SPIE Membership Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:
SPIE Photonics West 2019 | Call for Papers

2018 SPIE Optics + Photonics | Register Today



Print PageEmail PageView PDF

Remote Sensing

Tropical forest biomass assessment using multi-frequency radar imagery

Combining X- and L-band data allows accurate estimation of aboveground biomass and provides information on spatial variation across different forest ecosystems and forest degradation.
29 August 2011, SPIE Newsroom. DOI: 10.1117/2.1201108.003684

In the context of the international effort to reduce anthropogenic greenhouse gas emissions, several climate change mitigation mechanisms have been developed. Reducing emissions from deforestation and forest degradation (REDD), which aims to prevent deforestation, is one example. It was accepted at the climate change conference in Cancún in 2010. A precise quantification of aboveground biomass (AGB) is highly relevant to this endeavor, especially in tropical forests, which store huge amounts of carbon.

Because tropical forests are often difficult to access on the ground, satellite observations and measurements could become the primary source for monitoring AGB in tropical areas.1 Spaceborne synthetic aperture radar (SAR) sensors are active systems that transmit microwave energy at wavelengths ranging from 3.1cm (the X band) to 23.6cm (the L band). They are weather and daylight independent. This is very advantageous in tropical regions, which are often covered by clouds.

In this context, we investigated the potential of the combined use of TerraSAR-X (X-band) and ALOS PALSAR (L-band) data to estimate AGB in intact and degraded tropical forests in an especially carbon-rich forest ecosystem in Central Kalimantan, Borneo, Indonesia.2 Longer SAR wavelengths have proven to be more useful for AGB assessment because the backscatter range increases with changes in the biomass.1Biomass analyses using radar imagery are normally limited by saturation of the radar signal at high biomass values, which occurs when the biomass/backscatter slope approaches zero. Shorter wavelengths (e.g., the X band) saturate rather quickly—at approximately 50 tons per hectare (t/ha)—whereas longer wavelengths (e.g., the L band) saturate at high biomass levels (150t/ha, possibly up to 600t/ha).3, 4 However, the saturation level is influenced not only by the SAR frequency but also by the number and quality of forest inventory data points.

We collected and compiled AGB data based on forest inventories and airborne light detection and ranging (lidar) measurements, which is currently the most accurate method of estimating biomass.5 The database provided a huge number (n=3970) of AGB estimates over the entire biomass range from woody regrowth to mature pristine forest. We used these estimates for biomass regression modeling based on multi-frequency SAR data.

The relationship between radar signals and AGB was analyzed using spatially averaged backscattering coefficients over a grid size of 1ha, which offered the best tradeoff between radiometric accuracy and loss of spatial accuracy. To minimize the potential impact of rainfall, surface water, and soil moisture, only radar imagery acquired during the dry season (May to October) was processed. The regression models were then applied to the entire study area in order to assess the spatial complexity of biomass variation in different forest ecosystems. The results were compared with data from additional sources, such as Landsat, RapidEye, and field observations.

The regression modeling of AGB based on TerraSAR-X and ALOS PALSAR imagery showed that multi-temporal regression models are more accurate than mono-temporal models. Mono-temporal regression models are affected by weather conditions before and during image acquisition, and thus their spatial and temporal transferability are limited. Multi-temporal models compensate for varying weather conditions and are more feasible for temporal and spatial transfer.

Figure 1 .Aboveground biomass estimations of multi-temporal TerraSAR-X, ALOS PALSAR, and combined regression models of deforested areas with low biomass (upper panels) and peat swamp forest areas with high biomass (lower panels). Left panels depict a Landsat scene from 19 May 2008. The photographs show examples of different land covers indicated by arrows in the biomass map. t/ha: Tons per hectare. This figure was published previously.2(Photos © F. Siegert.)

The investigation of single-frequency relationships showed that X-band data is better for estimating low biomass values, and L-band data is better for estimating high biomass values. The combined use of TerraSAR-X and ALOS PALSAR imagery achieved the most accurate estimation over the entire biomass range (see Figure 1). The saturation of the multi-temporal TerraSAR-X and ALOS PALSAR combined model was 300t/ha, but the spatial pattern of AGB values and a comparison with multi-spectral imagery suggest that it might be possible to assess biomass levels up to 600t/ha. We are currently investigating this. Finally, we showed the temporal and spatial transferability of the multi-temporal X- and L-band combined biomass model. Appropriate AGB estimations were achieved for a similar forest ecosystem with different degradation levels in another study area in West Kalimantan when the same preprocessing and incidence angles, and only dry season SAR images were used. These constraints are very important for defining the spatial transferability of biomass regression models.6

In summary, the approach presented here allows us to estimate AGB values of to up to 300t/ha in tropical forests, and might enable us to visualize the spatial variability of AGB values of up to 600t/ha. Thus, it has the potential to meet the requirements for successful REDD in terms of accuracy and the trade-off between feasibility and cost.

Additional in situ biomass data would improve model calibration and validation. We are currently extending and updating forest inventories and lidar measurements. This process is expected to yield more accurate AGB results, which will allow more precise modeling. In particular, we will analyze the spatial transferability in more detail, especially that between different ecosystems.

Sandra Englhart
Department of Biology II and GeoBio Center
Ludwig-Maximilians-University Munich
Munich, Germany
Vanessa Keuck
Remote Sensing Solutions GmbH
Baierbrunn, Germany
Florian Siegert
Department of Biology II and GeoBio Center
Ludwig-Maximilians-University Munich
Munich, Germany
Remote Sensing Solutions GmbH
Baierbrunn, Germany

1. D. S. Lu, The potential and challenge of remote sensing-based biomass estimation, Int'l J. Rem. Sens. 27, no. 7, pp. 1297-1328, 2006. doi:10.1080/01431160500486732
2. S. Englhart, V. Keuck, F. Siegert, Aboveground biomass retrieval in tropical forests—the potential of combined X- and L-band SAR data use, Rem. Sens. Environ. 115, pp. 1260-1271, 2011. doi:10.1016/j.rse.2011.01.008
3. E. T. A. Mitchard, S. S. Saatchi, I. H. Woodhouse, G. Nangendo, N. S. Ribeiro, M. Williams, C. M. Ryan, S. L. Lewis, T. R. Feldpausch, P. Meir, Using satellite radar backscatter to predict above-ground woody biomass: a consistent relationship across four different African landscapes, Geophys. Res. Lett. 36, 2009. doi:10.1029/2009GL040692
4. J. M. Austin, B. G. Mackey, K. P. Van Niel, Estimating forest biomass using satellite radar: an exploratory study in a temperate Australian Eucalyptus forest, Forest Ecol. Manag. 176, no. 1–3, pp. 575-583, 2003. doi:10.1016/S0378-1127(02)00314-6
5. K. Kronseder, U. Ballhorn, V. Böhm, F. Siegert, Aboveground biomass estimation across forest types at different degradation levels in Central Kalimantan using LiDAR and field inventory data, J. Appl. Rem. Sens., 2011. submitted.
6. S. Freitas, M. Mello, C. Cruz, Relationships between forest structure and vegetation indices in Atlantic Rainforest, Forest Ecol. Manag. 218, pp. 353-362, 2005. doi:10.1016/j.foreco.2005.08.036