Project Details
Projekt Print View

Development of an integrated forest carbon monitoring system with field sampling and remote sensing for tropical forests in Indonesia

Subject Area Forestry
Term from 2012 to 2016
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 204152256
 
Final Report Year 2016

Final Report Abstract

Forests play a relevant role in mitigating climate change and different international policy programs aim at reducing carbon emissions from forests. A major issue, however, is the scientifically well founded, transparent and verifiable monitoring of achievements in forest carbon sequestration through reduction of deforestation, forest degradation and fostering sustainable forest management. This monitoring is particularly difficult in diverse and inaccessible humid tropical forest areas which are, on the other hand, under highest pressure. Efficient as well as reliable monitoring methods are required and knowledge about the different error sources and their contribution to the total error is key to further improve the monitoring systems both in statistical and economic terms. In this project, we analyzed and quantified the contribution of all major error sources when monitoring Above Ground Biomass (AGB) in tropical forests. We started with a literature review where we identified a large set of potentially suitable single tree allometric biomass models that are used in the context of forest monitoring in the tropics. An initial analysis of the variability in mean AGB estimates showed differences of up to 86% of the mean, indicating the high relevance of selecting an appropriate allometric biomass model. We then developed a new method for selecting suitable biomass models by means of statistical tests and showed that with only 10 sample trees cut in the target forest area such models can be identified. We further evaluated different plot designs including nested circular designs, fully mapped designs and Bitterlich plots. A simulation showed that Bitterlich plot design might introduce a potentially large bias in basal area estimates, particularly for small basal area factors due to limited visibility conditions in tropical forests. Based on a Monte-Carlo simulation we identified as the most critical error sources: the choice of biomass and tree height models in combination with the measurement of heights and the identification of tree species. For the latter we tested a new method to estimate wood density using a Pilodyn device. The new device was successfully tested in the Sebangau forest, and in a joint study with other projects from eight regions located in Brazil, Ecuador, French Guiana, Indonesia, New Caledonia and South Africa, it was shown that the wood penetration depth (P, mm) measured with the Pilodyn is a good predictor of wood density. We further evaluated three options for integrating remote sensing data into the forest monitoring with the aim to gain efficiency or increase precision of estimates: i) we studied the potential of RapidEye images and LiDAR data to directly predict AGB. Even though the model qualities were lower than reported in other studies, it could be shown that a small gain in precision could be achieved when integrating remote sensing based AGB predictions using a model-assisted inference framework. ii) We analyzed the use of remote sensing data to guide the plot selection in the field. This showed a high potential to gain efficiency as indicated by the results of a simulation in which the guided selection was 2.8 as precise as the standard systematic sampling, and iii) we tested different approaches to use remote sensing data to impute the values for non-response plots. The results of this study led to the conclusion that imputing missing values improves precision over just using the mean and that the non-parametric kNN Method with 𝑘 = 1 leads to unbiased estimates of the mean and variances for increasing numbers of missing observations. From the various results of the project we conclude that there is a high potential to improve forest monitoring systems for AGB by optimizing the sampling and plot design such, that full potential of remote sensing and field inventory data can be gained. This can be achieved by the application of guided selection procedures that incorporate the variability of the remote sensing auxiliary variables, and by the combination of different plot sizes. Furthermore, the observational design can be adapted to include measurements of wood density e.g. by using the Pilodyn or similar devices.

Publications

  • (2013). Above ground biomass spatial distribution in forests: from the tree position approach to a continuous view - Implications for Forest Inventory Sampling. 1. Southern Conference on Survey Methodology (SEMS), 12.-14.12.2013, Barcelona, Spain
    Pérez-Cruzado, C., Fehrmann L., Magdon, P., Seidel D., Kleinn C.
  • (2014) Comparing the precision of biomass estimates from a sample based forest inventory and a model-assisted approach utilizing small footprint LiDAR data. ForestSAT, 4-7. November, Riva del Garda, Italy
    Magdon, P., González-Ferreiro, E., Pérez-Cruzado, C., Purnama, E.S. and Kleinn, C.
  • (2014). Uncertainties of forest spectral response observed by earth observation satellites - A case study from Kalimantan peat swamp forest. International Workshop on Forest Carbon Emissions, 3-5-March 2015, Jakarta, Indonesia
    Magdon, P., Sarodja, D., Pérez-Cruzado, C.
  • Comparing the precision of biomass estimates from sample based forest inventory and a model-assisted approach utilizing small footprint LiDAR data. A case study from tropical peat swamp forests in Central Kalimantan. In Kleinn, C., Kleinn, A. and Fehrmann, L.(Eds) 2014: The Ecological and Economic Challenges of Managing Forests Landscapes in a Global Context, Proceedings of the 4th International DAAD workshop, 16.-22.2014, Bogar & Jarkarta, Cuvillier Verlag Göttingen
    Magdon, P., Setia, E., Sarodja, D. and Pérez-Cruzado, C.
  • Is it possible to monitor forest degradation with a single inventory? A case study in peat swamp forest in Indonesia. In Kleinn, C., Kleinn, A. and Fehrmann, L.(Eds) 2014: The Ecological and Economic Challenges of Managing Forests Landscapes in a Global Context, Proceedings of the 4th International DAAD workshop, 16.-22.2014, Bogar & Jarkarta, Cuvillier Verlag Göttingen
    Pérez-Cruzado, C.
  • Simulating the effects of limited visibility on basal area estimation from Bitterlich samples in a tropical peat swamp forest. In Kleinn, C., Kleinn, A. and Fehrmann, L.(Eds) 2014: The Ecological and Economic Challenges of Managing Forests Landscapes in a Global Context, Proceedings of the 4th International DAAD workshop, 16.-22.2014, Bogar & Jarkarta, Cuvillier Verlag Göttingen
    Sarodja, D., Pérez-Cruzado, C. and Magdon, P.
  • (2015) Uncertainty budget in large-scale forest inventories: a case study in peat-swamp forests in Indonesia. International Workshop on Forest Carbon Emissions, 3-5-March 2015, Jakarta, Indonesia
    Pérez-Cruzado, C., Magdon, P., Sarodja, D., and Kleinn, C.
  • (2015). On the sitelevel suitability of biomass models. Environmental Modelling & Software, 73, 14-26
    Pérez-Cruzado, C., Fehrmann, L., Magdon, P., Cañellas, I., Sixto, H., & Kleinn, C.
    (See online at https://doi.org/10.1016/j.envsoft.2015.07.019)
  • Stand-level biomass models and their use in forest monitoring: an example for Peat Swamp Forests in Indonesia. In Fehrmann, L., Kleinn, C. and Kleinn, A.(Eds) 2015: Bridging the gap between information needs and forest inventory capacity. Proceedings of the 5th International DAAD workshop, 06.-13.2015, Durban& Pietermaritzburg, Cuvillier Verlag Göttingen
    Magdon, P., Setia, E., Pérez-Cruzado, C., Sarodja, D, Melati, D., Magdon, P., Fehrmann, L., Nengah, S.J. and Kleinn, C.
  • (2016) Integrating field sampling and RapidEye spectral response signals for forest biomass estimation. EARSeL, 3th Workshop of the Special Interest Group on Forestry, 15-16-September 2016, Krakow, Poland
    Sarodja, D., Pérez-Cruzado C. and Magdon, P.
 
 

Additional Information

Textvergrößerung und Kontrastanpassung