The Sensitivity of Tropical Forests to Climate Variability and Change in Bolivia Christian Seiler Thesis committee Promotor Prof. Dr P. Kabat Professor of Earth System Science Wageningen University Director General and Chief Executive Officer at the International Institute for Applied Systems Analysis, Laxenburg, Austria Co-promotors Dr R.W.A. Hutjes Associate professor, Earth System Science Group Wageningen University Dr B. Kruijt Senior research scientist, ALTERRA Wageningen UR Other members Prof. Dr P.L. Ibisch, Eberswalde University, Germany Prof. Dr G.M.J. Mohren, Wageningen University Prof. Dr B. Smith, Lund University, Sweden Dr P.A. Zuidema, Wageningen University This research was conducted under the auspices of the SENSE Research School. The Sensitivity of Tropical Forests to Climate Variability and Change in Bolivia Christian Seiler Thesis submitted in fulfillment of the requirements for the degree of doctor at Wageningen University by the authority of the Rector Magnificus Prof. Dr M.J. Kropff, in the presence of the Thesis Committee appointed by the Academic Board to be defended in public on Monday 19 May 2014 at 1:30 p.m. in the Aula. Christian Seiler The Sensitivity of Tropical Forests to Climate Variability and Change in Bolivia, 157 pages. PhD thesis, Wageningen University, Wageningen, NL (2014) With references, with summaries in English, Spanish and Dutch ISBN 978-90-6173-923-0 Abstract This thesis studies the sensitivity of wet and dry tropical forests to climate variability and change in Bolivia. The complexity of the problem was approached by integrating information from climate observations, climate projections, biomass measurements, remote sensing data, and a dynamic vegetation model (LPJ-GUESS). Meteorological observations (1960 to 2009) revealed a warming trend (0.1°C per decade), as well as a drying trend since 1985 (-4% during the wet, and -10% during the dry season). Anomalies were significantly affected by the two climate modes Pacific Decadal Oscil- lation (PDO), and El Niño Southern Oscillation (ENSO). A multi-model ensemble of 35 general circulation models from the CMIP3 and CMIP5 archive projected signifi- cant increases in temperature (2.5 to 5.9°C), but conflicting changes of annual rainfall for the end of this century. Climate observations were then used to drive LPJ-GUESS in order to simulate Bolivia’s present-day potential vegetation as a baseline for assess- ing climate change impacts. Results were compared to biomass measurements from 819 plots, and to remote sensing data. Using regional parameter values for allometric relations, specific leaf area, wood density, and disturbance interval, a realistic spa- tial transition from the evergreen Amazon to the deciduous dry forest was simulated. The model reproduced reasonable values for seasonal leaf abscission, vegetation car- bon (cv), and forest fires. Modeled Gross Primary Productivity (GPP) and remotely sensed Normalized Difference Vegetation Index (NDVI) showed that dry forests were more sensitive to rainfall anomalies than wet forests. Modeled GPP was positively correlated to ENSO in the Amazon, and negatively correlated to consecutive dry days. Decreasing rainfall trends were simulated to decrease GPP in the Amazon. Using the same model configuration, the impacts of climate change were then assessed for two contrasting projections (wet and dry), revealing a large range of uncertainty. The loss of cv simulated under a dry scenario was primarily driven by a reduction in GPP, and secondarily by enhanced emissions from fires. In the wet forest, less precipitation and higher temperatures equally reduced cv, while in the dry forest, the impact of precip- itation was dominating. The temperature-related reduction of cv was mainly due to a decrease in photosynthesis, and only to lesser extent because of more autotrophic respiration and less stomatal conductance as a response to an increasing atmospheric demand. Tropical dry forests were simulated to virtually disappear, regardless of the potential fertilizing effect of rising concentrations of atmospheric carbon dioxide, sug- gesting a higher risk for forest loss along the drier southern fringe of the Amazon. The possibility of a forest dieback poses a serious threat to biodiversity, as well as rural livelihoods. Adaption measures for Bolivian forests should prioritize actions related to deforestation, fire management, grazing, and water resource management. Progress in climate change impact assessments for the Amazon will greatly depend on our ability to reduce uncertainties of future rainfall projections, possibly through physical reasoning rather than pure ensemble statistics. Otherwise, large uncertainties will re- main for climate change impact assessments in the Amazon, despite extensive model improvements. Contents 1. General Introduction 1 1.1. Context . 1 1.2. Climate variability and change . 3 1.3. Dynamic vegetation modeling . 4 1.4. Research questions . 6 1.5. Thesis outline . 7 2. Climate Variability and Trends in Bolivia 9 2.1. Introduction . 9 2.2. Study area . 11 2.3. Methods . 14 2.3.1. Data . 14 2.3.2. Data homogenization . 16 2.3.3. Climate means and extremes . 18 2.3.4. Variability . 18 2.3.5. Trends . 19 2.4. Results . 21 2.4.1. Climate means . 21 2.4.2. Climate extremes . 24 2.5. Discussion . 24 3. Likely Ranges of Climate Change in Bolivia 35 3.1. Introduction . 35 3.2. Study area . 37 3.3. Methods . 39 3.3.1. Data . 39 3.3.2. Emission scenarios . 40 3.3.3. Validation . 41 3.3.4. Ensemble weighting . 42 3.3.5. Multi-model agreement . 43 3.4. Results . 44 3.4.1. Validation and ensemble weighting . 44 3.4.2. Temperature projections . 46 3.4.3. Precipitation projections . 47 3.4.4. Shortwave radiation projections . 50 3.5. Discussion . 51 i 4. Measuring and Modeling Carbon Stocks 57 4.1. Introduction . 58 4.2. Study area . 59 4.3. Methods . 61 4.3.1. Model description . 61 4.3.2. Data . 64 4.3.3. Experimental design and analysis . 67 4.4. Results . 70 4.4.1. Plant Functional Types . 70 4.4.2. Carbon stocks . 74 4.4.3. Climate sensitivity . 76 4.5. Discussion . 80 5. The Sensitivity of Wet and Dry Tropical Forests to Climate Change 85 5.1. Introduction . 86 5.2. Study area . 89 5.3. Methods . 89 5.3.1. Model description . 89 5.3.2. Data . 92 5.3.3. Experimental design . 94 5.4. Results . 94 5.4.1. Changes in carbon pools . 94 5.4.2. Mechanisms of forest loss . 97 5.5. Discussion . 102 6. General Discussion 107 6.1. Research questions . 107 6.2. Climate change adaptation and mitigation measures . 112 6.3. Outlook . 114 A. Appendix 117 A.1. Climate Indices . 117 Bibliography 121 Summary 137 Resumen 141 Samenvatting 145 Acknowledgments 149 Curriculum Vitae 151 ii Publications 153 Education Certificate 154 Funding 156 iii 1. General Introduction 1.1. Context The Plurinational State of Bolivia is home to vast wilderness areas ranging from the Amazon rainforest to the glaciers of the Andes (Figure 1.1). Located at the southern fringe of the Amazon, Bolivia covers a transition zone of extraordinary biodiversity, with wet evergreen forests in the north, and dry deciduous forests towards the south (Ibisch and Mérida, 2003). Forests store about 6 Gt of carbon (Saatchi et al., 2011), and affect the water cycle of one of the main contributers of the Amazon river, the Madeira basin (Roche and Jauregui, 1988). Ecosystems are threatened by human pressures, including deforestation (Cuéllar et al., 2012), fires (Rodriguez, 2012a), live- stock grazing (Jarvis et al., 2010), and open-pit mining (Tejada, 2013). Conservation efforts intend to preserve natural habitats at varying scales, ranging from local land- use planning, to protected area management of a large-scale bio-corridor (Ibisch, 2005). In 1997, the Noel Kempff Mercado National Park was extended by 642,500 ha of tropical forests, terminating logging rights, and avoiding future deforestation in the respective region (Virgilio, 2009). Methods for carbon quantification were certified in 2005, creating one of the world’s first projects eligible under the REDD mechanism (reducing emissions from deforestation and forest degradation in developing countries; UNFCCC, 2010). A second sub-national REDD project was launched in 2010 in an indigenous territory of Bolivia’s Amazon (FAN, 2010). Carbon stocks protected by REDD could be threatened by a climate change induced forest die-back, as suggested by modeling experiments (Cox et al., 2004; Gumpenberger et al., 2010). The possibil- ity of such a forest loss triggered a discussion on the implications for biodiversity and long-term conservation strategies (Malhi et al., 2008). It must be noted that at the time of writing the Bolivian government opposes the carbon trading scheme of REDD projects, and is promoting an alternative mechanism including voluntary markets in- stead, namely the joint mitigation and adaptation mechanism for the comprehensive and sustainable management of forest and the mother earth (JMA; RREE, 2012). For simplicity, we continue referring to both projects as REDD, as they originally were designed as such. Model experiments exploring the impacts of climate change on vegetation dynam- ics are typically based on coupling general circulation models (GCMs) with dy- namic global vegetation models (DGVMs) under different emission scenarios. Results are subject to large uncertainties arising from conflicting climate change projections (Poulter et al., 2010), as well as different representations of the physiological responses 1 Chapter 1 General Introduction of plants to rising concentrations of atmospheric carbon dioxide ([CO2]), increasing temperatures, and water stress (Galbraith et al., 2010). The focus of previous studies has been mainly on the Amazon basin as a whole, and less on the drier transition zones from evergreen to deciduous forests along the southern margin of the basin. The impacts of climate change however may actually be higher in these transition zones, as species are closer to the extreme limits of their ecological requirements (Killeen et al., 2006).