Results and discussion

1.6.1 Remotesensing applications for quantifying and mapping ecosystemservices

The review of remote sensing applications in "quantifying and mapping ecosystem services supplies anddemands"provided the followinginsights. Quantifyingecosystem services using remote sensing requires establishing a theoretical link between the remote sensing data and the ecosystem in interest. However, there is no direct link between images andanecosystem service.Hence,onlyproxies can be used to estimate ecosystem services (e.g. Carbon storage) based on indicators derived from remote sensing data. Remote sensing systems vary in theirproperties whichmake selection of the sensor system and the methodological prerequisites for deriving proxies of ecosystem services (Eigenbrod etal. 2010). The scale atwhich ecosystem services are quantified depend on the spatial, temporal, spectral and radiometric resolution of the remotelysensed data (Andrew etal. 2014). Eventhough indicators forextensive areas can be defined based on operationally available data and well-established methods, indicators useful for exact quantification of ecosystem services can be only derived experimentally at local scale.

Quantifying and mapping proxies of ecosystem services using remotesensing involves uncertainties that come from intrinsic sources of errors such atmospheric influences, geometric distortions and drifts in the calibration coefficients of the sensors. Though these sources of errors can be corrected to a certain degree, there will still be errors resulting from the statistical model used to establish link between an ecosystem services parameter(e.g.standing biomass)and the remotesensingdata resulting inuncertainty of the final results. In general, the success in quantifying and mapping ecosystem services using remotesensing depends on the sensortypes, resolution, andfinancial as well as technical capacity. Moreover, there are uncertainties involved when using remotesensing data for quantifying andmapping ecosystem services and they needto be identified and managed.

1.6.2Prosopisjuliflorainvasion andits impacts onecosystemservices

In this section,theresults of the analysis ofP.juliflora invasion ofBaadu,locatedinthe Awash River BasinofEthiopia, are highlighted.Eventhough P.juliflorawas introduced to the area as an ecosystem engineermainlyforregulating soil erosion,itinvadedlands

that are crucial insupplyingecosystem services.The speciescontinuouslyspread to new un-invaded areas and dense coppices of P. juliflora emerged after previously invaded areas were cleared. Trends in the invasion of P. juliflora in the past decade showed drastic increment in the invaded area of wetlands (flood plains)and agricultural lands (Figure 14). In the year 2000 from the 45000 ha total area of wetlands, 3600 ha was invadedwhichamounts to 8%of the wetlands area.The invadedarea increased to over 8000 ha in2005making the proportionofinvaded area ofwetlandsabout 18%.In the year 2010 more than 13600 ha area (30 % of wetlands) was invaded. Analysis of the invasion in2013showedthat20000 ha ofwetlands (40%of the total area ofwetlands) was invaded.

Figure14 P. juliflorainvasionoverthelast decade(year 2000 to 2013).

The area of drylands that is invaded by P. juliflora in the year 2000 was 60 ha which 0

5000 10000 15000 20000 25000

2000 2005 2010 2013

P.juliflora invaded area (hectares)

Year Irrigated Agriculture Drylands

Wetlands

years 2005, 2010 and 2013 respectively. The proportion of irrigated land invaded in these years ranges from 2-4 % of the total areaof irrigated landduring the same time period.

The findings demonstratedthatwetlands ofBaadu are the most affectedwith P.juliflora invasion and yet are the most useful sources of ecosystem services on which the livelihoodof the Afar pastoralists depends on.The most threatenedecosystem services include provisioning services such as food and fodder, water, and loss of native tree species that supply timber, fuelwood and charcoal. The impacts of P. juliflora on the livelihoodofpeople varyamongdifferentusergroups such as mobile pastoralists, small-scale agro-pastoralists and large-scale farmers.

1.6.3 Undercovercroplandinside forests

In the Bale Mountains of Ethiopia, cropland was found as an undercover inside the remnants of forests forming a belt in the upper escarpments though there is no undercover cropland in the upper most extremes of the site. Field observations confirmed that this belt is dominated with J. procera trees. The area of cropland per pixel of 250 m resolution gridranges from 0 to 6 hectares. Cropland was observed inside J. procera forests including very steep terrains that were entirely covered with forest and/orwith some open areas thatwere meadows previouslyusedforlivestockgrazing.

The undercover cropland forms vertical stratawith cropland as anundergrowth andJ.

procera being the upper canopy. As it was observed in the field, the major crop cultivatedinside the J.proceraforest is wheatdue to the high marketdemandforwheat andprovision of improvedseeds andfertilizer bygovernmentdue to its recent plan to improve crop production.

The relative influence ofdifferent factors on undercovercropland area calculatedfrom RapidEye images and field estimated percent cover was assessed using BRTs. The results of the BRTs model fitting demonstrated that the highly influential factors for undercover cropland area are elevation,distance to settlements,slope,East Aspectand distance to national park. Amongall the factors tested the most influential is elevation which contributed to the highest values of deviance explained by the BRTs model.

Undercover cropland area showedincrement with increasingelevation,slope, distance

to majorsettlements while itdecreases with increasingdistance from the national park.

However,aftercertainlimit the graph remains constant with a valueof 0showingthat there is noundercover cropland above such limits. Similarpatterns of the relationships between aspect and undercover cropland areawere observed.

Finally, the details of the analysis, the results, discussion and specific conclusions of studies 1,2and3 are presentedin the manuscripts listedinsection 1.7andincludedin chapters 2 to 4 respectively. The research questions and hypotheses presented in section 1.3 are addressed by the first three manuscripts listed in section 1.8 below.

1.7List of manuscripts andspecifications ofindividual contributions