limitations of topological encoding schemes

1) The direct correlation of polygonal watershed, linear river, and point river confluence or monitoring station features is facilitated based on common values employed to encode features between layers.
2) The topology of the hydrological network, both within and between feature data layers, can with training be viewed at a glance directly from the tabular attribute data.
3) The topological encoding can be used to conduct analyses directly from the tabular data, allowing standard processing tasks to be completed both more efficiently and cost effectively.
4) Because topological encoding schemes are by their nature hierarchical, results are again more easily interpreted and summarized for hardcopy tabular reporting.

1) Since topological encoding is numerically based: it and may not conform to existing naming or cartographic conventions; and may distract from efforts to attribute an actual name effectively limiting rather than enhancing user efforts to identify a specific area of interest.
2) The use of such encoding is redundant once the connectivity between features classes has been established, i.e. the hard-coding of unique downstream identifier.
3) Although topological encoding standards facilitate hydrological analysis outside of a GIS and allow the effective summarisation and sorting of results, references in the literature aside, most casual users find the codes difficult to interpret without extensive training and then constant reference to supporting documentation.

CHAPTER 11. SATELLITE IMAGERY, ORTHORECTIFIED MOSAICS, LAND COVER AND VEGETATION DATA

Unlike many of the vector data libraries discussed earlier, there is not as great of a potential overlap for the inventory of raster satellite image databases. Further, substantive progress has been made in the last two year towards the availability of satellite imagery which has undergone extensive value-added processing.

For actual satellite imagery, this availability includes full band imagery of individual satellite scenes and/or mosaics at pixels resolutions of 1 km, 500 m, 80 m, 28.5 m, and 28.5/15 metres. While for land cover and vegetation data, derivative 1 km data, 500 m and higher resolution data are or very soon will be available. Both the imagery and derivative land cover data are available for multiple time periods.

Due to time constraints, only a brief summary of satellite mage data could be prepared for this section and the inventory of derivative land cover and vegetation could not be completed as fully as the utility of these data warrant.

11.1 SATELLITE IMAGE MOSAICS AND ORTHORECTIFIED IMAGERY

Since the middle of 2001, an increasing amount of satellite image mosaics and orthorectified imagery has started to become available. For the first time, these data put value-added derivative versions of fairly high resolution satellite imagery into the public domain. Due primarily to the USG as the source of funding and various international "buy-ins", these data may be distributed unencumbered by copyright or licensing restrictions. The satellite image data identified for the inventory are listed below and summarized in Table 11.1. In addition to these data, UNCS has acquired NGA's full archive of 10 m digital ortho imagery (DOI10). However, although these data are in the public domain, because they currently provide only consistent coverage of Europe, the Middle-East, Southeast Asia, and Korea, they are not inventoried.

The time periods covered by the data identified for the inventory include:

The five LandSat based image data sources listed in Table 11.1 comprise what can be termed the NASA LandSat Orthorectified Image Library (LOIL). Each of the image sources in this library have been orthorectified to a common set of control points to enable direct comparisons between the locations over the approximate 30-year time period covered by the three sets of reference imagery. When imagery has been orthorectified, it is systematically "flattened" to account for vertical relief and corrected to match known ground control points. After the orthorectification process, the LandSat 4/5 and seven other data sources have a purported positional accuracy of better than 50 metres, while the accuracy of the MSS imagery is considered better than 100 m.

Table 11.1
Satellite image mosaics and orthorectified imagery

Data Type/Source

URL

Extent

Scale

Availability

Notes

SATELLITE IMAGE MOSAICS

NASA/USGS-EDC MODIS Image Mosaic (Year ~2001)

ftp://visibleearth.nasa.gov/pub/earthviz

Global

30as
nominal 1 km

Public Domain

An 2.5d enhanced global MODIS image mosaic of 30 m image scenes generalized to ~1 km

NASA/USGS-EDC Monthly MODIS Image Mosaics

http://glcf.umiacs.umd.edu/data

Global

250 m
500 m
1000 m

Public Domain

Monthly mosaics 11/2000 - 6/2003, and potentially onward band 1-7 & NDVI500 & 1000 m, with more recent 250 m band 1-2 & NDVI

NASA/USGS-EDC ~1990 LandSat-4/5 Orthorectified (OR) Image Mosaics as RGB images

http://edcdaac.usgs.gov/dataproducts.asp
http://glcf.umiacs.umd.edu/data
https://zulu.ssc.nasa.gov/mrsid

Near -Global

28.5 m

Public Domain

~6 x5 mosaics of image scenes in RGB GeoTiffor MrSID format based on bands 7, 4, 2 respectively

NASA/USGS-EDC~2000 LandSat-7 OR Image Mosaics

http://edcdaac.usgs.gov/dataproducts.asp http://glcf.umiacs.umd.edu/data

Global

14.25 m

Public Domain

RGB mosaics as above, scheduled for availability in June 2004

FULL BAND ORTHORECTIFIED SATELLITE IMAGERY

NASA/USGS-EDC~1980 LandSat-MSS Image Scenes

http://edcdaac.usgs.gov/dataproducts.asp
http://glcf.umiacs.umd.edu/data

Near -Global

57 m

Public Domain

2 visible and 2 near-infrared bands of orthorectified imagery

NASA/USGS-EDC~1990 LandSat-4/5 T M Image Scenes

http://edcdaac.usgs.gov/dataproducts.asp
http://glcf.umiacs.umd.edu/data

Near -Global

28.5 m
114 m

Public Domain

3 visible: 1 near-infrared and 2 mid-infrared at 28.5 m, and 1 thermal-infrared at 114 m orthorectified

NASA/USGS-EDC~2000 LandSat-7 ETM+ Image Scenes

http://edcdaac.usgs.gov/dataproducts.asp
http://glcf.umiacs.umd.edu/data

Global

28.5 m,
14.25 m,
57 m

Public Domain

3 visible: 1 near-infrared and 2 mid-infrared at 30 m, 1 thermal-infrared at 60 m, panchromatic 15 m orthorectified

NASA/USGS-EDC Aster Image Scenes {Not Orthorectified}

http://edcdaac.usgs.gov/dataproducts.asp

Near -Global

30 m

Public Domain

Extent of archive currently not known, expected global by 2005

Both the UNCS and FAO-SRDN have currently acquired access to the full archives of the LandSat full-band orthorectified imagery scenes. Although, the archives of both organizations contain the LandSat-4/5 TM image mosaics in GeoTiff or MrSID formats, it could not be determined if the LandSat-7 Enhanced Thematic Mapper Plus (ETM+) circa year 2000 image mosaics are as yet included[30]. According to UNGIWG reports, UNCS also expects to add the full NASA/USGS archive of available MODIS-ASTER multispectral imagery to their archive.

The Imagery listed in Table 11.1 are in a variety of formats. The image source data listed as containing full band imagery employ either the GeoTiff or BSQ format and use separate files for each multispectral channel. The LandSat image mosaic source data are available in either an uncompressed GeoTiff or MrSID compressed format for RGB imagery, using the bands 7 4 2 respectively. It should be noted however, that even using the MrSID compression format, the LandSat-4/5 TM Mosaics for Africa alone require eight CD-ROM disks or 5.2 gigabytes. Also that for Africa alone, the full archive of all the imagery listed in Table 11.1, again for Africa alone, is comprised of over 500 gigabytes of data, requiring roughly 0.6 terabytes of storage.

Figure 11.1 depicts the availability of the LandSat orthorectified image data sources available from the University of Maryland's Global Land Cover Facility which contains an on-line distributional archive of all but the first and last datasets listed in Table 11.1. FAO is in the process of making these data available via the GeoNetwork spatial data portal as well as via a seamless server.

Figure 11.1
Coverage extents of available LandSat orthorectified imagery

As depicted in Figure 11.1, there are wide areas of the globe where the LandSat MSS data were not available for inclusion in the NASA-LOIL. These gaps are attributed to the degradation of the MSS tape storage media from the Brazilian LandSat ground control station. Similar problems also resulted in a lack of coverage for the LandSat-4/5 TM Image Scenes and OrthoTM Image Mosaics over eastern Siberia. Both the LandSat-7 ETM+ full band Image Scenes and RGB Mosaics will provide global coverage for these areas, (Compton et. al. 2004).

To facilitate access to NASA's LOIL in developing countries, USGS has started to distribute the library to selected regional remote sensing centres. This effort will not however, of necessity, help researchers and potential users outside of these centres. UNCS on the other hand, has undertaken the potential distribution of the NASA-LOIL to interested parties or institutions using inexpensive 250-500 gigabyte "snap-servers" to bulk load and deliver the data. In addition to FAO's seamless server, the UN may want to consider this approach for making these data available to field offices and cooperators on a country-by-country basis.

Given the many possible applications for the data comprising the NASA-LOIL, other then the issues discussed in Section 11.1.1 below, no LOE estimates can be made in relation to any of the data listed in Table 11.1.

11.1.1 Issues related to the use of NASA's Orthorectified Image Library for base mapping

With the exception of the first two data sources listed in Table 11.1, all of the other image data inventoried have been projected into specific Universal Transverse Mercator (UTM) projection systems or zones. Each UTM zone covers 6° of longitude and effectively limits the extent to which the imagery can be organized for the creation of standard and seamless base maps. The boundary index of the OrthoTM image mosaics covering Africa can be seen in the inset graphic at the right. In this figure it can be seen that the OrthoTM data appear to have a symmetry to their orientation from both the north to south, as well as from the east to the west. What is not apparent however is that each of the north/south belts shown represent a different UTM projection. When data are in a UTM projection, it is possible for them to align with other data in the same UTM projection seamlessly from north to south, but unfortunately from east to the west - across UTM zones - the data will not align properly for the purposes of seamless base mapping.

Given the fairly high resolution of the imagery contained in the NASA-LOIL, these data can be used as image backdrops to support topographic base mapping at scale of 1:250 000 or larger. The maximum viewing scales for these imagery can be approximated as follows: LandSat-MSS Imagery, 1:250 000; LandSat-TM Imagery, 1:130 000; and, LandSat-ETM+ Pan-Sharpened Imagery, 1:75 000. As large scale topographic base maps commonly use the UTM projection, the imagery comprising NASA's OLIL is already in a suitable projection system for this use. However, there may be some reasons for the UN to consider the processing of these data into a projection system, or systems, which will support seamless base mapping.

For example, a reference was made at the 4th Annual UNGIWG Conference to examine the potential utilization of OrthoTM data for providing a homogeneous baseline, or master ground reference, for the demarcation of political boundaries, (SALB, 2003b). This potential application, as well as the need to produce country base maps at various scales may argue for the selective projection of imagery from the NASA-LOIL into either Decimal Degrees or continentally specific Lambert Azimuthal Equal Area projection systems. The least-cost method for facilitating the use of imagery from the NASA-LOIL for seamless base mapping would be to simply project the existing RGB OrthoTM image mosaics into the projection system chosen. This method has the advantage of not requiring any mosaicing during the processing, as the existing mosaics can simply be projected into seamless "tiles".

Assuming, the GeoTIFF formatted mosaics are used as a baseline, the LOE for projecting the LandSat-4/5 mosaics globally could be estimated at eight days. Given the 14.25 m pan-sharpened resolution of the LandSat-7 ETM+ image mosaics, a similar projection of these data would be estimated at a minimum of ten days. Unfortunately, the contrast enhancement and colour-balancing which was undertaken to ensure that the various source image scenes comprising each mosaic match-up radiometrically at the seams within and then between each mosaic, does not always result in attractive background images for display or use on base maps. This can be seen in Figure 11.1.1.

Figure 11.1.1 depicts the area surrounding the town of Karibib within the Namibian AOI. In addition to portraying the relative scales at which the imagery within NASA's OLIL can be displayed, this image also shows both the "raw" and then a post-processed version of the RGB OrthoTM image mosaic data. Although, the post-processing of hue and colour balance to the OrthoTM imagery required the use of professional imaging software, the result is a more realistic representation of both the vegetation and landscape surrounding the town. The addition of this extra step in the process of creating image mosaics suitable for seamless base mapping will however, increase the estimated LOE from eight days globally to four to five days per continental "mosaic" for the LandSat 4/5 mosaics alone.

In conjunction with the development and review of the CGDB inventory, the UNCS identified a product under development by EarthSat Corporation entitled “Natural View”. This product uses the 15 m panchromatic band of the LandSat 7 orthorectified data to sharpen and produce a more realistic, i.e. natural looking, version of the circa 2000 OLIL imagery based on a combination of the 28.5 m ETM+ bands. The results of this commercial effort will be similar to the colour correction process described above for the LandSat-4/5 mosaics. Although this effort would likely result in better control and standardization over the outputs, further analysis as to the cost effectiveness of purchasing a commercially licensed product versus some concerted effort to process these data in the public domain would be required. UNCS has indicated that the Natural View product will be purchased selectively in support of UN Peace Keeping efforts.

A number of Internet based satellite image viewing and GIS sites came on-line during 2004 subsequent to the preparation of the inventory. These sites provide a more up-to-date reference base concerning the availability of baseline or potentially improved editions of the satellite imagery discussed above. Although not inclusive, and in addition to the University of Maryland site listed in Table 11.1 the on-line sites include: the NASA portal, http://onearth.jpl.nasa.gov/landsat.cgi; the USGS Portal, http://glovis.usgs.gov, and the NGA portal, http://earth-info.nga.mil.

Figure 11.1.1
Examples of NASA Orthorectified Image Library

11.2 SATELLITE DERIVATIVE LAND CLASSIFICATION AND VEGETATION DATABASES

As discussed previously, time constraints prohibited the completion of this section of the inventory. The following databases were, however, identified as the three most prominent derivative global databases which might be evaluated for future consideration. Although derived for the most part from composite normalized difference vegetation indexes (NDVI) of the satellite image baselines above, the actual NDVI image composites and mosaics used as inputs would demand consideration under a separate interim section of the inventory which time constraints also prohibit.

CHAPTER 12. CLIMATIC DATA: TEMPERATURE, RAINFALL, AND ATMOSPHERIC-EMISSIONS

Similar to the above section on Satellite Derivative Land Classification and Vegetation Databases, time constraints also limited as detailed a consideration of potential sources of climatic data as potentially warranted in the inventory. Although as yet not finalized, subsections outlined as necessary to complete this topical section would include:

12.1 Temperature databases

12.1.1 Ocean temperature databases
12.1.2 Terrestrial temperature databases

Monthly normal temperatures, recent 30-year averages

12.1.3 Surface water body temperature databases

12.2 Rainfall databases

12.2.1 Time-series
12.2.2 Climatological

12.3 Atmospheric and potential evaporation databases

12.3.1 Solar radiation or sunshine indexes
12.3.2 Air moisture or dew point
12.3.3 Windspeed
12.3.4 Potential evaporation

12.4 Global emissions databases.

12.4.1 Greenhouse gases
12.4.2 Carbon and sulphur dioxides
12.4.3 Methane and chlorofluorocarbons

Taking the above limitation into account, in order to provide at least an overview of climatic data within the inventory, in place of addressing the topical divisions listed above a summary of perhaps the most often cited source of integrated climatological data is provided. This section is followed by a brief outline of a point meteorological station database and modelling toolset sponsored in part by FAO and detailed in Appendix B.

12.1 INTEGRATED CLIMATOLOGY DATABASE

The source of integrated climatological databases selected for discussion in the inventory is the Climatic Research Unit (CRU) of the University of East Anglia in the United Kingdom. While certainly not the sole source of such data, data either directly sourced or derived from the CRU is often cited in data compendiums focused on land, water, and agricultural modelling. In addition, the CRU works closely with a number of other institutions globally and is a primary partner in the World Meteorological Organization and UNEP coordinated Intergovernmental Panel on Climate Change (IPCC) and more specifically the IPCC Data Distribution Centre (DDC). The IPCC-DDC is administered by the CRU in joint cooperation with the Deutches Klimarechenzentrum (DKRZ/MPI) in Germany, and the Center for International Earth Science Information Network (CIESIN) at Columbia University in the US, http://ipcc-ddc.cru.uea.ac.uk/. The data centre is in part hosted by the CRU and is the result of recommendations made by the IPCC Task Group on Data and Scenario Support for Impact and Climate Analysis.

The IPCC-DDC appears to distribute a smaller range of data at a more aggregate pixel resolution, than those data made available directly via the CRU high-resolution gridded data archive located at www.cru.uea.ac.uk/cru/data/hrg.htm. From this URL, a table outlining the various editions of global gridded Climatological, Time-Series and Global Climate Change Scenario databases available from the CRU can be accessed. From this table, the 0.5° degree (55 km) and 10’ minute (18.5 km) databases that are based on timeframes reflecting the available historical record, specific time-series or projections can be either researched or downloaded. Of specific interest to this section of the inventory, are the currently available 10-minute climatological database, CRU CL-v2.0, based on the 1961-1990 historical record and the 0.5° time-series database, CRU TS-v2.0, recording data from 1901 to 2000. The CRU CL-v2.0 database contains monthly grids of: precipitation, wet days, mean temperature, mean diurnal temperature range, relative humidity, sunshine, ground frost, 10 m windspeed, and elevation with grids for vapour pressure and cloud cover still pending release of CL-v2.1. The CRU TS-v2.0 database contains either 100 or 10 year compilations for each of the first three climate variables listed above, as well as the two pending variables, as separate grids; station data files are also available for this database by time-series.

Both the CRU CL and TS databases are delivered as generic compressed ASCII flat-files. The LOE for processing the CRU CL database into ESRI Arc-Grid format would not be expected to exceed two days. However, for the TS database, given the number of individual grids comprising the 100 year time-series, the LOE would likely lie between three to five days.

12.2 THE LOCCLIM V2.0 METEOROLOGICAL STATION DATABASE AND TOOLSET

Further to any future review of other existing sources of climate data that may address the more detailed topical climate data headings listed in the introduction to this section, the following section was prepared to provide a discussion of the robustness and direct applicability of gridded data at higher resolutions than the CRU five degree and ten minute databases already provide. Specifically, some assessment of one of the major factors constraining the production of more robust gridded data, i.e. the current lack of input data of consistent quality and/or spatial distribution upon which higher resolution gridded outputs could be based.

Appendix A to this report outlines the pending publication of a point data archive or baseline of climatic data that also includes an interpretive geospatial toolset. Together this data archive and toolset have been named “New LocClim”. New LocClim is currently in beta testing and is expected to be jointly released by FAO/SDRN in collaboration with the Global Precipitation Climatology Centre of Deutscher Wetterdienst (the German Weather Service) by the time this manuscript is published. New LocClim, or LocClim v.2, constitutes the convergence of FAOCLIM Ver.2 with LocClim Ver.1. The LocClim v.2 toolset includes all of the standard interpolation methods, e.g. IDW, Kriging, Shepard, Thin-Plate Splines, etc. and a number of added functionalities. One of the most useful of these is the possibility for users to interpolate their own climate data and to prepare output grids at any spatial resolution.

Additionally, LocClim v.2 can also be used in conjunction with FAO Crop Specific Soil Water Balance models to produce a number of outputs such as water balance variables such as soil moisture versus actual evapotranspiration over the vegetative phase, water stress at flowering, and other variable outputs. These outputs can be “mapped” separately for the purposes of crop monitoring or to create composite derivatives such as agroclimatic databases using the Satellite Enhanced Data Interpolation (SEDI) methodology. SEDI takes advantage of the correlation between point data references, e.g. monthly temperature and an “environmental variable” that is available as a grid, for instance a digital terrain model. Other examples include point rainfall with a satellite-based cloud index, or crop yield with Normalized Difference Vegetation Indices.


[26] A fifth database published by the International Commission on Large Dams (ICOLD), www.icold-cigb.net, was identified but access to these data could not be established within the timeframe of the inventory. The ICOLD 2003 World Register of Dams contains information on 33 105 dams based on 4 108 published references. The database is copyrighted with distribution limited to licensed users based on a fee of 180 euros.
[27] It should be noted that in addition to the three polygonal data sources identified in Section 10.1 and the two satellite based sources of SWB listed in Table 10.1.3, it may be possible to derive polygonal representations of SWBs having a minimum length of 600 m and rivers greater than 185m wide from the newly released corrected or finished SRTM 1as and 3as near-global DEM databases. This facility could not, however, be evaluated within the timeframe of the inventory.
[28] A WS model can be distinguished from a simple WS delineation by attribute encoding that allows for the differentiation of upstream and downstream flow regimes in regards to any specific WS within the model. Implicit to such a differentiation, a WS model would by definition contain levels, whereby the attribute encoding for each watershed allows the determination or topological referencing of broader river systems or megabasins. Using a three tier/level model as an example and a watershed surrounding Lake Victoria in eastern Africa as a base, it should be possible to determine that the Lake Victoria watershed (Lvl-3), is part of the broader White Nile River Basin (Lvl-2), which is in turn part of the Nile Megabasin (Lvl-1).
[29] The following summary is based on information originally compiled for Section 1.2: Issues related to codification standards for hydrological networks, in an internal FAO-Report, Suggested Hydrological Standards & Base Mapping, (Dooley, 2003).
[30] Each OrthoTM mosaic can be comprised of up to fifteen individual LandSat scenes.

Previous Page Top of Page Next Page