Abstract | Keywords | Export | Availability | Bookmark

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The demand foragricultural land in the Congo basin is expected to yield substantialdeforestation over the coming decades. In addition to thebiogeochemical impact (release of carbon stocks), forest removalaffects the surface energy and moisture balance, and consequently theregional climate. Climate sensitivity to these biogeophysicalprocesses has been modeled during the past few decades usingcoarse-scale global climate models. Although useful as “thoughtexperiments” which yielded insight in the climatic impact ofdeforestation, such studies cannot be considered as realisticprojections for the coming decades, since they are all based on theassumption of total, basin-wide forest removal with furthermore acompletely homogeneous conversion to e.g. bare soil or pasture.Recently, the focus of numerical impact studies has been shiftedsteadily towards more detailed future projections which are oftenregionally investigated. There is, however, still room forimprovement, most notably by implementing observed successionalvegetation in the deforestation scenario. Apart from that, recentdevelopments aiming at more physical realism remain scattered amongdifferent modeling studies and research groups, and are rarelycombined.The overallobjective of this dissertation is to quantify and understand regionalclimatic impact of future deforestation in the Congo Basin, takinginto account several improvements required to achieve more realism.For this purpose, we use the COSMO-CLM regional climate model andprescribe a high-resolution spatially-explicit scenario of futureforest removal for this particular region. Model integrations areperformed with a small-scale grid resolution of 0.22° (~25km). Asopposed to assumptions such as a complete basin-wide conversion offorest to pasture or crops, the total amount of forest loss in thisstudy lies within the range of plausible estimates for deforestationin the next few decades. A state-of-the-art SVAT (soil vegetationatmosphere transfer) scheme is used, including detailed soil andvegetation input datasets and complex process parametrizations.Finally, the removed portion of primary forest is replaced by acombination of successional fallow vegetation types typical of theCongo Basin, based on field observations. It is the first time thatall these improvements are incorporated together. In addition we arethe first to introduce observed successional vegetation in our modelwhich is an important contributor to the overall improvement ofphysical realism.An extensiveevaluation of the model precedes the actual impact assessment andreveals good performance compared to in-situ and satelliteobservations. The model consists of an atmospheric part (COSMO-CLM)which is coupled to a SVAT scheme. First, two SVAT schemes are run instandalone mode (decoupled from the atmospheric model) and forcedwith meteorological in-situ measurements obtained at several tropicalAfrican sites. Model performance is quantified by comparing simulatedsensible and latent heat fluxes with eddy-covariance measurements.The simulations from Community Land Model correspond more closely tothe micrometeorological observations, reflecting the advantages ofthe higher model complexity and physical realism. The maindeficiencies identified in TERRA-ML consist of (i) a dry-seasonunderestimation of evapotranspiration, caused by erroneous defaultinput data (root depth) deviating largely from the actualregion-specific values (tropical evergreen forest), (ii)overestimations of both latent and sensible heat fluxes, caused byinaccurate leaf area index and albedo (which simply depend onhard-coded model constants), and (iii) an unrealistic fluxpartitioning caused by overestimated superficial water contents(improper parametrization of hydraulic conductivity). Community LandModel is by default more versatile in its global application ondifferent vegetation types and climates.Additionally, theSVAT schemes are tested by coupling them to COSMO-CLM. As expected,some biases of the TERRA-coupled model can be attributed toquestionable default values of input data, for instance root depth ofthe rainforest and albedo of desert sand. The implementation of amore realistic set of input parameters (EcoClimap) causes even worsesimulations in many cases, indicating a wrong model tuning. Theresults of the stand-alone validation already indicated bettersimulations of the Community Land Model compared to TERRA-ML. Thecoupled model validation now confirms the beneficial effects of usingCommunity Land Model, as it indeed also delivers the best coupledmodel results corresponding well with the observations. Hence, basedon its superior performance, the model coupled to Community LandModel is selected to perform the long-term present-day and futureclimate simulations. In one of these future simulations, the default(reference) vegetation map within the model is perturbed by adeforestation scenario. Therefore, an existing spatially-explicitdeforestation scenario is fine-tuned to match currently observeddeforestation rates, and complemented by typical successionalvegetation as observed in the Congo Basin.Successional landcover types are identified and their areal proportions are quantifiedin regions deforested during the past 37 years around the city ofKisangani, D.R.Congo. The fallow vegetation continuum is categorizedin different stages, adapted from existing classifications. Groundtruth points describing the present-day vegetation are obtainedduring a field campaign and used for supervised and validated landcover classification of these categories, using the Landsat image of2012. Areal proportions of successional land cover types are thenderived from the resulting land cover map. To illustrate the use ofthese results, the relative areal proportions are used to re-fine adeforestation scenario and apply it on existing datasets of LAI andcanopy height. Assuming a simple conversion of forest to cropland,the deforestation scenario projected a reduction ofgrid-cell-averaged canopy height from 25.5m to 7.5m (withindeforested cells), whereas the re-fined scenarios that we proposeshow more subtle changes with a reduced canopy height of 13m. Thisillustrates the importance of taking successional land covercorrectly into account in environmental and climatological modelingstudies. In our impactassessment we compared the different model simulations to each other.Differences in average climatology between present-day/future andreference/deforested simulations quantify the long-term impact offuture greenhouse gases and deforestation on the regional climate.Model integrations indicate that the deforestation, expected for themiddle of the 21th century, induces a warming of about 0.7°C. Thisis about half the greenhouse gas-induced surface warming in thisregion, given an intermediate forcing (A1B) with COSMO-CLM driven bythe ECHAM5 global climate model. This shows the necessity of takinginto account deforestation to obtain realistic future climateprojections. The deforestation-induced warming can be attributed toreduced evapotranspiration, but this effect is mitigated by increasedalbedo and increased sensible heat loss to the atmosphere.Precipitation is also affected: As a consequence of surface warmingdue to deforestation, a regional heat low develops above therainforest region. Resulting low-level convergence causes aredistribution of moisture in the boundary layer and a stabilizationof the atmospheric column, thereby reducing convection intensity andhence precipitation by 5 to 10% in the heat low region.

academic collection --- 551.583 --- 504.064.2 --- 504.732.052 --- Variations of climate. Climatic change --- Environmental conservation. Environmental analysis --- Overexploitation of terrestrial vegetation. Desertification. Desert formation --- Theses --- 504.732.052 Overexploitation of terrestrial vegetation. Desertification. Desert formation --- 504.064.2 Environmental conservation. Environmental analysis --- 551.583 Variations of climate. Climatic change