Environment and development
in coastal regions and in small islands

Coastal management sourcebooks 3
Part 6 Cost-effectiveness of Remote Sensing


Cost-effectiveness of Remote Sensing for Coastal Management Part A

Summary Habitat mapping is an expensive undertaking and using remote sensing to augment field survey is the most cost-effective means of achieving outputs for scientific and management purposes. Four types of cost are encountered when undertaking remote sensing: (1) set-up costs, (2) field survey costs,(3) image acquisition costs, and (4) the time spent on analysis of field data and processing imagery. The largest of these are set-up costs, such as the acquisition of hardware and software, which may comprise 40Ц72% of the total cost of the project depending on specific objectives.

For coarse-level habitat mapping with satellite imagery, the second most important cost is field survey which can account for ca 25% of total costs and over 80% of costs if a remote sensing facility already exists (i.e. in the absence of set-up costs). Field survey is a vital component of any habitat mapping programme and may constitute ca. 70% of the time spent on a project.

Detailed habitat mapping should be undertaken using digital airborne scanners or interpretation of colour aerial photography. The cost of commissioning the acquisition of such imagery can be high (£15,000Ц £26,000 even for small areas of 150 km2) and may constitute 33Ц45% of total costs (64Ц75% if set-up costs are excluded). For a moderate sized study larger than ca. 150 km2, about a month will be spent deriving habitat classes from field data and processing digital image data Ц irrespective of the digital data used (though, if staff need to be trained, this will increase time requirements considerably).

Since the set-up costs, field survey costs and analyst time do not vary appreciably between satellite sensors, the selection of cost-effective methods boils down to map accuracy and the cost of imagery, the latter of which depends on the size of the study area and choice of sensor. Using a case study of mapping coarse-level coastal habitats of the Caicos Bank, it appears that SPOT XS is the most cost-effective satellite sensor for mapping sites whose size does not exceed 60 km in any direction (i.e. falls within a single SPOT scene). For larger areas, Landsat TM is the most cost-effective and accurate sensor.

The relative cost-effectiveness of digital airborne scanners and aerial photography are more difficult to ascertain because they are case-specific. Where possible, we recommend that professional survey companies be approached for quotes. In our experience, the acquisition of digital airborne imagery such as CASI is more expensive than the acquisition of colour aerial photography. However, this must be offset against the huge investment in time required to create maps from aerial photograph interpretation (API). If habitat maps are needed urgently, say in response to a specific impact on coastal resources, API might take too long and therefore be inappropriate. For small areas of say 150 km2, a map could be created within 120 days using CASI but might take almost twice this time to create using API. We estimate that, in such a scenario, API is only cheaper if the staff costs for API are less than £75 per day. If consultants are used, this is unlikely to be the case. Further, as the area that needs to be covered by the survey increases, the cost of API is likely to rise much faster than the cost of a digital airborne scanner survey, making API progressively less cost-effective as area increases. In cases where the costs of API and digital airborne scanners are similar, the latter should be favoured because they are likely to yield more accurate results than API.

Ball-park figures of the costs involved to map an area of ca 150 km2 are given below:

Habitat detail Method Cost £
(including set-up)
Cost £
(excluding set-up)
Time taken
(person days)
Digital airborne
scanner (e.g. CASI)


The cost-effectiveness of a remote sensing survey is perhaps best assessed in relation to alternative means of achieving the same management objectives. The three primary objectives of remote sensing surveys identified by end-users were: providing a background to management planning, detecting change in habitats over time, and planning monitoring strategies (see Chapter 2). In all cases the expected outputs are habitat/resource maps of varying detail. Once the political decision has been taken that coastal resource management needs to be strengthened and that, as part of this process, an inventory of coastal ecosystems is required, the question becomes СDoes a remote sensing approach offer the most cost-effective option to achieve this objective?Т

The only alternative to remote sensing for mapping marine and shoreline habitats is use of boat-based or land-based surveys where the habitat type is recorded at each point in a grid and boundaries are fitted using interpolation methods (Figure 19.1; for more details see Box 19.1). Unlike remote sensing methods that sample the entire seascape, errors arise from sampling a grid of points because of the possibility of overlooking some habitats between adjacent survey sites. The probability of missing habitats decreases if a finer grid is surveyed, but a finer grid requires much greater survey effort. If say a 10% risk of missing boundaries is tolerated, the appropriate sampling density is approximately half the mean distance between habitat boundaries (Burgess and Webster 1984a, b).

  1. A three-person team in a boat 
    samples across a grid at a 
    constant sampling frequency.
  1. Point habitat data are entered 
    into mapping software.
  1. Interpolation of point data to 
    produce the final habitat map.

Figure 19.1 Overview of boat-based habitat mapping methods. 
A three-person team collects field data (1) across a grid and 
notes the habitat type at each point on the grid (2). The point 
samples are interpolated to create a thematic habitat map (3).  

Box 19.1

Estimating survey effort for a boat-based survey of marine habitats

It is possible to create a marine habitat map purely from field survey. The habitat type is recorded at each point within a large grid and habitat boundaries are fitted using interpolation methods. Unlike remote sensing methods, which sample the entire seascape, sampling a grid of points entails error and there is a possibility of overlooking some habitats between adjacent survey sites. The probability of missing habitats decreases if a finer grid is surveyed, but a finer grid requires much greater survey effort. In practice, a 10% risk of missing boundaries tends to be adopted (Burgess and Webster 1984a) and the appropriate sampling density for this risk of error is approximately half the mean distance between habitat boundaries (Burgess and Webster 1984b).

Ideally the mean boundary spacing would be determined in a prior reconnaissance survey, but we may obtain a surrogate measure directly from remote sensing. This entails the seemingly fair assumption that remotely sensed data can be used to identify the boundaries between habitats although this does not require that the habitats be identified correctly. Therefore, Landsat TM data were used to estimate the mean boundary spacing of habitats of the Caicos Bank based on a coarse descriptive resolution (coral, sand, seagrass, and algal habitats). To estimate the spatial frequency of habitats described in greater detail (i.e. 9 habitats), measurements were taken from CASI data of Cockburn Harbour, which encompassed a variety of lagoon and reef environments. Mean boundary spacings are listed in Table 19.1.

In the light of guidelines produced by Burgess and Webster (1984b), Table 19.1 suggests that, for a constant boundary omission risk of 10%, the interval for a synoptic boat survey mapping at a fine descriptive resolution should be 10 m and 152m for a survey at coarse descriptive resolution. It is also clear from Table 19.1 that there are considerable differences in the average sizes of different habitats. Large areas of the Caicos Bank are covered with bare sand (mean boundary spacing by class = 951 m), whereas dense seagrass, for example, occurs in comparatively small beds (mean boundary spacing by class in Cockburn Harbour = 51 m). Therefore, a survey conducted at a sampling frequency of 105 m would miss more of the dense seagrass boundaries than the sand boundaries, thus under-sampling smaller habitats and producing a map biased in favour of larger habitats.  

Table 19.1 Boundary spacing for fine and coarse habitats of
 Cockburn Harbour and the Caicos Bank respectively with 
appropriate grid spacing for habitat mapping.  

Descriptive resolution  

Fine Coarse

Mean boundary spacing (m)
Number of habitat polygons used  
in calculation
Range of mean boundary spacing  
by class (m)
Appropriate grid spacing (m)
Number of points to cover 16 km2
at fine and the whole Caicos Bank  
at coarse descriptive resolution







By calculating the spatial frequency (average patch size) of habitats on the Caicos Bank we were able to estimate the costs of using boat survey to achieve maps of similar detail to those obtained using remote sensing. The analysis is conservative with respect to boat-survey because an additional reconnaissance survey of boundary spacings would be required to plan the sampling strategy for a purely boat-based survey. Using remotely sensed data as a surrogate reconnaissance survey makes the seemingly fair assumption that remotely sensed data reveal the boundaries between habitats although this does not include the unrealistic notion that the habitats are identified correctly. At the scale of the Caicos Bank (15,000 km2), coarse descriptive resolution (corals, sea-grass, algae, sand) is more feasible than detailed habitat mapping (e.g. coral assemblages, seagrass standing crop) and data from the sensor Landsat TM were used to estimate mean boundary spacings from 2650 polygons on the image.

For a 10% risk of missing boundaries, the interval for a synoptic boat survey of the Caicos Bank mapping at a coarse descriptive resolution would be 152 m, which translates to ca 190,000 sites at an estimated cost of ca £380,000. This would take a survey team of three more than 8љ years to complete!

At the scale of marine protected areas (MPA median size 16 km2 ; Kelleher et al. 1996), where detailed habitat mapping is more feasible, CASI would be a more appropriate remote sensing method than Landsat TM (Mumby et al. 1998). Although the mean boundary spacing of habitats will vary according to the location of the MPA, a mean boundary spacing of 20 m was obtained from CASI data of a representative fringing reef and lagoon in the TCI (Cockburn Harbour). However, a corresponding grid spacing of 10 m would barely be possible given the 5 m positional errors of differential global positioning systems (August et al. 1994). A more realistic grid spacing of 25 m (half the mean boundary spacing of dense seagrass, which had the largest patches) would still require 25,600 points taking a team of three surveyors 170 days and would cost ca £21,250 in boat charges. CASI hire for such a small area would be approximately half this cost (£12,000) and field survey would be reduced to 1 day. Although image analysis may take about a month, boat-based survey would still be less accurate, more expensive, and involve an extra 16 months of person-days of effort!

In summary, while almost 70% of questionnaire respondents considered the cost of remote sensing to be a main hindrance to using it for coastal habitat mapping (Chapter 2), the issue is not that remote sensing is expensive but that habitat mapping is expensive. Remote sensing is just a tool that allows habitat mapping to be carried out at reasonable cost. Therefore, the main issue facing practitioners is: СWhich is the least expensive remote sensing method to achieve a given habitat mapping task with acceptable accuracy?Т

Any generic discussion of the costs and cost-effectiveness of remote sensing is somewhat limited by the large number of variables to take into account such as hardware availability, the size of the study site, the technical expertise of staff, the quotations from aerial survey companies for obtaining airborne data for a given region, and so on. Therefore, the chapter begins by setting out the costs incurred during remote sensing in as generic a format as possible, thus allowing readers to estimate costs to suit their own requirements.

Costs are divided into four main categories:

  1. Set-up costs (e.g. hardware and software requirements).

  2. Field survey costs.

  3. The time required for image processing and derivation of habitat classes.

  4. Cost of imagery (discussed in detail in Chapter 5 and implicit in latter sections of this chapter)  

The term СprocessingТ relates to the combined time of the operator and computer when manipulating imagery and should not be confused with the speed with which a computer will undertake computations.  

The chapter then draws on the Turks and Caicos Islands case study of coastal habitat mapping in the Caribbean to make a comparison of the cost-effectiveness of various sensors. We point out upfront, however, that this comparison is case-specific and that our conclusions must not be accepted blindly without considering the similarity between our study and that of the reader. The Turks and Caicos Islands are an ideal site for remote sensing of coral reefs and seagrass beds because the banks are relatively shallow (average depth ca 10 m) and the water clarity is high (horizontal Secchi distance 30Ц50 m). Therefore, the accuracies quoted for habitat maps are likely to represent the maximum accuracies possible for such habitats. However, the mangrove areas in the TCI are fairly small and higher accuracies may be expected in areas with a better developed mangal where patterns of species zonation would be larger and more appropriate for the larger pixels of satellite sensors.

In an attempt to consolidate the cost information, the chapter ends with a cost checklist that should help practitioners budget for a remote sensing facility or campaign.  

Costs incurred during remote sensing  

This part of the chapter discusses the costs associated with set-up, field survey and the analysis of field data, image acquisition and processing. Although each type of cost will be dealt with separately, it is worth beginning with an overview to compare the relative importance of set-up, field survey and analysis costs. All monetary costs are given in pounds sterling (£) and time is quoted in person-days (pd). Costs are based on 1999 prices and are subject to change (particularly the costs of computing equipment where a given level of performance costs less each year). Time requirements assume that staff are proficient in field survey, field data analysis and image processing. Estimates of the additional time required for training in remote sensing can be obtained from the appropriate chapters of the book.  

Overview of costs incurred during remote sensing  

The overall costs of remote sensing have been simulated for an area of ca 150 km2 based on the costs and time spent conducting remote sensing in the Turks and Caicos Islands (Table 19.2).

Table 19.2 The cost of mapping submerged marine habitats based on a case study of 150 km2 in the Turks and Caicos Islands. The relative proportions of set-up (S), field survey (F), and image acquisition (I) costs are given for two scenarios - the costs required to start from scratch (i.e. including set-up costs) and the cost given existing remote sensing facilities (excluding set-up costs). API = aerial photographic interpretation. To convert costs and time estimates to those pertinent to mapping a similar area of mangrove habitats, subtract 57 days from each time estimate and subtract £2200 from each total cost (i.e. remove boat hire costs and net difference in field requirements - Table 19.4). If required, add the cost of either purchasing or hiring a small craft (e.g. kayak) for transport with the mangal. The main assumptions of the table are described below.  
All costs in £



Airborne Digital 


Total costs incl. set-up 
Cost component
% total cost
S  F  I
72 28 1
S   F   I
68 26 6
S   F   I
69 27 4
S   F   I
68 26 5
S   F   I
40 15 45
S   F   I
48 19 33
Total  costs  incl.  set-up 
Cost component
% total cost
F    I
98   2 
F     I
82   18 
F     I
86   14 
F     I
83   17 
F     I
25   75 
F     I
36   64
Time taken (person-days) 97 98 97 95 117 229
Set-up costs assume that a full commercial image processing software licence is purchased and that the software is run on a UNIX workstation. Field survey costs assume that a differential GPS is included (see following section for further explanation), that 170 ground-truthing sites and 450 accuracy sites are visited, and that boat costs are £125 day-1. Estimates of time are based on a 3-person survey team and include the time required to derive habitat classes from field data and process imagery / photo-interpret and digitise photographs into a geographic information system. Field survey time is estimated at 69 person-days in each case with additional time being that required for image acquisition and processing (Table 19.6) or digitisation in the case of API. Total costs (including set-up) for API: some remote sensing software (e.g. Erdas Imagine 8.3 Professional) is particularly user-friendly for hand digitising polygons from aerial photographs and thus the cost of the platform and software has been included in these costs.  

Although set-up costs could be avoided (to some extent) by contracting consultants to undertake the work , this may prove to be a false economy given that much of the set-up equipment would be needed if an institution is to make effective use of the output habitat maps in a geographic information system or image processing environment. Looking at the costs including set-up, it can be seen that imagery would rep resent a small proportion of the total cost (4Ц6%) if coarse-level habitat mapping using appropriate satellite sensors (Landsat TM and SPOT XS) was the objective. This rises to about 33Ц45% of project costs if accurate fine-level habitat mapping is required, in which case only digital airborne scanners (e.g. CASI) or API of colour aerial photography are adequate. If set-up costs are not included, the total costs fall dramatically and the acquisition of SPOT or Landsat TM satellite images account for between 10Ц20% of total costs.

Field survey constitutes a significant proportion of remote sensing costs even if set-up costs are included. However, the time required to undertake processing of digital remotely sensed images is largely independent of the sensor used (whether using satellite-borne or airborne scanners). Even for a small area of 150 km2, aerial photograph interpretation (API) is at least twice as time consuming as digital remote sensing Ц and this disparity in time would grow as the study site becomes larger.  

Set-up costs  

Set-up costs are defined here as the cost of those items that have to be purchased only once to enable a mapping campaign to be instigated. Under this definition, set-up costs are independent of the type of imagery used, the duration of fieldwork and the number and distribution of sites. The most significant costs are hardware and software but training and reference materials should also be considered (see Table 19.3).

Table 19.3 Example breakdown of remote sensing set-up costs for mapping submerged tropical habitats.
Prices are approximate.  
Set-up Requirement

Cost (£)

SUN Ultra 10 with 21Ф monitor, 256MB RAM and 18GB hard disk
HP DesignJet Printer (A4ЦA1 capacity) with stand and extras
Ink and paper consumables
8 mm tape drive (7GB EXABYTE with SCSI card)
8 mm tapes to backup and store data




ERDAS Imagine Professional 8.3 licenceЖ


Training and reference materials  
Image processing textbooks and ERDAS Imagine training course
Charts (1:60,000 and 1:200,000)
Ordnance survey maps (1:25,000) (@ ca. £10 per map)
Ancillary aerial photographs to aid image interpretation (@ ca. £25 per photo)




Ж This quote is for a commercial UNIX-based licence. A commercial 
PC-based licence would cost about £7,000. Academic and non-profit 
users can usually negotiate a substantial discount from ERDAS.

A UNIX work station is recommended in preference to a high-specification PC because data processing is faster. This is a considerable bonus when conducting processing that requires many stages, or if using whole satellite images or digital airborne data that have large data volumes.  

The rapid development of affordable high-specification PCs is narrowing the advantages of UNIX based workstations: however, at the same time, UNIX workstations are steadily becoming more and more affordable.  

Field survey costs 

Field survey costs can be divided into fixed (e.g. equipment) and variable categories, the latter of which depend on the number of field sites required.  

Fixed costs of field survey

A major equipment cost for field survey is a global positioning system (GPS). Hand-held stand-alone instruments can be purchased cheaply for several hundred pounds but their spatial accuracy is poor (Chapter 4). August et al. (1994) found that 95% of position fixes from GPS were within 73 m of the true position Ц equivalent to several pixels of most satellite sensors! A differential facility (DGPS) reduced this error margin considerably to just 6 m. While DGPS is ca £1800 more expensive than GPS (Table 19.4), the extra expenditure is well worth the gains in accuracy. After all, field survey is expensive and, although the use of GPS may reduce capital expenditure, any saving is likely to be outweighed by the costs of discarding some field data due to poor positional accuracy (i.e. wasting information that was expensive to collect).  

Table 19.4 Example of fixed costs for field survey equipment


Cost (£)

General survey equipment
DGPS (preferred option)
Notebook computer for data entry and analysis
(10 GB hard-disk and 128 MB RAM)
Laminated colour prints of imagery 
(@ ca £3 per print at 1:80,000 for Landsat TM)
Recording equipment (water-resistant paper)
Alkaline batteries for DGPS and sonar
Identification guides




sub-total (including DGPS)


Additional equipment for marine surveys
Diving equipment (per set)
Hand-held depth sonar  




Additional equipment for mangrove surveys
Hemi-spherical densiometer
(to measure percent canopy closure)
Pair of PAR incident light sensors
(to measure leaf area index)
Telescopic measuring pole
(to measure mangrove height)






Variable costs of field survey  

The number of ground-truthing sites required per habitat is difficult to quantify and depends on the size of the area and distribution (i.e. complexity) of habitats: smaller study areas and areas of relatively uniform habitat (e.g. sand banks) require less ground-truthing. For example, 10 sites per habitat may be accept able for a single bay whereas 30 may be required to map an entire coastline. However, whereas ground-truthing requirements may vary, it is imperative that an adequate number of sites are visited for accuracy assessment Ц ideally at least 50 independent sites per habitat. 

The following section concerns the costs of visiting field sites, usually by boat. First, the number of sites must be estimated and expressed in terms of the number of boat days. The cost is then calculated for boat hire/fuel costs and the corresponding staff time.

Apart from stating that adequate accuracy assessment is required, it is difficult to generalise how much field survey should be undertaken. To provide some insight, how-ever, we use the Turks and Caicos Islands case-study to examine the affect of different amounts of fieldwork on the accuracy of marine habitat maps using Landsat TM. A total of 157 sites we re used to ground-truth this image. To assess the effect of varying amounts of field data on accuracy, the supervised classification procedure was repeated using a sample of 25%, 5 0 % , and 75% of these data. Each simulation was repeated three times with a random selection of field data and the accuracy of the resulting habitat maps was assessed using independent field data.

Figure 19.2 shows that the amount of field survey used to direct supervised classification profoundly influenced the accuracy of outputs although the increase in accuracy between 75% and 100% of field survey inputs was not significant. 75% of the signature file in Figure 19.2 would give about 30 sites per class at coarse-level (4 habitat classes). An additional 50 sites per habitat class should be surveyed for accuracy assessment (i.e. a total of 80 sites per habitat).  

Figure 19.2 The effect of increasing amounts of 
fieldwork on classification accuracy (expressed as
the tau coefficient) for Landsat TM supervised