in coastal regions and in small islands
Coastal management sourcebooks
Part 5 Quantitative Measurement of Ecological Parameters and Marine Resource Assessment
|17||Assessing Mangrove Leaf Area Index and Canopy Closure|
Summary The previous chapter introduced the different ways in which ecological parameters can be predicted from remotely sensed data, and illustrated the use of an empirical model to measure seagrass standing crop. This chapter describes the way in which a semi-empirical model can be used to measure two ecological parameters of a mangrove canopy: leaf area index (LAI) and percentage canopy closure. The accuracy and sensitivity of the mangrove models were assessed using data from airborne and satellite sensors.
The superior spatial resolution and spectral versatility of CASI allows mangrove LAI to be measured more accurately and with greater precision than with satellite sensors. However, if large areas need to be surveyed, satellite sensors are likely to be more cost-effective. The difference between LAI measured in situ and from remotely sensed data was typically less than 10%. Compared to direct methods of measuring LAI, remote sensing allows information to be obtained rapidly and minimises the logistical and practical difficulties of fieldwork in inaccessible mangrove areas.
Practical guidelines for the monitoring of mangrove canopies using remote sensing are presented.
In Chapter 13 methods for mapping mangroves using remote sensing were discussed. Airborne multispectral sensors such as Compact Airborne Spectrographic Imager (CASI) were shown to be capable of distinguishing different mangrove community types and mapping these with considerable precision (Figure 10.9, Plate 11). By contrast, satellite sensors such as Landsat TM could only separate tall Rhizophora from ‘other mangrove’, whilst SPOT XS could not distinguish mangrove from thorn scrub in the study area. However, in areas with well-developed mangroves, better results would be expected from both satellite sensors. In this chapter we describe the way in which semi-empirical models can be used to map two parameters which describe mangrove canopy structure: leaf area index (LAI, Box 17.1) and percentage canopy closure. As in the previous chapter, the method relies on gathering considerable field data both to calibrate image data – establish a significant relationship between values on an image and field measurements – and to check the accuracy of predicted values of LAI and canopy closure derived from imagery. However, the status of large areas of impenetrable mangrove can be assessed for a relatively modest investment in field survey. This chapter describes both the field survey and image processing methods required with examples based on SPOT XS and CASI imagery.
Leaf area index (LAI)
LAI is defined as the single-side leaf area per unit ground area and as such is a dimensionless number. The importance of LAI stems from the relationships which have been established between it and a range of ecological processes (rates of photosynthesis, transpiration and evapotranspiration: McNaughton and Jarvis 1983, Pierce and Running 1988; net primary production: Monteith 1972, Norman 1980, Gholz 1982, Meyers and Paw 1986, 1987; rates of energy exchange between plants and the atmosphere: Botkin 1986, Gholz et al. 1991). Measurements of LAI have been used to predict future growth and yield (Kaufmann et al. 1982) and to monitor changes in canopy structure due to pollution and climate change (Waring 1985, Gholz et al. 1991). The ability to estimate leaf area index is therefore a valuable tool in modelling the ecological processes occurring within a forest and in predicting ecosystem responses.
Mangrove canopy structure
Here we describe how semi-empirical models were derived to estimate LAI and percentage canopy closure from remotely sensed data of the Turks and Caicos Islands (refer to Chapter 16 for an explanation of empirical, semi-empirical and analytical models). This approach is particularly advantageous in mangrove areas where access to the interior is usually difficult or when alternative methods are laborious and difficult to replicate properly over whole forests.
Leaf area index
Many methods are available to measure LAI directly and are variations of either leaf sampling or litterfall collection techniques (Clough et al. 1997 and references in Chason et al. 1991). There are problems associated with both however: leaf sampling involves the destructive harvesting and measurement of leaf area for all leaves within a vertical quadrat down through the entire canopy. And litterfall collection is better suited to deciduous forests that have a single leaf fall as opposed to evergreen canopies. All direct methods are similar in that they are difficult, extremely labour intensive, require many replicates to account for spatial variability in the canopy and are therefore costly in terms of time and money. Consequently, many indirect methods of measuring LAI have been developed (see references in Nel and Wessman 1993). Techniques based on gap-fraction analysis assume that leaf area can be calculated from the canopy transmittance (the fraction of direct solar radiation which penetrates the canopy). This approach to measuring LAI uses data collected from along transects beneath the forest canopy (e.g. at 10 m intervals along 300–400 m transects in coniferous forests, Nel and Wessman 1993).
Mangroves are intertidal, often grow in dense stands and have complex aerial root systems which make the sampling regimes described so far difficult to carry out. However, Clough et al. (1997) estimated LAI from a ground area of 900 m2, their objective being a comparison between an estimation of mangrove LAI based on gap-fraction analysis with a direct collection method. To measure the LAI of mangroves over large areas would require measurements at many different locations, an extremely time-consuming process. The difficulty of moving through dense mangrove stands and the general inaccessibility of many mangrove areas would clearly pose a further problem.
The interception, scattering and emission of radiation is closely related to the canopy structure of vegetation. Spatial aspects of LAI can be indirectly measured from the spectra of mangrove forests, and so many of the problems associated with obtaining LAI values for entire mangrove forests are avoided. A semi-empirical model described in this section provides an indirect estimation of mangrove LAI with gap-fraction analysis being used as ground-truthing information to calibrate remotely sensed data. Thematic maps of LAI for the entire area covered by mangroves are derived to a high level of accuracy, without the need for large numbers of ground measurements.
Image processing methods
Landsat TM and SPOT XS imagery
Mangrove and non-mangrove vegetation in the Turks and Caicos Islands could not be separated accurately using SPOT XS imagery (Chapter 13) but, if this can achieved by other methods (Box 17.2), there is no reason why SPOT XS data cannot be used to measure LAI.
Image processing for the estimation of LAI in the Turks and Caicos Islands
Mangrove areas were identified in the Landsat TM scene using the band ratio/principal components analysis method (Method V, Chapter 13). This image was resampled to a pixel size of 20 m and recoded to produce a mangrove/non-mangrove mask. Using the near infra-red band 3 of SPOT, submerged areas were masked from the XS scene. Non-mangrove areas were then masked out using the mangrove mask derived from processing the Landsat TM data. It was necessary to use Landsat TM imagery as a mangrove mask for SPOT XS because in the Turks and Caicos Islands mangrove and non-mangrove vegetation cannot be separated accurately using SPOT XS imagery (Chapter 13). This process is illustrated in Figure 17.1.
Figure 17.1 A diagrammatic representation of the semi-empirical model used to derive LAI from remotely sensed data (illustrated for SPOT XS) and assess its accuracy. Mangroves were separated from other terrestrial vegetation on the SPOT XS image using a mask derived from Landsat TM. This happened to be necessary in the Turks and Caicos but will not be necessary everywhere. Once such a mask exists (and is known to be sufficiently accurate) then, by resampling to the appropriate pixel size, it can be applied to any other imagery of the same area and map projection.
Once non-mangrove vegetation had been masked out of the Landsat TM and SPOT XS images the remaining mangrove image data was converted into LAI using the following method. Field measurements suggest a linear relationship between normalised difference vegetation index, NDVI (see Box 13.1) and mangrove LAI (Ramsey and Jensen 1995, 1996). NDVI was therefore calculated using the red (SPOT band 2, TM band 3) and infra-red (SPOT band 3, TM band 4) bands of the XS and TM data. Values of LAI estimated from in situ measurements of canopy transmittance (see measurement of canopy transmittance and calculation of LAI) were regressed against values of NDVI derived from the image data. The equations of this linear regression model were then used to calibrate the NDVI image into a thematic map of LAI.
Water areas were removed from the image using a mask derived from band 7 (near infra-red). Since NDVI is calculated using near infra-red and red bands there were four options for calculating NDVI from the CASI data (see Table 17.1 for wavebands) with combinations of Bands 5 to 8. NDVI was calculated for all band combinations and regressed against values of LAI estimated from in situ measured canopy transmittance.
17.1 Band settings used on the Compact Airborne Spectrographic
Imager (CASI) for
|Band||Part of electromagnetic spectrum||Wavelength (nm)|
Measurement of canopy transmittance and calculation of LAI
LAI is a function of canopy transmittance, the fraction of direct solar radiation (Box 17.3) which penetrates the canopy. Canopy transmittance is given by the ratio I c /I o where I c = light flux density beneath the canopy and I o = light flux density outside the canopy. LAI can then be calculated, and corrected for the angle of the sun from the vertical, using the formula
where LAI = leaf area index, q = sun zenith angle in degrees (this can be calculated from time, date and position), k = canopy light extinction coefficient, which is a function of the angle and spatial arrangement of the leaves. The derivation of this formula is given in English et al. (1997). Nel and Wessman (1993) and Clough et al. (1997) should be consulted for a full discussion of the assumptions on which this model is based. For each field site, loge(Ic/Io) was calculated for pairs of simultaneous readings and averaged (Plate 22). A value for k of 0.525 was chosen as being appropriate to mangrove stands (B. Clough, personal communication).
Most spreadsheet packages require angles to be entered in radians, in which case should be multiplied by .
and diffuse radiation
The radiation at any point under the canopy is a mixture of direct and diffuse radiation. Direct radiation has arrived at that point directly, passing through only the canopy perpendicularly above the position, and the atmosphere. Diffuse light has arrived at that point by being reflected off clouds, ‘sideways’ through adjacent areas of canopy and back up from the soil or understory vegetation. Direct light predominates on clear days and at high sun angles (i.e. when the sun is more or less directly overhead). Diffuse light predominates on cloudy days and at low sun angles. Measurement of canopy transmittance using the field survey method described below, assumes that all the light recorded at a point under the canopy is direct radiation. It is possible to measure and correct for diffuse radiation. Two more sensors inside collimating tubes would have been required to measure diffuse radiation simultaneously with the readings of Ic and Io (see Chason et al. 1991, Nel and Wessman 1993). No attempt was made to correct for diffuse radiation however because (i) uncorrected measurements have been shown to yield estimates of LAI similar to direct methods, (ii) uncorrected measurements would under-estimate, rather than over estimate LAI, (iii)the cost of extra detectors was high, and (iv) a reduction in mobility through the mangroves would result from a system of four inter-connected detectors and cables.
Field survey methods
Measurements were taken on clear sunny days between 1000 and 1400 hours, local time. The solar zenith angle was judged to be sufficiently close to normal (i.e. more or less vertically overhead) two hours either side of noon for directly transmitted light to dominate the radiation spectrum under the canopy (Box 17.3). At other times the sun is too low and diffuse light predominates. Photosynthetically active radiation (PAR) was measured using two MACAM™ SD101Q-Cos 2PAR detectors connected to a MACAM™ Q102 radiometer. One detector was positioned vertically outside the mangrove canopy on the end of a 40 m waterproof cable and recorded Io. The other detector recorded Ic and was connected to the radiometer by a 10 m waterproof cable. If the mangrove prop roots and trunks were not too dense to prevent a person moving around underneath the canopy this detector (which was attached to a spirit level) was hand-held. If the mangroves were too dense then the Ic detector was attached to a 5.4 m extendible pole and inserted into the mangrove stand. The spirit level was attached to the end of the pole to ensure the detector was always vertical (Plate 22). All recordings of Ic were taken at waist height, approximately 0.8 m above the substrate. Eighty pairs of simultaneous readings of Ic and Io were taken at each site, the ratio Ic/Io calculated and averaged for input into Equation 17.1.
The estimation of LAI from remotely sensed data
Figure 17.2 is a scatter plot of in situ LAI values against the NDVI information derived from the SPOT XS data for 29 field sites surveyed in 1995. A linear regression was fitted to these data and a good coefficient of determination obtained (r2 = 0.74, P < 0.001, n = 29). The F-test for the model and t-test for the slope estimate were both significant at the 0.001 level of confidence, indicating a strong relationship which can be used to convert NDVI values to LAI.
Figure 17.2 The relationship between leaf area index (LAI) estimated from in situ measurements of mangrove canopy transmittance and normalised difference vegetation index (NDVI) derived from SPOT XS data for 29 sites near South Caicos, Turks and Caicos Islands. A linear regression has been fitted to the data and is significant at the 0.001 level.
The NDVI model was then used to estimate values of mangrove LAI for the entire image. LAI ranged from 0.83 to 8.51, with a mean value of 3.96. This produced a thematic image of LAI for the mangrove areas of the Caicos Bank. Field data from 32 accuracy sites surveyed in 1996 were used to test the accuracy of this image. The 95% confidence intervals of values of LAI predicted from NDVI were calculated for the regression model (Zar 1996). The accuracy of the LAI image was defined as the proportion of 1996 sites at which the value of LAI estimated from in situ measurements of canopy transmittance lay within the 95% confidence interval for that value of NDVI (Figure 17.3). Figure 17.3 shows that the thematic maps of LAI are adequately accurate; 88% of the LAIs predicted from NDVI were within the 95% confidence interval (Table 17.2). The mean difference between predicted LAI and the value estimated from field measurements was only 13% for the accuracy test sites.
Figure 17.3 A plot of the 95% confidence intervals of the LAI:NDVI regression model for SPOT XS imagery. In situ measured values of LAI for 32 accuracy sites have been superimposed over this plot. The accuracy of the LAI image was defined as the proportion of 1996 accuracy sites where the LAI value lay within the 95% confidence interval for that value of NDVI. For example, accuracy site number 16 (indicated by an arrow) has a NDVI of 0.22. At this NDVI the 95% confidence interval of a predicted value of LAI is 3.81-4.48. However the LAI at that site was calculated from in situ measurements of canopy transmittance to be 4.9. Therefore, the estimation of LAI at accuracy site number 16 was deemed inaccurate. LAIs of 28 of the 32 accuracy sites do lie between the appropriate confidence intervals. The accuracy of the SPOT XS derived thematic map of LAI is thus deemed to be 88%.
17.2 The accuracy of LAI and percentage canopy
closure prediction using three image types.
|Landsat TM||SPOT XS||CASI|
|Leaf area index
An estimate of the precision of the LAI prediction was obtained by expressing the standard error of a predicted LAI value as a percentage of LAI estimated from in situ measurements of canopy transmittance. Figure 17.4 shows that the precision is high (< 10%) for values of LAI of 1.5 and greater.
Figure 17.4 The precision of the LAI prediction. Precision is defined as the standard error/mean (see Equations 16.1-16.3). Good precision is indicated by low values of the standard error/mean, that is the error of measurement is small in relation to the mean. The standard errors of predicted values of LAI have been calculated as a percentage of mean measured LAI and are plotted against measured LAI. The precision of LAI prediction is high (< 10%) for LAI values greater than about 1.5-2.0.
The accuracy and precision of the LAI image was assessed in the same manner as for SPOT XS but was lower, 71 % (Table 17.2).
The relationship between NDVI calculated from bands 8 (near IR, 776.3–785.4 nm) and 5 (red, 630.7–643.2 nm), and values of LAI estimated from in situ measured canopy transmittance, was not significant (Table 17.3). Neither was a model using NDVI calculated from bands 8 and 6. However, there was a significant relationship when LAI was regressed against NDVI calculated either from bands 7 and 6 or 7 and 5 (Table 17.3). The former model was deemed more appropriate for the prediction of LAI because (i) it accounts for a much higher proportion of the total variation in the dependent variable, and (ii) the accuracy with which the model predicts the dependence of LAI on NDVI is higher (the standard error of the estimate is lower).
Table 17.3 A summary of the four ways used to calculate NDVI from CASI bands using combinations of bands 5 to 8. NDVI calculated from CASI data was regressed against LAI estimated from in situ measurements of canopy transmittance. Bands 5 and 6 are in the visible red portion of the electromagnetic spectrum, and bands 7 and 8 in the near infra - red (for exact wavelengths refer to Table 17.1). r2 = coefficient of determination, P = probability of the F-test for the model, NS = not significant at the 5% level, SE = standard error of the estimate. Degrees of freedom = 1, 29 in all cases. For models marked * the F-test and t-test for the slope estimate we re both significant at the 0.1% level, indicating a strongly significant relationship which could be used to convert NDVI values to LAI.
|(Band 8 - Band
5)/(Band 8 + Band 5)
(Band 8 - Band 6)/(Band 8 + Band 6)
(Band 7 - Band 6)/(Band 7 + Band 6)*
(Band 7 - Band 5)/(Band 7 + Band 5)*
The accuracy and precision of the LAI image was assessed in the same manner as for SPOT XS. Some 94% of the LAIs predicted from NDVI were within the 95% confidence interval (Table 17.2). In other words, anyone using this thematic image knows that there is a 94% probability that the 95% confidence interval of any value of LAI predicted from CASI data includes the value of LAI which would be obtained by field measurements of canopy transmittance. The mean difference between predicted LAI and the value estimated from in situ measurements of canopy transmittance was only 9% for the accuracy sites.
A thematic map of LAI predicted for mangroves on South Caicos using CASI imagery (using bands 6 and 7) is shown in Figure 17.5 (Plate 21). Comparison with Figure 10.9 (Plate 11) shows that the highest LAI occurred in areas of tall Rhizophora mangle.
The use of indices other than NDVI in semi-empirical models of LAI
The results of modelling the relationship between LAI and more atmospherically robust indices (Angular Vegetation Index – AVI and Global Environment Monitoring Index – GEMI, see Box 13.1) are given in Table 17.4 for SPOT XS and CASI. In the interests of simplicity only the results of the model using NDVI are discussed further although readers should note that these indices would be preferable to NDVI if LAI is to be monitored over time with several images (see Box 13.1 for more details on the use of different vegetation indices).
17.4 Modelling the relationship between LAI and atmospherically
indices; Angular Vegetation Index (AVI) and Global Environment Monitoring Index (GEMI),
calculated from SPOT XS and CASI data. Refer to Box 13.1 for more information on AVI
The canopy closure (expressed on a percentage scale) of a mangrove stand is highly correlated with NDVI and other indices (Jensen 1991). Canopy closure can therefore be measured from remotely sensed data in the same way as LAI (i.e. field measurements can be regressed against NDVI derived from the imagery and the resulting model used to calibrate the entire scene). Jensen (1991) proceeded to use canopy closure as a surrogate measure of tree density and determine the sensitivity of different mangrove areas to oil spill on the basis that oil would penetrate further into less dense stands.
Field techniques for measuring canopy closure
Wherever access underneath the mangrove canopy was possible percentage canopy structure was measured using a hand held semi-hemispherical mirror (a spherical densiometer) which had a grid graticule engraved on its surface. Percentage canopy closure was then estimated by computing the proportion of the mirror area covered by the reflection of leaves and stems (Plate 22). Eighty readings were taken at each site, converted into percentage canopy closure and averaged. A hemi-spherical densiometer was purchased fro m: Forest Densiometers, 5733 SE Cornell Drive, Bartlesville, OK 74006, USA (Tel: +1 918 333 2830).
Summary of results
NDVIs of approximately 0.60 or above were obtained from sites with 100% canopy closure. Below 0.60 the relationship between NDVI and percentage canopy closure was linear for all three image types. NDVI calculated from CASI bands 6 and 7 was again a superior predictor of percentage canopy closure (r 2 = 0.92, P <0.001, n = 19) to NDVI calculated from other band combinations. Regression models were fitted to scatter plots of percentage canopy closure against NDVI. The equations of these regression models were then used to predict mangrove canopy closure over the entire mangrove area.
Accuracy of this estimation of percentage canopy closure was assessed in the same manner as the LAI (see the estimation of LAI from remotely sensed data). The most accurate estimation was obtained from CASI, the least from Landsat TM (Table 17.2). The mean difference between predicted canopy closure and the in situ measured value was 4% for CASI,11% for SPOT XS and 17% for Landsat TM.
Comparison of LAI derived from remotely sensed data with other methods
Clough et al. (1997) have published LAI values for mangroves from the west coast of Peninsular Malaysia. They obtained indices ranging from 2.2 to 7.4 (mean 4.9) by direct measurement and a mean value of 5.1 when LAI was estimated indirectly from light transmission measurements over four transects. Our values of LAI derived from CASI and satellite data of Caribbean mangroves are similar to their findings and other published values for mangrove LAI (Table 17.5). The LAI of surrounding vegetation is usually quite different from mangrove. Tropical rainforest typically has a higher LAI than mangrove (~10). The difference has been attributed to the shade intolerance of mangroves, the dense aggregation of foliage in the upper portion of the mangrove canopy and the absence of an understory (Cintrón and Schaeffer-Novelli 1985). In contrast the acacia scrub in the Turks and Caicos Islands has a lower LAI than mangroves (mean 2.0), a feature of the more open canopy of that type of vegetation
17.5 Values for mangrove leaf area index estimated using airborne
Leaf Area Index
et al. (1997)
Clough et al. (1997)
Cintrón and Schaeffer-Novelli (1985)
Cintrón et al. (1980)
The lower LAI values reported from the Turks and Caicos Islands reflect the inclusion of some sites with sparse mangroves and more open canopies. Care should be taken when interpreting low (< ~1.5) values of LAI derived from remotely sensed data, especially if the understory vegetation is not uniform. The spectral signature of such sites will contain a larger proportion of light that has been reflected from the understory than if the canopy was denser. Ground cover beneath sparse mangrove sites was variable in the Turks and Caicos Islands. White sand, dark organic detritus and dense green mats of the succulent Salicornia perennis were all recorded, ground covers with presumably very different optical properties. The LAI of mangroves with relatively open canopies might be under-estimated if they were growing over white sand, or over-estimated if dark organic detritus covered the sediment. LAI values higher than previously published probably reflect the effectiveness of using remote sensing because they include especially dense mangrove areas which others might not have been able to access.
Above an LAI of 1.5–2.0 satellite data can be used to estimate mangrove LAI with considerable precision. The extent to which LAI obtained in this way can be used to model ecological processes in mangrove forests may only be limited by the availability of supporting data. If appropriate data exist then LAI may be used to produce thematic maps of, for example, rates of photosynthesis, transpiration, respiration and nutrient uptake. Such thematic maps could then be combined with other spatial data in a GIS for the management of mangrove areas.
Maps of mangrove LAI could be incorporated in a GIS and presented in 3-D in the same fashion as seagrass standing crop (see Selection of methods for mapping and monitoring seagrass beds, p. 243, and Figure 16.8, Plate 20).
The selection of an appropriate sensor for monitoring mangrove LAI
LAI can be estimated from CASI data at a significantly greater level of accuracy than is possible from either SPOT XS or Landsat TM (Table 17.2) but this must be interpreted in the context of the area covered. The area of CASI cove rage used in this Chapter was slightly more than 0.5 km2 . Although calibration of CASI imagery can be carried out at higher accuracies than satellite data, the latter cover an area that is approximately 104 or 105 times as large. There appears to be a trade-off between accuracy and cove rage. CASI is also relatively expensive in both financial cost and processing time For example, in the Turks and Caicos study CASI costs (£ sterling km-2) were approximately 400 times as much as SPOT XS, whilst acquisition and correction of imagery took about twice as long. Although CASI offers extremely high spatial resolution, great care should be taken to decide whether high resolution is really necessary. Reducing pixel size from (say) 3 m to 1 m will have a direct affect on the width of the area surveyed along each flight line (approx. 1.5 km to 0.5 km). At a resolution of 1 m the error present in each position fix, even with a DGPS, means that site-specific information has to be analysed from a 5 x 5 block of pixels. Other considerations include (i) the ratio of useful signal to noise (Box 5.1), which tends to be less for small pixel sizes, (ii) the frequent need to smooth images through filtering to facilitate visual interpretation of the final product, and (iii) reduced spectral resolution (8 rather than 18 bands for CASI). Weighed against these considerations is the ability to detect small, subtle features in CASI data.
High spectral resolution and the ability to select the location and width of the bands are considerable advantages to CASI, though ancillary data will frequently be necessary to exploit this feature fully. It is clear from the work of Ramsey and Jensen (1995,1996), who have modelled mangrove canopy reflectance in Florida at species compositions very similar to the Turks and Caicos, that there is a sharp increase in reflectance at wavelengths of 710–720 nm (the mangrove ‘red edge’, Figure 17.6). With hindsight a better configuration for CASI might have been to place two red bands and two infra-red bands either side of this red edge (i.e. at approximately 680–690 nm, 700–710 nm,720–730 nm and 740–750 nm). Bands 7 and 6 were either side of the red edge and this probably explains why NDVI calculated from them was a better predictor of LAI and canopy closure than the more spectrally distant Bands 8 and 5 (Figure 17.6).
Figure 17.6 The measured mangrove canopy reflectance between 400 and 1100 nm for mixed Florida mangroves (~40% Rhizophora, 30% Avicennia and 30% Laguncularia). The bandwidths of CASI (top graph) and SPOT XS (bottom graph) have been superimposed over the canopy spectral profile. CASI bands 6 and 7 are either side of the mangrove ‘red edge’ which occurs between about 700 and 750 nm. Canopy reflectance values based on: Ramsey, E.W., and Jensen, J.R., 1996, Remote sensing of mangrove wetlands: relating canopy spectra to site-specific data. Photogrammetric Engineering and Remote Sensing, 62 (8), 939-948.
An inevitable consequence of high spectral resolution is that image processing can be complicated and considerably extended as a result of the various combinations of bands which can be used to generate signatures, calculate indices etc. Undoubtedly, high spectral resolution can be immensely useful but readers need to be aware of these drawbacks.
Advantages of remote sensing for estimating LAI
The major advantage of remote sensing methods is that estimates of LAI for large areas of mangrove (a SPOT XS scene covers 3600 km2 ) can be obtained without the need for extensive field effort in areas where logistical and practical problems can be severe. Rhizophora mangle, in particular, can grow extremely densely and access to the interior of mangrove thickets is physically impossible in many cases. Remote sensing is therefore the only way that LAI can be estimated non-destructively. In addition, the LAI data is obtained relatively quickly; the field measurements presented here required a team of two working for a total of 34 person-days; processing time involved another 22 person-days.
Araújo, R.J, Jaramillo, J.C., and Snedaker, S.C., 1997, Leaf area index and leaf size differences in two red mangrove forest types in South Florida. Bulletin of Marine Science, 60 (3),643–647.
Botkin, D.B., 1986, Remote sensing of the biosphere. National Academy of Sciences, Report of the Committee on Planetary Biology, National Research Council, Washington D.C.,USA.
Chason, J.W., Baldocchi, D.D., and Huston, M.A., 1991, A comparison of direct and indirect methods for estimating forest canopy leaf area. Agricultural and Forestry Meteorology, 57, 107–128.
Cintrón, G., and Schaeffer-Novelli, Y., 1985, Características y desarrollo estructural de los manglares de Norte y Sur America. Ciencia Interamericana, 25, 4–15.
Cintrón, G. , Lugo, A.E., and Martínez, R., 1980, Structural and functional properties of mangrove forests. A symposium signalling the Completion of the ‘Flora of Panama’. Universidad de Panamá Monographs in Systematic Botany, Missouri Botanical Garden.
Clough, B.F., Ong, J.E., and Gong, G.W., 1997, Estimating leaf area index and photosynthetic production in canopies of the mangrove Rhizophora apiculata. Marine Ecology Progress Series, 159, 285–292.
English, S., Wilkinson, C., and Baker, V., 1997, Survey Manual for Tropical Marine Resources. 2nd Edition. (Townsville: Australian Institute of Marine Science).
Gholz, H.L., 1982, Environmental limits on above-ground net primary production, leaf area and biomass in vegetation zones of the Pacific Northwest. Ecology, 63, 469–481.
Gholz, H.L., Vogel, S.A., Cropper, W.P., McKelvey, K., Ewel,K.C., Teskey, R.O., and Curran, P.J., 1991, Dynamics of canopy structure and light interception in Pinus elliottiistands, North Florida. Ecological Monographs, 6, 33–51.
Green, E.P., Mumby, P.J., Edwards, A.J., and Clark, C.D., 1996, A review of remote sensing for the assessment and management of tropical coastal resources. Coastal Management, 24, 1–40
Jensen, J.R., Ramset, E., Davis, B.A., and Thoemke, C.W., 1991, The measurement of mangrove characteristics in south-west Florida using SPOT multispectral data. Geocarto International, 2, 13–21.
Kaufmann, M.R., Edminster, C.B., and Troendle, C., 1982, Leaf area determinations for subalpine tree species in the central Rocky Mountains. US Department of Agriculture and Rocky Mountains Forestry Rangers Experimental Station General Technical Report, RM–238.
Meyers, T.P., and Paw, U.K.T., 1987, Modelling the plant canopy micrometeorology with higher-order closure principles. Agricultural and Forestry Meteorology, 41, 143–163.
Meyers, T.P., and Paw, U.K.T., 1986, Testing of a higher-order closure model for modeling airflow within and above plant canopies. Boundary-Layer Meteorology, 37, 297–311.
McNaughton, K.G., and Jarvis, P.G., 1983, Predicting effects of vegetation changes on transpiration and evaporation. In Water Deficits and Plant Growth. Vol. 7., edited by T.T. Kozlowski, (London:Academic Press),pp. 1–47.
Monteith, J.L., 1972, Solar radiation and productivity in tropical ecosystems. Journal of Applied Ecology, 9, 747–766.
Nel, E.M., and Wessman,C.A., 1993, Canopy transmittance models for estimating forest leaf area index. Canadian Journal of Forestry Research, 23, 2579–2586.
Norman, J.M., 1980, Photosynthesis in Sitka Spruce (Picea sitchensis). Radiation penetration theory and a test case. Journal of Applied Ecology, 12, 839–878.
Pierce, L.L., and Running, S.W., 1988, Rapid estimation of coniferous forest leaf area index using a portable integrating radiometer. Ecology, 69, 1762–1767.
Ramsey, E.W., and Jensen, J.R., 1995, Modelling mangrove canopy reflectance by using a light interception model and an optimisation technique. In Wetland and Environmental Applications of GIS, (Chelsea,Michigan:Lewis Publishing),pp. 61–81.
Ramsey, E.W., and Jensen, J.R., 1996, Remote sensing of mangrove wetlands: relating canopy spectra to site-specific data. Photogrammetric Engineering and Remote Sensing, 62 (8), 939–948.
Waring, R.H., 1985, Estimates of forest growth and efficiency in relation to canopy leaf area. Advances in Ecological Research, 13, 327–354.
Zar, J.H., 1996, Biostatistical Analysis. Third Edition. (New Jersey, USA:Prentice Hall).