If you are interested only in canopy you can use C-band. HH polarization will give you good results.Since C-band has shorter wavelength than L-band it will not penetrate through the trees and there will be less interaction with the ground. If you are interested in trunks and other part of trees then you can use L-band. The phase difference HH and VV also gives good results.

Ref:

Classification Accuracy of Multi-Frequency and Multi-Polarization SAR Images for Various Land Covers

Turkar V, Deo R, Rao Y.S, Mohan S, Das A

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing (impact factor: 1.49). 06/2012; 5(3):936- 941.

It is crucial to understand signal scattering within forest before applying classifications.

Some baselines to understand how C-, L and P-band radar signals interact with mangrove canopies are available from

Proisy, C., Mougin, E., Fromard, F., & Karam, M.A. (2000). Interpretation of polarimetric radar signatures of mangrove forests. Remote Sensing of Environment, 71, 56-66. and

Mougin, E., Proisy, C., Marty, G., Fromard, F., Puig, H., Betoulle, J.L., & Rudant, J.P. (1999). Multifrequency and multipolarization radar backscattering from mangrove forests. IEEE Transactions on Geoscience and Remote Sensing, 37, 94-102.

I would suggest to compare radar results with those obtained using canopy grain analysis from high resolution optical images.

Proisy, C., Couteron, P., & Fromard, F. (2007). Predicting and mapping mangrove biomass from canopy grain analysis using Fourier-based textural ordination of IKONOS images. Remote Sensing of Environment, 109, 379-392.

Overall, ground truth and participation to the improvement of allometric relationships are still necessary. http://www.earthzine.org/2012/04/23/linking-remote-sensing-information-to-tropical-forest-structure-the-crucial-role-of-modelling/

To complement some point already arosen I like to exploit your question as a hint on the discussion about remote sensing.

First of all, remote sensing is a three legs approach where you have to deal with measurements (including data quality and sensor selection), modelling (that is the relationship between the observable quantity and the geophysical ones) and inversion.

The three problems are inter-related.

First of all you have to define the geophysical qunatity of your interest (say one just to simplify discussion). You say for example biomass, or vegetation cover....

It sounds good enough but according to your environmental application in mind you need to get this information at a proper scale (scale change may complete change the problem in some cases).

Then you have to consider the sensor and teh model, i.e. which is the observable to have sensitivity to your geophysical quantity of interest....

At this stage you should check if temporal and spatial scale are adequate to the geophysical dynamics.....(not an easy task in several real cases....),

then finally try to extract info by data (the inversion) with associate uncertainties, ambiguities....

Sound crazy but physics works as http://www.youtube.com/watch?v=CRChVEO8IKI

:)