int. j. remote sensing, 2002, vol. 23, no. 20, 4197– 4210
On the in uence of canopy structure on the radar backscattering of mangrove forests C. PROISY* Institut de Recherche pour le Developpement, UMR AMAP, Route de Montabo, BP 165 97323 Cayenne cedex, France
E. MOUGIN Centre d’Etudes Spatiales de la Biosphe`re, CNES/CNRS/UPS/IRD, bpi 2801 18 avenue E. Belin, 31401 Toulouse Cedex 4, France
F. FROMARD, V. TRICHON Laboratoire d’Ecologie Terrestre, CNRS/UPS, 13 avenue du Colonel Roche, 31055 Toulouse Cedex, France
and M. A. KARAM GenCorp Aerojet, Electronic Systems Division, 1100 West Hollyvale Street, Azusa, CA 91702, USA (Received 22 May 2000; in nal form 9 August 2001) Abstract. This paper is the third of a series which aims to evaluate the eVects of canopy structure on the polarimetric radar response of mangrove forests. It complements the experimental and theoretical study of closed canopies presented in the previous papers by analysing two diVerent mangrove stands of equal biomass but which greatly diVer in their structure. For the three considered frequencies (C-, L- and P-band), experimental observations show that the backscattering from the open declining stand is higher than that of the closed forest. The corresponding enhancement factor increases with wavelength and shows maximum values for the HH polarization. The identi cation of the scattering mechanisms occurring between the incident radar wave and the forest components was performed with the assistance of a polarimetric scattering model based on a radiative transfer approach. For the co-polarizations, results of the simulation study con rm that the backscatter enhancement is mainly due to an increase of either the surface scattering or the interaction component. For the crosspolarization HV at L- and P-bands, the increase of the volume component, originating from a stronger interaction with bigger branches, is found to be responsible for the observed enhancement. These ndings con rm the large eVect of the canopy structure on the forest backscatter and give rise to two important applications. First, the mapping of open declining mangrove stands appears feasible by using either the backscattering coeYcient values, especially at P-HH and P-HV, or the HH-VV phase diVerence at P-band. Second, the use of the s°–biomass statistical relationships must be restricted to homogeneous closed canopies. *e-mail: [email protected]
Internationa l Journal of Remote Sensing ISSN 0143-1161 print/ISSN 1366-590 1 online © 2002 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080/01431160110107725
C. Proisy et al.
Introduction In tropical regions, monitoring of the dynamics of mangrove forests is of particular interest owing to their considerable ecological and economic importance. In particular, mangrove forests play a major role in supplying organic nutrients to coastal marine ecosystems (e.g. Hutching and Saenger 1987). The trophic relationships between mangroves and coastal ecosystems can be characterized by the biomass and the productivity of the mangrove forests; these parameters are themselves closely linked to the structure of the forests. For the regions under consideration, biomass and productivity data are scarce, mainly because of the diYculties associated with eld measurements. In this context, data delivered by space-borne and airborne radars present numerous advantages since they can provide information on the extent of mangrove surfaces as well as on their structural parameters (Hess et al. 1990). Observations performed over the French Guianese mangroves have recently shown that there is a positive relation between the measured backscattering coeYcient s° and the total standing biomass of homogeneous canopies up to a biomass threshold for which s° saturates (Mougin et al. 1999, Proisy et al. 2000). The largest sensitivity was found for the cross polarization HV giving saturation values of about 70, 140 and 160 tons of dry material per hectare (t DM haÕ 1 ) at C-, L- and P-band, respectively. Using a statistical model derived from the P-HV response, above-ground biomass could be estimated with a mean error of 25%. In French Guiana, mangrove forests stretch over 350 km with a total extent of 600 km2. Multifrequency and multipolarization AIRSAR images acquired along the coast exhibit almost constant backscattering response due to the overall high vegetation density of mangrove forests ( gure 1). However, patches of very bright returns are observed at the lowest frequencies over open declining mangroves ( gure 1 (b) and (c)). Similar observations have been made over various ooded forests in the world (MacDonald et al. 1980, Richards et al. 1987, Wang and ImhoV 1993, Wang et al. 1995 ). Recently, we provided an interpretation of the backscatter response from those homogeneous closed mangrove forests of French Guiana (Mougin et al. 1999, Proisy et al. 2000). Our present study is to investigate how canopy structure aVects the radar backscattering of mangrove forests as well as the s°–biomass relationships. It is based on an experimental and a theoretical analysis of polarimetric data acquired over two mangrove stands of equal biomass but which greatly diVer in their structural characteristics. The second section describes the study area and the experimental data. The third section presents the modelling study. Finally, recommendations for forest biomass retrieval are indicated. 2. The study site and the eld data collection 2.1. Description The study site (52°19ê W, 4°52ê N ) is located along the coast of French Guiana. The topography of the area is nearly at. A detailed description of the area is given in Mougin et al. (1999). On this site, various development stages of mangrove forests are present: pioneer, mature and declining stages. The pioneer stage consists of a very homogeneous canopy dominated by the grey mangrove (L aguncularia racemosa). Tree density is high, ranging from about 40 000 to 10 000 stems haÕ 1 (from the sea inland ). Corresponding values of DBH (diameter at breast height) range from 0.8 to 5.3 cm and mean tree height from 0.8 to 8 m, respectively. The mature stage is dominated in biomass by the white mangrove (Avicennia germinans) with a tree
Radar backscattering of mangrove forests
Figure 1. Multipolarization AIRSAR images of the French Guianese mangrove forests acquired on 11 June 1993 (Sinnamary region). (a) C-band, (b) L- band and (c) P-band. For each image, red=HH, green=VV and blue=HV. The bright returns correspond to declining mangrove forests.
density ranging between 500 and 2000 stems haÕ 1 and DBH ranging from 8.1 to 20.1 cm. Mean tree height is about 15 m reaching a maximum of 25 m for the dominant species. The declining stage shows two strata: a high single-species stratum composed of the white mangrove and a lower stratum of the red mangrove (Rhizophora ssp.). Tree density is low, from 300 to 600 stems haÕ 1 for trees of DBH ranging from 35 to 16 cm, respectively. 2.2. Ground data In a previous study, we selected 12 stands, representing diVerent successional stages of mangrove forest dynamics (Mougin et al. 1999). On the whole, these stands consist of closed and homogeneous canopies. The eldwork was conducted aiming at characterizing the geometric properties of the stands considered. A detailed description of the ground data collection can be found in Fromard et al. (1998) and Proisy
C. Proisy et al.
et al. (2000 ). In particular, the dimensions of leaves were measured as well as the main characteristics of the woody components (trunks and branches), including size and orientation. Branches were repartitioned into 1 to 4 classes according to the considered stand. Moreover, partitioned biomass for trunks, branches and leaves were derived from allometric relationships (Mougin et al. 1999). Trunks, branches and leaves were cut and weighed on site. Their gravimetric moisture content, as well as wood density, were determined after oven drying of subsamples. In this study, an open declining stand was added to the 12 previous stands. This open stand was selected on large aerial photographs that were acquired over the same study site (Trichon et al. 1998). Stereoscopic analysis of the photographs allowed us to distinguish between the two species, Rhizophora ssp. and Avicennia germinans, and to estimate their height compared with reference data (ground measurements) . Tree density and vegetation cover fraction were also derived from aerial photographs. Table 1. Stand no.
Total above-ground dry biomass of the mangrove stands.
Total biomass (t haÕ 1 )
1 2 3 4
L aguncularia racemosa
5 6 7 8
239.5 159.2 233.3 437.4
9 10 11 12
Avicennia germinans +Rhizophora ssp.
141.8 392.8 230.1 356.8
Avicennia germinans +Rhizophora ssp.
General characteristics of the two considered stands. For stand 13, the two numbers corresponds to the two strata.
Stand no. 3 13 Table 2b.
Stand no. 3 13
5.0 31.5 71.9 92.9
Tree density (N haÕ 1 )
Basal area (m2 haÕ 1 )
Mean tree DBH (cm)
Mean tree height (m)
Canopy coverage (%)
11 944 31/250
Partitioned above-ground dry biomass of the two considered stands. Total biomass (t haÕ 1 )
Leaf biomass (t haÕ 1 )
Branch biomass (t haÕ 1 )
Trunk biomass (t haÕ 1 )
Radar backscattering of mangrove forests
Finally, biomass data were derived from the allometric relationships using height–DBH curves (Fromard et al. 1996, 1998). For diVerent stands, the estimated biomass values, expressed in t DM haÕ 1, are given in table 1. Total above-ground biomass ranges from 5 t haÕ 1 for a stand at the pioneer stage to a maximum of about 437 t haÕ 1 for the oldest mature stand. For stand 3 (pioneer) and 13 (open declining), the structural parameters and the partitioned biomass values are shown in table 2a and b. Note that stand 3 and stand 13 have the same total biomass but diVer greatly in both their structural characteristics and their degree of canopy closure. Stand 3 is a closed homogeneous pioneer forest composed of even-aged L aguncularia trees mixed with younger Avicennia individuals. This stand is characterized by a high tree density (about 12 000 stems haÕ 1 ), a limited height (