This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 182.73.193.34

This content was downloaded on 22/06/2016 at 11:55

Please note that terms and conditions apply.

Rising floodwaters: mapping impacts and perceptions of flooding in Indonesian Borneo

View the table of contents for this issue, or go to the journal homepage for more

2016 Environ. Res. Lett. 11 064016

(http://iopscience.iop.org/1748-9326/11/6/064016)

Home Search Collections Journals About Contact us My IOPscience

Environ. Res. Lett. 11 (2016) 064016 doi:10.1088/1748-9326/11/6/064016

LETTER

Rising floodwaters: mapping impacts and perceptions of flooding inIndonesian Borneo

Jessie AWells1,2, Kerrie AWilson3, Nicola KAbram1,2,4,5,MalcolmNunn1, David LAGaveau6,RebeccaKRunting1,7, Nina Tarniati2, Kerrie LMengersen8 andErikMeijaard1,2

1 ARCCentre of Excellence for Environmental Decisions, University ofQueensland, Brisbane,QLD4072, Australia2 Borneo Futures, People andNatureConsulting International, CountryWoods 306, Jl.WR Supratman, Ciputat, Jakarta, 15412, Indonesia3 School of Biological Sciences, TheUniversity ofQueensland, St Lucia, Brisbane,QLD, 4072, Australia4 Living Landscape Alliance, 5 JupiterHouse, Calleva Park, Aldermaston, Reading, RG7 8NN,UK5 HUTAN/KinabatanganOrang-utanConservation Programme, POBox 17793, KotaKinabalu, Sabah 88874,Malaysia6 Center for International Forestry Research, POBox 0113 BOCBD, Bogor 16000, Indonesia7 School of Geography, Planning and EnvironmentalManagement, University ofQueensland, Brisbane,QLD4072, Australia8 School ofMathematical Sciences, QueenslandUniversity of Technology, Brisbane, Australia

E-mail: emeijaard@gmail.com

Keywords:Borneo, flooding, land use change, watershed ecosystem services

Supplementarymaterial for this article is available online

AbstractThe roles of forest andwetland ecosystems in regulating flooding have drawn increasing attention inthe contexts of climate change adaptation and disaster risk reduction.However, data onfloods arescarce inmany of the countries where people aremost exposed and vulnerable to their impacts. Here,our separate analyses of village interview surveys (364 villages) and news archives (16 sources) showthatfloods havemajor impacts on lives and livelihoods in Indonesian Borneo, andflooding risks areassociatedwith features of the local climate and landscape, particularly land uses that have seen rapidexpansions over the past 30 years. In contrast with government assessments, we find thatflooding isfarmorewidespread, and that frequent, local, events can have large cumulative impacts. Over threeyears, local news agencies reported floods that affected 868 settlements, 966 times (including 89 inurban areas), inundated at least 197 000 houses, and displacedmore than 776 000 people, possibly asmany as 1.5million (i.e. 5%–10%of the total population). Spatial analyses based on surveys in 364villages show thatflood frequency is associatedwith land use in catchment areas, including forestcover and condition, and the area of wetlands,mines (open-cut coal or goldmines), and oil palm. Theprobability thatfloods have becomemore frequent over the past 30 years was higher for villages closertomines, and inwatersheds withmore extensive oil palm, but lower inwatershedswith greater coverof selectively-logged or intact forests.We demonstrate that in data-poor regions,multiple sources ofinformation can be integrated to gain insights into the hydrological services provided by forest andwetland ecosystems, andmotivatemore comprehensive assessment offlooding risks and options forecosystem-based adaptation.

1. Introduction

Global initiatives such as The Economics of Ecosys-tems and Biodiversity, the Intergovernmental Plat-form on Biodiversity and Ecosystem Services and theIntergovernmental Panel on Climate Change haveemphasised the importance of ecosystem services forhuman well-being and climate adaptation in tropical

developing countries. It is in these countries, however,that changes in ecosystem processes and the relation-ship between changes in land use and land cover(LULC) and the provision of ecosystem services areleast understood.

Rapid changes in LULC are causing widespreaddegradation and loss of tropical ecosystems, impactingmany of the services these ecosystems provide [1, 2].

OPEN ACCESS

RECEIVED

18November 2015

REVISED

30April 2016

ACCEPTED FOR PUBLICATION

1 June 2016

PUBLISHED

20 June 2016

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2016 IOPPublishing Ltd

Impacts on forest and wetland ecosystems are particu-larly important for services linked to the hydrologicalcycle. These services depend on the landscape’s capa-city to regulate the quantity of freshwater flows, miti-gate damages from flooding and erosion, andinfluence transport of pollutants, sediments or nutri-ents [3, 4].

Linkages between LULC change (LULCC) andchanges in flooding regimes form a complex andactive field of research [5]. Recent reviews haveshown that flood frequency and severity are not sim-ply determined by changes from forest to non-forestland cover. Instead, flooding is influenced by theextents of change in vegetation structure and soilqualities, and the scales of the surrounding watershedand rainfall events [6–9]. The most extreme rainfallevents can lead to major floods, regardless of landcover, if extensive or prolonged rainfall overwhelms alandscape’s capacity to store or slow themovement ofwater, and the capacity of river channels to conductfloodwaters away [10]. Nonetheless, when consider-ing rainfall events that are less extreme (and so morecommon), ecosystems can strongly influence floodpeaks and velocities [11–14]. Conversion of naturalecosystems to more intensive land uses alters water-shed hydrology in three main ways: (1) changes invegetation and soils alter water cycling, storage andmovement by reducing canopy interception of rain-water, soil saturated water content, infiltration ratesand hydraulic conductivity, and rates of evapo-transpiration [6, 15]; (2) impervious surfaces and lin-ear networks such as roads or footpaths alter thepaths and speed of water movement, especially if theyact as direct conduits for run-off to streams [16]; (3)erosion, landslips, sedimentation and scouring asso-ciated with land cover change alter river channelmorphology [17].

Both major floods and smaller, more frequentfloods can have large impacts on human lives, healthand livelihoods [18]. Impacts include direct loss of life,injury, disease and displacement; damage to agri-culture, industries and the built environment; and dis-ruption of essential services such as water, healthcareand education [18]. Changes in LULC can also influ-ence human vulnerability to the impacts of flooding,through changes in human settlement patterns andsocietal capacity to mitigate or adapt to flood-ing [19, 20].

For tropical climates and developing countries,most of the existing studies relating flood occurrenceor impacts to LULCC and/or socio-economic factors[e.g. 21, 22] have been based on global datasets[23, 24]. However, these datasets are highly incom-plete for many countries, even for the high-impactevents that are their focus [9, 25]. Given this scarcity ofdata, local knowledge can form a vital source ofinsights into current impacts and drivers of change inflooding events [26].

In this study, we employ novel data sources fromvillage interviews and news archives to assess the fre-quency and impacts of flooding on local communitiesin a tropical biodiversity hotspot, Indonesian Borneo[27]. Studies of flooding or relations to LULCC havebeen rare, and to our knowledge, this is the first tocover all of Indonesian Borneo. Past studies consist ofhistorical descriptions of floods in the 17th–19th Cen-turies [28]; and hydrological simulations in specificriver basins, for example showing increasing down-stream flooding due to degradation and subsidence ofpeatlands in Central Kalimantan [29]; and higher wetseason discharge (and lower dry season discharge) dueto forest and peatland degradation in Kapuas, WestKalimantan [11].

Our objectives were to: (1)map patterns of flood-ing across Indonesian Borneo, through statisticalmodelling of recent flood frequencies and perceivedtrends in frequency over the past 30 years; (2) comparethe maps of flood hazards generated from two inde-pendent sources of data on recent flooding (interviewsand news archives); and (3) explore the relationshipsbetween flooding and the social and environmentalfeatures of each watershed, including LULC. Wedemonstrate the large scope for information frominterview and news sources to rapidly improve theaccuracy and resolution of flooding assessments. Thisapproach has great utility in data-poor environmentswhere floods have substantial impacts on human live-lihoods andwellbeing.

2.Methods

2.1. Study regionBorneo is the world’s third largest island, located inSouth-East Asia, encompassing regions of three coun-tries: Indonesia, Malaysia and Brunei Darussalam(4.242° S–6.965°N, 108.883° E–118.990° E). Borneo’snative ecosystems host some of the highest levels ofbiodiversity globally and include various (and oftenthreatened) vegetation types such as mangroves, low-land dipterocarp tropical forests, peat swamp forestsand montane rainforests [30]. These forested systemsnot only harbour some of the world’s most iconicspecies such as the Bornean orangutan (Pongo pyg-maeus) [31] and proboscis monkey (Nasalis larvatus)[32], but also provide essential and varied ecosystemservices to some 25 million people across Borneo[26, 33]. Current rates of deforestation and forestdegradation, however, are among the highest in theworld [34–36], thereby threatening such services.Mean annual rainfall is generally high yet variable[range: 1520–4820 mm; 37]. Heavy rainfalls can leadto severe flooding, especially in lowland and coastalareas [38], which are often targeted for conversion tooil-palm agriculture [36, 39]. Our study focused onriverine and flash flooding in Kalimantan (i.e.

2

Environ. Res. Lett. 11 (2016) 064016

Indonesian Borneo), which covers 76% of the island’stotal area.

2.2. Rivers andwatershed dataWedelineated rivers and watersheds in Borneo using ahydrologically corrected Digital Elevation Model at 3arc-second resolution, and ArcHydro 2.0 tools inArcGIS 10.2 [40] (see S1—supplementary methods,S1.1). We defined ‘major rivers’ by a minimumdrainage area of 200 km2, and the finer network of ‘allrivers’ by a minimum drainage area of 20 km2. Wedefined watersheds as areas draining to the ocean, anddelineated subwatersheds as the area draining to eachlink of the major rivers. For any point of interest, the‘relative watershed’ consists of the surrounding sub-watershed, combined with any upstream subwater-sheds (i.e., all areas that contribute to flow through thefocal subwatershed).

2.3. Interview surveys—villagers’perceptions offloodingWe surveyed village leaders’ perceptions of flooding in364 villages across Kalimantan (figure 1, and figureS1), through interviews with the village head or otherofficial. These interviews were conducted duringApril–October 2009 (341 villages) and April–October2012 (23 villages), andwere drawn from a larger surveyon villagers’ perceptions of forests and wildlife, invillages selected at random within 10 km of forests in

the approximate range of the Bornean orangutan (i.e.sampling was random with respect to past or presentflooding). The larger survey comprised two sets ofinterviews: (1) the village-level interviews, which weanalyse; and (2) interviews on villagers’ individualperceptions with 7–12 respondents per village, whichwe do not analyse, and which formed the basis forprevious studies of wildlife and ecosystem ser-vices [26, 33].

Our study selected the 364 villages within Kali-mantan, for which village-level interviews gave detailedresponses to questions on flooding. A ‘flood’ wasdefined as either a riverine flood orflash flood, inwhichfloodwaters covered the village’s main road or path atthe centre of the village. We coded responses regarding(1) the frequency of flooding over the past five years(n = 302 villages; coded as 5 classes representing fre-quencies of 0, 0.1, 0.5, 1 and 2+ floods per year) and (2)trends in flood frequency over the past 30 years(n = 260; responses to the question ‘has the frequencyof floods declined, stayed the same, or increased overthe past 30 years?’were coded as: decline, no change, orincrease in flood frequency). For the flood trends, a‘decline’was reported in only 2 of 260 villages (S2.1), sowe excluded this class, and modelled flood trends aspresence/absence data, i.e. whether flood frequencyincreased (presence) or showed no change (absence).Further details of the interviewmethods, quality assess-ments, and coding are given in S1.2.

Figure 1.Map of interview surveys in 364 villages in Indonesian Borneo (conducted 2009–2012), land use and land cover in 2010(SarVision andBorneo Futures), and Province borders.

3

Environ. Res. Lett. 11 (2016) 064016

2.4. Newspaper reports offlood events and damagesWeobtained flooding reports from the online archivesof six news agencies in Kalimantan, covering 16regional newspapers, using the search keyword ‘banjir’(flood). We collated data from all articles that reportedflooding in Kalimantan over 3 years, 20 April 2010–29April 2013, and that gave estimates of flood heightand/or flood impacts (see S1.3). We georeferencedflooding locations using Google Earth and censusadministrative boundaries. We estimated floodimpacts for each flooded settlement as the number ofpeople affected, and number of houses flooded. If anumeric value was not reported directly for a settle-ment, we assigned low, median and high estimates(based on the median, 10th and 80th percentiles fromthe statistical distributions of the number of housesand number of people affected per flooded settle-ment), enabling calculation of summed impacts acrossall the separateflood events (table S1).

2.5. Relationshipswith LULCand topographicvariablesWe performed spatial analyses separately for each ofthe three datasets: (1) village flood frequency fromvillage interviews over the past five years (n = 302villages); (2) village flooding trends over the past 30years (n = 260 villages); and (3) news reports of floodevents (n = 380 settlements, i.e. villages, towns orurban areas) (figure S1). We analysed each dataset inrelation to 35 spatial predictor variables (‘landscapevariables’—see S1.6) to assess statistical relationshipsbetween flooding and LULC in each watershed (20LULC classes), soil types and water storage capacities,topography (slope, elevation), climate (temperatureand rainfall seasonality and monthly maximum),infrastructure (impervious surface area, and roaddensities), and socio-economic factors (populationdensity in 2011, and proportions of the population byreligion and ethnic group). Initial models included allspatial predictors; final models included those con-tributing>1%of explained variance.

We performed statistical modelling using BoostedRegression Tree (BRT) models with five-fold cross-validation [41] in the R statistical environment[42, 43]. BRTmethods combinemany regression treesto form an ensemble model, providing a flexible androbust method for both (1) quantifying relationshipsand contributions (including nonlinear relationshipswith the response, and interactions among pre-dictors), and (2) generating predictions for new areasor datasets [41]. Further details of model fitting andassumptions are given in S1.4. We used the resultingmodels to generate predictions for all populated areasacross Kalimantan (S1.4). Model performance wasassessed via deviance reductions, cross-fold correla-tions and estimates of classification error (detailed inS1.4, S2.2).

3. Results

3.1. Villagers’perceptions offloodingIn 58% of villages (176/302), respondents reportedflooding frequencies of one or more times every yearover the past 5 years (S2—supplementary results, tableS2). Only 10% of villages reported no experience offlooding. Approximately 20% of all villages reportedthat floods had becomemore frequent over the past 30years (table S2; as mentioned in methods 2.3, only twovillages or 0.8% reported declines in frequency, andwere excluded from modelling of flood trends). Theassociation between flood trends over 30 years, andcurrent frequency over 5 years, was positive but nottight (figure S2); i.e. villages that perceived an increasein flood frequency, reported current frequencies thatstill span a wide range, from occasional to more thanonce per year. Current flood frequencies were gen-erally consistent for villages on the same stretch of ariver, and either similar or lower for villages furtherupstream. For example, when looking at the villageswith annual or more frequent floods, other villagesfurther upstream (>3.2 km) usually reported similaror lower flooding frequencies (mean differencebetween frequencies−0.24 floods per year, range 1 to−2; figure S3).

3.2. Newspaper reports offlood events and damagesOver the period 20 April 2010–29 April 2013 (3 years),414 news items reported at least 142 distinct floodevents affecting 966 settlements in Kalimantan. Quan-tifying the total impacts over the 142 flood eventsbetween 2010 and 2013 (table S7), we estimate that197 000 to 360 000 houses were flooded (based on themedian and 80th percentile of the distribution ofhouses flooded per event, respectively, table S1),directly displacing 776 000 people (possibly as manyas 1.5million, based on the 80th percentilemethod).

Flood events on average affected five settlements,and 60 settlements were flooded in two or more of theevents. Flooding was reported in 32major urban areasand 836 other settlements. Observed locations offlooding based on newspaper reports (2010–2013) andvillage interviews (2008–2011) showed strong spatialagreement. Almost all interviews from villages close tosites of newspaper-reported floods (within 4 km over-land or 60 km downstream of villages) reportedexperiences of moderate to high flood frequencies(every two years ormore frequent).

The most frequently quantified impact of floodingreported was the number of houses flooded, and thenumber of people directly affected. Other impactswere frequently described, but not quantified, forexample ‘residents could not work’, ‘schools wereclosed’, or croplands, plantations, schools, businessesor health facilities were flooded. Financial impactswere often mentioned, but rarely enumerated. Gov-ernment budget allocations, however, are clearly high:

4

Environ. Res. Lett. 11 (2016) 064016

for example the Government of East Kalimantan allo-cated 605 billion Rupiah (c. US$50 million) for floodmitigation in one city alone from 2011–2013, pri-marily for four systems of structural defences along theMahakamRiver [44].

3.3. Relationships betweenflooding and landscapevariablesThe village interviews and news-reported floodsprovide two distinct lines of evidence for the existenceof frequent and widespread flooding, and for examin-ing the relationships between flooding and landscapevariables. BRT models showed very high performancefor all datasets (table S3), and key predictors are shownfor each analysis in figure 2. The influence of each classof landscape variable (LULC, climate, topography etc)is summarised in table 1. The influence of eachlandscape variable is computed from substantialcross-validation, giving a robust estimate of thevariable’s contribution to themodel’s predictive good-ness-of-fit. Interactions among predictors are natu-rally incorporated in the tree structure of themodels.

LULC variables accounted for 24.0%–29.8%of thetotal variance in each dataset (corresponding to28.4%–50.7% of the variance explained by eachmodel). We note that some of the effect of soil oredaphic changes will be represented here by LULCvariables, due to the nature of peatlands and wetlandsas both landcovers and hydroecosystems.

Predictions for settlements across Kalimantan aremapped for village flooding frequency in figure 3, vil-lage flooding trends in figure 4, and news-reportedfloods infigure 5.

LULC variables consist of distance and areametrics for 20 LULC classes. Hydrography consists ofdistance to the nearest river systems, and watershedextents. Topography consists of slope, elevation, anddistance to coastline. Population and infrastructureconsist of population densities (2011), road densities,and urban cover (2010). Climate refers to long-termmean precipitation and temperature variables(1950–2000). Soil variables consist of soil family (oxi-sol, histosol etc) and changes in soil saturated watercontent under non-native LULC.

As expected, the frequency of recent flooding (inthe past five years) was lower in villages located farfrom rivers and at higher elevations (figure S4). Boththe recent frequency, and the likelihood of a long-termincrease in frequency (trend over the past 30 years),were lower for villages near to natural wetlands andwhen natural wetlands covered a large proportion ofthe watershed (figure 2, S4, S10). A long-term increasein flood frequency was more likely for areas <200 kmfrom the coastline or on flatter slopes, and in areaswith lower long-term average rainfall of the wettestmonth (less than 400 mm; possibly reflecting lowerhistorical flood frequencies that made any increases in

frequency more noticeable), or with higher rainfallseasonality (figure S10).

The frequency of recent floods and the probabilityof flooding trends over the past 30 years were both lar-ger for villages closer to mines (figure 2), and all 31 vil-lages in watersheds with >0.5% cover of mines werepredicted to experience flooding at least every 2 years(figure S6). Flood frequencies were also higher for vil-lages close to other open or disturbed landcovers (e.g.bare or grassed areas and open peatlands).

Recent flood frequencies were slightly lower inwatersheds with larger areas of peatland upstream(figure S5), while villages located very close to peat-lands were more likely to have experienced an increasein flood frequency over the past 30 years (figure 2).This difference between villages in watersheds withupstream peatlands, versus villages within or nearpeatlands themselves, is consistent with peatlands hav-ing a role in regulating the floodwaters that reachdownstream areas, even if flooding of peatlands them-selves has increased in some areas over the past 30years.

For flooding trends, an increase in flood frequencyover the past 30 years was less likely in watersheds withgreater cover of logged or intact forests, and morelikely in watersheds with more extensive oil palmplantations (figure S11). Recent flood frequencieswere lower in watersheds with higher cover of loggedforests or agroforest/regrowth cover (figure S7).Regarding distance to logged forests, villages very closeto the nearest logged forests (<5 km)were predicted tohave higher mean recent flood frequencies (figure 2),but were nomore likely to have experienced long termincreases in frequency. The association with recentfrequencies could arise, for example, if logged foreststhat occurred in areas that naturally flood frequently,weremore likely to be left standing (and not convertedto other land uses), than forests in less frequently floo-ded areas.

Of the news-reported flood events, strong positiverelationships were identified with the distance fromrivers (especially major rivers) and urban areas, higherimpervious cover and lower forest cover (figure 2,table S9). Flooding probability increased sharply asimpervious surface cover increased from 0%–3%(measured at subwatershed level), without furtherincreases beyond this threshold. Floods were alsomore likely in watersheds with higher oil palm cover,or at larger distances from intact forest or agroforest/regrowth (table S9). In relation to wetlands, news-reported floods were less likely in subwatersheds withmore wetlands (>4% cover), and floods were slightlyless likely in watersheds with greater peat soil areas.Despite strong relationships with urban and imper-vious cover, the news reports did not appear to bebiased to only reporting floods in densely populatedareas. This is based on three observations, (i) thereports spanned the full range of population densities,and 71% of flooded settlements were rural villages or

5

Environ. Res. Lett. 11 (2016) 064016

towns (not cities); (ii) reports also covered areas220 km from the coastline (figure S14); (iii) predictedflooding was more likely within 50 km of an urbanarea, but showed no further declinewith distance.

Population variables (2011 population density inthe local area or the watershed) contributed very littleto any of the BRTmodels. Population effects per se canbe a concern in studies of flooding trends because ofthe risk that perceived trends are due to changes inflood detection (exposure and observation), rather

than flood occurrence. However, we conclude that vil-lage leaders’ perceptions of trends in flood frequencyover recent decades are largely independent of popula-tion per se, and flooding trends relate more to biophy-sical changes in the landscape that are captured byother variables. This conclusion is based on our defini-tion of a village flood event (flooding of the main roador path, which is less sensitive to population size thanother definitions such as a number of houses flooded),the very small effects attributable to current

Figure 2. Influence of selected landscape variables on (a) village flood frequency over the pastfive years, (b) village trends of increasingflood frequency over the past 30 years, and (c) likelihood of a news-reportedflood during 3 years April 2010–2013. Each plot shows theBRTmodel’s fitted function for a flooding variable (y axis) versus values of a predictor variable (x axis).Mines are open-cut coal orgoldmines. All distances are inmetres.Wetlands SWS,Oil palm SWS, or Logged area SWS gives the cover of wetlands or oil palm orlogged forest, as a percentage of the area of the local subwatershed. Plots in each row share the same scale on the y axis, and show themagnitude of change in predicted values (when integrated over all other variables), relative to themean prediction (the grey linethrough zero). Forflood frequency (top row), predictions are power-scale frequencies scaled to amean of zero. Forflooding trendsand news-reported floods, the values are logit-transformed probabilities, scaled to amean of zero. Each predictor variable’scontribution to themodel is given in brackets (%of explained variance), and the observed distribution of the predictor variable isshown by vertical lines above the x axis,marking deciles of the distribution across settlements. Expanded figures showing the top sixpredictors for each analysis are given in S2 supplementary results.

6

Environ. Res. Lett. 11 (2016) 064016

population size, and the lack of detectible effects ofpopulation growth at district level (see supplementaryresults 2.1).

3.4. Comparison of predictions fromalternativedata sources: village interviews versus news-reportedfloodsComparingmodels based on the two independent datasources (village interviews and news reports), we notestrong similarities in the relationships estimatedbetween flooding and landscape variables, and strongspatial concordance in those areas at low or highflooding hazards (figure 6, S16). Regarding relation-ships with landscape variables, some variables differ inthe relative strengths of their influence between theanalyses (i.e. % of explained variance), for example,the news dataset spans a larger range of populationdensities and impervious cover, giving greater powerto detect their influences. However, even where thepercentage of variance differs, the relationships showsimilar functional forms in both analyses (figure 2 andsupplementary figures S4, S10 and S14).

Comparing predictions of village flood fre-quencies versus news-reported floods (figure 6), we seestrong spatial concordance in the areas predicted byboth datasets to have high flooding hazards (darkreds), or low flooding hazards (pale greens). News-reported floods were very rarely predicted to occur inareas where the village models indicated low flood fre-quencies (pale reds). The main difference in predic-tions occurs in areas where village interviews predictedhigh flood frequencies, but news-reported floods werepredicted to be absent (dark greens). These areas areusually further from major rivers or urban centres,and may reflect lower coverage by news reports ofareas that are more remote, or of floods that are veryregular or affect very small populations. Furtherdetailed comparisons are given in S2.6.

Comparing predictions of village flood trends, wefound that news-reported floods were more likely inareas that were also predicted to have experienced a

trend in village flooding (i.e. where village flood fre-quencies were predicted to have increased over thepast 30 yr, table S13). Spatial concordance is marked(figure S16), though not as strong as for village floodfrequencies versus news reports (figure 6).

4.Discussion

The experiences of villagers interviewed in this studyindicate that flooding patterns have changed consider-ably over recent decades. Nearly 60% of villagesreported flooding one or more times per year, andfloods were perceived to have become more frequentover the past 30 years in approximately a quarter ofvillages that ever experienced floods. Historically,flooding has formed an important natural process onBorneo. Early flood impacts on Bornean societies arereflected in oral traditions and historical writings [28].In areas of seasonal or periodic flooding, people havetraditionally lived in dwellings high above the ground,and some flooding may have been welcomed forbeneficial consequences such as nutrient input intoalluvial areas, or abundant fish supplies followingfloods [45]. Larger or unexpectedfloods however oftenhad clearly negative impacts on swidden agriculture,food supplies or caused epidemic disease outbreaks[46, 47]. Currently available data do not enable aquantitative analysis of changes through time, basedon empirical measurements of flood heights orvolumes. Nonetheless, our independently collecteddatasets on local knowledge and news reports give astrong indication of the direction of change, and oflarge, current impacts.

Landscapes throughout the tropics have under-gone rapid changes over recent decades, and growingpressures on land and water resources give greaturgency to understanding the roles of natural andmodified ecosystems in regulating flooding. Relation-ships between flooding, vegetation and soils are com-plex, and oversimplifications have been the source ofmajor contention [7]. If only a single distinction ismade between ‘intact versus modified’ land covers,this ignores immense variation within these classes,for example intensive cropping versus multi-storyagroforests [48]. Secondly, it is clear that extremes ofprolonged and heavy rainfall can overwhelm storagecapacities of soils and aquifers, and lead to floodingwhatever the dominant land cover. Acknowledgingthis complexity, recent hydrological studies haveemphasised that forest ecosystems, especially thosewith intact soils and groundcover, can slow watermovement and reduce flood peak volumes and velo-cities from rainfall events that are smaller but occurmore often than those causing massive disasters[9, 11, 13, 14, 49].

Our analyses have found that flooding prob-abilities and reported trends in flooding are related tolandscape features, especially the extents of

Table 1. Influence of landscape variables onflooding patterns, as apercentage of total variance. Values give the percentage of total var-iance that is attributed to each class of landscape variables (i.e. thepercentage of the dataset’s total variance, not explained variance).The percentages sum to the total variance accounted for by eachmodel.

Variable class

Village flood

frequency

Village

flood

trends

News-

reported

floods

LULC 29.8 24.5 24.0

Hydrography 11.5 4.5 36.0

Topography 7.1 6.3 7.5

Population and

infrastructure

4.2 3.8 12.7

Climate 3.2 9.1 1.5

Soils 3.0 1.6 2.7

Model total 58.8 49.9 84.4

7

Environ. Res. Lett. 11 (2016) 064016

impervious cover, open coal and gold mines, oil palmplantations, the extent and condition of forests andwetlands, and changes in soil water storage capacity.These findings are supported by the similarity ofresults from two independent data sets.

Wetlands and peatlands appeared to have a protec-tive role in relation to flooding in village survey

analyses, and to a smaller extent in the news analyses.Larger upstream areas of wetlands and peatlands areassociated with lower flood frequencies. Logging orclearing of peatlands may have contributed to increas-ing flood frequency over the past 30 years, since thesetrends were associated with larger areas of logged,open canopy or bare peatlands, and were common

Figure 3.Map offlood frequencies from village interviews, andmodelled predictions for populated areas of Kalimantan. Interviewdata is from302 villages. Flood frequency predictions were generated for populated areas (centre points of 1 km cells with populationdensity>1.2 km–2, except dense urban areas) fromBRTmodels relating the villages’flood frequency to 35 environmental and socialpredictor variables. Insetmap: detailed view of the area ofWest Kalimantan (expands the area of the grey box in themainmap).

8

Environ. Res. Lett. 11 (2016) 064016

among villages surveyed in eastern Central Kali-mantan where large peatland areas have been con-verted to plantation agriculture [e.g. 50]. This isconsistent with studies from across Indonesia on thehydrological consequences of clearance and drainage

of wetlands and peatlands [29, 51]. In peatlands ofSumatra and Kalimantan, those studies show thatdeforestation and drainage clearly increases localflooding and leads to higher flood peaks (and lowerbaseflow) within and downstream of peatlands [52].

Figure 4.Map offlooding trends over the past 30 years (decline, no change or an increase inflood frequencies over the last 30 years),showing observed trends from interviews in 232 villages, andmodelled predictions for populated areas of Kalimantan. Trendpredictions estimate the relative likelihood that an area has experienced an increase inflood frequencies. Crimson and red areas arethosewhere a trend is predicted (i.e. themodelled value is above the threshold optimised for distinguishing absence versus presence ofa trend). Predictions were generated for populated areas (except dense urban areas) fromBRTmodels relating presence/absence of anincrease inflood frequency (from village interviews) to 35 environmental and socio-economic variables. Insetmap: detailed view ofthe area ofWest Kalimantan (expands the area of the grey box in themainmap).

9

Environ. Res. Lett. 11 (2016) 064016

For peatlands at the lowest elevations, subsidence isprojected to lead to loss of drainability and eventuallyto permanent inundation [29, 53]. A recent study ofPeninsular Malaysia’s river basins [9] found that whileflooding (days flooded per month) was increased by

deforestation of inland forests, flooding appeared tohave reduced when the forests converted were wetlandforests (peat swamps and mangroves). The authorssuggested that improved drainage of peatlands mayreduce flood peaks by reducing antecedent water

Figure 5.Map of news-reported floods andmodelled flood probabilities for populated areas of Kalimantan.Points show the locationsof 380 news-reported floods over the period April 2010–April 2013. Predictions were generated for populated areas fromBRTmodelsrelating the presence/absence of reportedfloods to 35 environmental and socio-economic variables. Predictions give the relativeprobability of a flood occurring over a 3 yr period, under conditions similar to those experienced in April 2010–April 2013.Mid anddark red areas are thosewhere flooding is predicted (i.e. themodelled value is above the threshold optimised for distinguishingabsence versus presence of a reported flood, 0.42). Insetmap: detailed view of the area ofWest Kalimantan (expands the area of thegrey box in themainmap).

10

Environ. Res. Lett. 11 (2016) 064016

Figure 6. Spatial comparison of predicted flooding hazards based on village interviews versus news reports. Darker colours indicatehigher flood frequencies predicted from village interviews (flood frequency over the past 5 years,mainly 2004–2009). Red coloursindicate areas where news-reported floodswere predicted to occur (higher relative likelihood of a news-reported flood over the 3 yrsApril 2010–April 2013). Darker reds therefore indicate areas where both data sources gave predictions of highflooding hazards. Greencolours indicate areas where floodingwas not predicted fromnews reports. Pale greens therefore indicate the areas agreed by both datasources to have lowflooding hazards. Dark green indicates high village flood frequencies in areas where news-reported floodswerepredicted to be absent.Model details: Predictions were generated for populated areas fromBRTmodels relating either village floodfrequencies (over the past 5 yrs,mainly 2005–2009), or news-reportedflooding (presence/absence of reported floods over the 3 yrsApril 2010–April 2013), to 35 environmental and socio-economic variables. Predictions from eachmodel (continuous values for1 km2 pixels)were grouped into classes to enable visual comparison. News-reported floodswere predicted to occurwhere pixel valueswere above the threshold of 0.427. Insetmap: detailed view of the area ofWest Kalimantan (expands the area of the grey box in themainmap).

11

Environ. Res. Lett. 11 (2016) 064016

levels, but this effect appears to have been small orabsent in the Indonesian studies. It is also possible thatflooding observations in [9]may not reflect conditionswithin or downstream from peatlands, since the spa-tial relations between the peatlands and the observa-tion points for river heights are unknown.

The hydrological effects of land use change occurin large part through changes in soil physical proper-ties and ground-level cover [6, 54, 55]. In our analyses,flooding increased slightly with changes in soil waterstorage capacity of the upstream watershed (estimatedfrom present versus natural LULC), but effectsappeared small in relation to the effects of peatlandsand other wetlands. However, available soil data isvery limited in information on depth and especially oncurrent condition (affecting infiltrability, conductivityand network connectivity). We were only able to esti-mate soil degradation via association with LULC clas-ses, and although this association is generally strong,there are some land uses within which soil conditionmay vary widely (especially agroforest/regrowth), andthe temporal dynamics of soil degradation and recov-ery can also have important consequences for hydro-logical services [8].

Our analyses highlight several relationships, andmotivate further studies to clarify hydrologicalmechanisms, and the effects of specific managementsystems, and to enable quantitative projections underalternative land use and climate change scenarios. Inparticular, they highlight an urgent need to betterunderstand the hydrological effects of oil palm planta-tion design and management [56], and demonstratethat the spatial distributions and roles of wetlands andpeatlands in regulating flooding, deserve strong focusin future research. Existing maps of peatlands showlarge underestimations and uncertainties compared tothe small areas where detailed depth studies have beenconducted [29]. New initiatives for mapping of peatdepths [57] have been completed only for small pilotareas, but further mapping will also be essential toimplement new legislation requiring protection ofpeatlands [58].

By analysing an independent dataset of newsreports over three years, we start to gain a picture offrequent and extensive impacts from flooding. Weemphasise that our quantitative estimates understatethe full magnitude and range of impacts beyond flood-ing of houses, which include effects on agriculture,education, business, and health. Our analyses consideronly riverine and flash flooding, and overall floodingimpacts would be higher when considering coastalflooding from storm surges and tidal inundation(alone or in interaction with riverine floods). Evenwhile the full social and economic impacts of floodingon Borneo are not yet known [39, 59], it is clear thatplanning formitigation and adaptation is not yet givenadequate priority by local and national governments.Our survey of news articles found they were mostoften focused on immediate impacts and emergency

responses, but often also made links to land use andespecially deforestation andmining. Examples includearticles from South Kalimantan, where disaster pre-paredness officials attributed floods to mining onmountain slopes [60], and from East Kalimantan,where many reports attributed floods to impacts ofcoal mines [61, 62]. Mining, especially of coal, hasundergone extremely rapid expansion in Kalimantansince 2004 [63]. In our analysis, areas downstreamfrom coal or gold mines experienced higher flood fre-quencies and likelihood of increases over the past 30years, indicating a strong need for quantifying miningimpacts on hydrology and river systems, and compre-hensive collation of data on industrial scale and small-scale, often informal operations.

4.1. Comparisonwith government records and riskassessmentsGovernment knowledge of flood events in Borneoappears to be fragmented and highly incomplete.While national and regional governments on Borneoacknowledge that floods represent major economicand social costs [64, 65], quantitative data and assess-ments are lacking, especially for regional areas (S2.7).Indonesia’sNational ActionPlan forDisasterManage-ment 2010–12, for example, does not include adetailed flood risk assessment [66]. We compared ourestimates from news-reported floods with recordsfrom Indonesia’s publicly accessible online DisasterLoss Database (DiBi—Data dan informasi BencanaIndonesia) for the same time period. This nationaldatabase is a model of data transparency and accessi-bility, however the data it contains is highly incom-plete, and dramatically underestimates the number oflarge flooding events and their impacts (tables S14,S15). For example, the single largest event in ournewspaper survey was reported by the regional SouthKalimantan Disaster Management Agency to haveflooded 24 889 houses in 299 villages along theMartapura River in April 2010 (table S14). The samedisaster is recorded in the national database as affect-ing 929 people, and causing 5 deaths. Overall, thenumber of people reported in the DiBi database to beimpacted by flooding is 10–100 times smaller than thatestimated from the newspaper records, for all pro-vinces except Central Kalimantan (table S15).

Indonesia’s national methods for flood reportingand risk assessment are geared to prioritisation ofemergency response to intensive disasters, i.e., highimpact events that affect large, spatially concentratedpopulations, such as those in the capital Jakarta. How-ever, they have been based on brief and incompletetime series of events, and need to better integrate themonitoring efforts of local and regional agencies(S2.7). Governments do not adequately assess or acton large cumulative risks from events that affect morespatially distributed populations, or occur at high fre-quency, and add up to large impacts over a region or

12

Environ. Res. Lett. 11 (2016) 064016

time period. We also note that the existing method forrisk assessment only considers areas with major settle-ments or agricultural production, and for which �1major event appear in the (incomplete) national data-base, giving a map with no predicted hazard or risklevel for the majority of the land area of Kalimantan(figure S18). The importance of disasters of lowermagnitude but high frequency has been highlighted inrecent studies in Indonesia, Bolivia, Mexico, Mozam-bique, Nepal, the Philippines, and Vietnam showingnegative impacts on children’s education, health, andaccess to services such as water and sanitation [18].Note that we do not advocate replacing disasterreporting mechanisms with news surveys. These com-parisons indicate the need to support consistentrecording at the local level (and maintain existingefforts by agencies such as the South Kalimantan Dis-aster Management Agency). Secondly, they indicatethe scope for better communication of flood eventsand impacts among local and Provincial authorities,and incorporation into national databases.

5. Conclusions

The results from our studies of both interviews andnews reports show that flooding is an important andwidespread social and economic problem in Kaliman-tan. We have identified large areas that experience highflooding hazards, and associations between floodingfrequency (and perceived trends) andLULC that call forfurther investigation. Furthermore, flooding hazardsare projected to increase over this century due to climatechange intensifying regional water cycles and rainfallevents [67–69], sea level rise [68], and modification ofcoastal ecosystems including drainage of lowland peats,leading to rapid subsidence [53]. Finally, these changesin flooding hazard are likely to combine with increasesin flooding exposure and vulnerability as populationdensities increase in cities and many rural towns andvillages over the coming decades [70].

Clearly there is a need, alongside emergency pre-paredness and response [71], for long-term risk reduc-tion and adaptation. Deeper understanding of therelationships between LULCCs and the risks of flood-ing would help inform flood mitigation and adapta-tion strategies, and enable a fuller understanding of theconsequences of conversion or conservation of naturalandmodified ecosystems. Recognising the longer termimpact of land use decisions on environmental ser-vices such as flood prevention would make an impor-tant contribution towards securing a more sustainablefuture in Borneo.

Acknowledgments

We thank the many respondents who volunteeredinformation in the interview surveys, and thank TheNature Conservancy for funding the surveys and

providing access to this dataset.We are grateful to stafffrom CIFOR (The Center for International ForestryResearch), the World Agroforestry Centre, and theRiver Basin Planning section of the IndonesianGovernment Ministry of Public Works, for discussingdata sources and their current research on floods andforests. KAW and JW acknowledge funding from TheUniversity of Queensland and the Australian ResearchCouncil (ARC) Future Fellowship FT100100413 andARC Centre of Excellence program CE110001014.KM also acknowledges support from the ARC. TheBorneo Futures network acknowledges financial sup-port from the United Nations Environment Program(DEPI Regional Office for Asia and the Pacific) and theArcus Foundation.

References

[1] MillenniumEcosystemAssessment 2005Ecosystems andHumanWell-Being: Synthesis (Washington,DC: Island)

[2] FAO2011 State of theWorld’s Forests 2011 (Rome: Food andAgricultureOrganization of theUnitedNations)

[3] BraumanKA,Daily GC,Duarte TK andMooneyHA2007The nature and value of ecosystem services: an overviewhighlighting hydrologic servicesAnnu. Rev. Environ. Resour. 3267–98

[4] Buschman FA,HoitinkA J F, de Jong SM,Hoekstra P,HidayatH and SassiMG2012 Suspended sediment load in thetidal zone of an Indonesian riverHydrol. Earth Syst. Sci. 164191–204

[5] McIntyreN et al 2014Modelling the hydrological impacts ofrural land use changeHydrol. Res. 45 737–54

[6] Bruijnzeel L A 2004Hydrological functions of tropical forests:not seeing the soil for the trees?Agric. Ecosyst. Environ. 104185–228

[7] vanDijk A I JM, vanNoordwijkM,Calder I R,Bruijnzeel S LA, Schellekens J andChappell NA2009 Forest-flood relation still tenuous—comment on ‘Global evidencethat deforestation amplifiesflood risk and severity in thedevelopingworld’ byC JABradshaw,NS Sodhi, K S-HPehandBWBrookGlob. Change Biol. 15 110–5

[8] vanNoordwijkM,Agus F, Verbist B,HairiahK andTomichTP 2007Watershedmanagement ed S Scherr andJMcNeely Farming withNature: The Science and Practice ofEcoagriculture (Washington, Covelo, London: Island)pp 191–212

[9] Tan-Soo J-S, AdnanN, Ahmad I, Pattanayak SK andVincent J R 2014 Econometric evidence on forest ecosystemservices: deforestation and flooding inMalaysia Environ.Resour. Econ. 63 25–44

[10] CameronC,Norrington-Davies G, Back E andMitchell T 2012SREX: Lessons for theWater Sector (Agulhas: AppliedKnowledge, OverseasDevelopment Institute)

[11] WahyuA,KuntoroAA andYamashita T 2010Annual andseasonal discharge responses to forest/land cover changes andclimate variations inKapuas River Basin, Indonesia J. Int. Dev.Coop. 16 81–100

[12] Bathurst J C et al 2011 Forest impact onfloods due to extremerainfall and snowmelt in four Latin American environments:II.Model analysis J. Hydrol. 400 292–304

[13] Lusiana B,WidodoRH,VanNoordwijkM,Mulyoutami E andNugrohoDA2008RapidHydrological Appraisal: TwoCaseStudies in Integrating Knowledge ofWatershed Functions inIndonesia (Bogor, Indonesia:WorldAgroforestry Centre(ICRAF) Southeast Asia RegionalOffice)

[14] Ogden F L, CrouchTD, Stallard R F andHall J S 2013 Effect ofland cover and use on dry season river runoff, runoff efficiency,and peak storm runoff in the seasonal tropics of CentralPanamaWater Resour. Res. 49 8443–62

13

Environ. Res. Lett. 11 (2016) 064016

[15] NagyRC, Lockaby BG,Helms B, Kalin L and Stoeckel D 2011Water resources and land use and cover in a humid region: thesoutheasternUnited States J. Environ. Qual. 40 867–78

[16] Ziegler AD, BruunTB,Guardiola-ClaramonteM,Giambelluca TW, LawrenceD and LamNT2009Environmental consequences of the demise in Swiddencultivation inMontaneMainland Southeast Asia: hydrologyand geomorphologyHum. Ecol. 37 361–73

[17] Hamilton L S,DudleyN, GremingerG,HassanN, LambD,Stolton S andTognetti S 2008 Forests andWater: AThematicStudy Prepared in the Framework of the Global Forest ResourcesAssessment 2005 (Rome: Food andAgricultureOrganization oftheUnitedNations)

[18] UNISDR2011Global Assessment Report onDisaster RiskReduction (Geneva, Switzerland:UnitedNations InternationalStrategy forDisaster Reduction)

[19] Pramova E, Locatelli B, DjoudiH and SomorinOA2012Forests and trees for social adaptation to climate variability andchangeWiley Interdiscip. Rev. Clim. Change 3 581–96

[20] Merz B et al 2014 Floods and climate: emerging perspectivesforflood risk assessment andmanagementNat. Hazards EarthSyst. Sci. 14 1921–42

[21] Ferreira S andGhimire R 2012 Forest cover, socioeconomics,and reportedflood frequency in developing countriesWaterResour. Res. 48W08529

[22] WinsemiusHC, VanBeek LPH, JongmanB,Ward P J andBouwmanA 2013 A framework for global river flood riskassessmentsHydrol. Earth Syst. Sci. 17 1871–92

[23] BrakenridgeGR 2014Global active archive of largefloodeventsDartmouth FloodObservatory (Colorado,USA:University of Colorado) (http://floodobservatory.colorado.edu/Archives/index.html)

[24] Guha-SapirD, BelowR andHoyois P 2014EM-DAT:International Disasters Database (Louvain LaNeuve, Belgium:Université Catholique de Louvain) (http://emdat.be/)

[25] Doocy S, Daniels A,Murray S andKirsch TD2013The humanimpact offloods: a historical review of events 1980–2009 andsystematic literature review PLOSCurr. Disasters (doi:10.1371/currents.dis.f4deb457904936b07c09daa98ee8171a)

[26] AbramNK,Meijaard E, AncrenazM, Runting R K,Wells J A, GaveauD L A, Pellier A-S andMengersen K 2014Spatially explicit perceptions of ecosystem services and landcover change in forested regions of Borneo Ecosyst. Serv. 7116–27

[27] Runting RK et al 2015Alternative futures for Borneo show thevalue of integrating economic and conservation targets acrossbordersNat. Commun. 6 6819

[28] KnapenH2001 Forests of Fortune? The Environmental Historyof Southeast Borneo, 1600–1880 (Leiden: KITLVPress)

[29] QANS2013PeatlandMaps for Indonesia—Accuracy Assessment—Recommendations for Improvement—ElevationMapping–Evaluation of Future FloodRisk, AReport from theQuickAssessment andNationwide Screening (QANS) of Peat andLowland Resources andAction Planning (Jakarta, Indonesia:BAPPENAS&Ditjen SumberDayaAir, KementerianPerkerjaanUmum,Government of Indonesia & Partners forWater Programme—TheNetherlands)

[30] de BruynM et al 2014 Borneo and Indochina aremajorevolutionary hotspots for Southeast Asian biodiversity Syst.Biol. 63 879–901

[31] Wich SA et al 2012Understanding the impacts of land-usepolicies on a threatened species: is there a future for theBorneanOrang-utan? PLOSONE 7 e49142

[32] Sha J CM,Henry B andNathan S 2008 Status and conservationof proboscismonkeys (Nasalis larvatus) in Sabah, EastMalaysiaPrimate Conserv. 23 107–20

[33] Meijaard E, AbramNK,Wells J A, Pellier A-S, AncrenazM,GaveauDLA, Runting RK andMengersenK 2013 People’sperceptions on the importance of forests on Borneo PLOSONE 8 e73008

[34] GaveauDLA et al 2013Reconciling forest conservation andlogging in Indonesian Borneo PLOSONE 8 e69887

[35] MargonoBA, Potapov PV, Turubanova S, Stolle F andHansenMC2014 Primary forest cover loss in Indonesia over2000–2012Nat. Clim. Change 4 730–5

[36] GaveauD, Sloan S,Molidena E,HusnayaenH, Sheil D,AbramN,AncrenazM,Nasi R,WielaardN andMeijaard E2014 Four decades of forest persistence, clearance and loggingonBorneo PLOSONE 9 e101654

[37] Hijmans R J, Cameron S E, Parra J L, Jones PG and Jarvis A2005Very high resolution interpolated climate surfaces forglobal land areas Int. J. Climatol. 25 1965–78

[38] CaseM, Ardiansyah F and Spector E 2007Climate Change inIndonesia—Implications for Humans andNature (Jakarta,Indonesia:WWF—WorldWildlife Fund)

[39] AbramNK et al 2014 Synergies for improving oil palmproduction and forest conservation infloodplain landscapesPLOS ONE 9 e95388

[40] ESRI 2013ArcGISDesktop: Release 10.2[41] Elith J, Leathwick J R andHastie T 2008 Aworking guide to

boosted regression trees J. Anim. Ecol. 77 802–13[42] RCore Team2014R: A Language and Environment for

Statistical Computing (Vienna, Austria: R Foundation forStatistical Computing)

[43] Hijmans R J, Phillips S, Leathwick J R and Elith J 2013 Dismo v08-11CRAN—Comprehensive RArchiveNetwork)

[44] PemprovKaltim 2013 Bantuan Pemprov untuk PenangananBanjir di Samarinda Rp605Miliar—Provincial GovernmentAssistance for FloodManagement in Samarinda 605 BillionRupiah 2011–2013, Pemerintah Provinsi KalimantanTimur—Provincial Government of East Kalimantan (http://kaltimprov.go.id/berita-257-bantuan-pemprov-untuk-pananganan-banjir-di-sanarinda-rp605-miliar-.html)

[45] Mallinckrodt J 1925 EthnografischeMededeelingen over deDajaks in de Afdeeling Koealakapoeas. (Res. Zuider- EnOosterafd. VanBorneo)Bijdr. tot Taal-, Land- enVolkenkd.vanNed. 82 165–310

[46] KnapenH1997 Epidemics, droughts, and other uncertaintieson Southeast Borneo during the eighteenth and nineteenthcenturies ed PBoomgaard, FColombijn et al Paper Landscapes:Explorations in the Environmental History of Indonesia (Leiden:KITLVPress) pp 121–52

[47] Enthoven J J K 1903Bijdragen Tot deGeografie vanBorneo’sWester-afdeeling vol 2 (Leiden, TheNetherlands: E J Brill)

[48] vanNoordwijkM2008RapidHydrological Appraisal (RHA):An Integrated Approach to AssessWatershed Functions andManagementOptions (Bogor, Indonesia:WorldAgroforestryCentre—ICRAF, SEARegionalOffice)

[49] Bathurst J C et al 2011 Forest impact onfloods due to extremerainfall and snowmelt in four Latin American environments: I.Field data analysis J. Hydrol. 400 281–91

[50] FullerDO,HardionoMandMeijaard E 2011Deforestationprojections for carbon-rich peat swamp forests of CentralKalimantan, IndonesiaEnviron.Manage. 48 436–47

[51] Wetlands International 2011 Peatlands in Indonesia’s NationalREDD+ Strategy. Responding to the Public Consultation of theDraft National REDD+ Strategy of Indonesia Issued onAugust18, 2011. Version 2

[52] Hooijer A, van der VatM, PrinsenG,VernimmenR,Brinkman J andZijl F 2008 Master Plan for the RehabilitationandRehabilitation of the Ex-Mega Rice Project Area inCentralKalimantan. Technical Report No. 2. Hydrology of the EMRPArea:WaterManagement Implications for PeatlandsEuroconsultMottMacDonald andDeltares—DelftHydraulics)

[53] Hooijer A, Page S, Jauhiainen J, LeeWA, LuXX, Idris A andAnshari G 2012 Subsidence and carbon loss in drained tropicalpeatlandsBiogeosciences 9 1053–71

[54] Sidle RC, Ziegler AD,Negishi J N,NikAR, SiewR andTurkelboomF2006 Erosion processes in steep terrain—truths,myths, and uncertainties related to forestmanagementin Southeast Asia For. Ecol.Manage. 224 199–225

[55] ZimmermannB, Papritz A and ElsenbeerH 2010Asymmetricresponse to disturbance and recovery: changes of soil

14

Environ. Res. Lett. 11 (2016) 064016

permeability under forest-pasture-forest transitionsGeoderma159 209–15

[56] Comte I, Colin F,Whalen J K, Grüberger O andCaliman J-P 2012 Agricultural practices in oil palmplantations and their impact on hydrological changes,nutrient fluxes and water quality in Indonesia: a reviewAdv. Agron. 116 71–124

[57] ICCC2013PeatlandDefinition and PeatlandMappingMethodology Assessment—Final Project Report December 20,2013 (Jakarta: Indonesia Climate Change Center) (http://iccc-network.net)

[58] President of the Republic of Indonesia 2014 IndonesianGovernment Regulation 2014–71 on Protection andManagement of Peat Ecosystems—Peraturan PemerintahRepublik IndonesiaNomor 71 Tahun 2014 TentangPerlindunganDan Pengelolaan EkosistemGambut (Jakarta,Indonesia: Government of the Republic of Indonesia)

[59] KTAT2003National Register of River Basins. Final ReportVolume 2:Updating of Condition of Flooding inMalaysia (KualaLumpur,Malaysia: Jabatan Pengairan dan Saliran,Government ofMalaysia, andKTATenaga SdnBhd)

[60] Media Indonesia 2010 Banjir KalselMeluas—Widespreadflooding in SouthKalimantanKalimantanNews (http://kalimantan-news.com/berita.php?idb=176)

[61] Maimunah S 2013 Banjir DanKeselamatanWargaIndoprogress 4 Februari 2013 (http://indoprogress.com/

banjir-dan-keselamatan-warga/)[62] Sabarini P and SulaimanN2013 Samarinda floods blamed on

mining Jakarta Post (http://thejakartapost.com/news/2013/01/18/samarinda-floods-blamed-mining.html)

[63] Lucarelli B 2010The history and future of Indonesia’s coalindustry: impact of politics and regulatory framework onindustry structure and performance Program on Energy and

Sustainable Development,Working Paper#93 StanfordUniversity, Stanford

[64] BNPB2013Report from Indonesia’s National Consultation onHFA2 and Post-2015Development Agenda, February 11, 2013(Jakarta, Indonesia: IndonesiaNational Agency forDisasterManagement—BNPB)

[65] MKN2012Malaysia: National Progress Report on theImplementation of theHyogo Framework for Action (2011–2013)—Interim (Putrajaya,Malaysia: National Security Council ofMalaysia—Majilis KeselamatanNegara)

[66] Bappenas andBNPB 2010National Action Plan for DisasterRisk Reduction 2010–2012 (Jakarta, Indonesia: Government ofthe Republic of Indonesia)

[67] Katzfey J,McGregor J, NguyenK andThatcherM2010Regional climate change projection development andinterpretation for Indonesia Final Report for AusAIDTheCentre for AustralianWeather andClimate Research, Apartnership betweenCSIRO and the Bureau ofMeteorology

[68] IPCC2012 Special Report onManaging the Risks of ExtremeEvents andDisasters to Advance Climate Change Adaptation, ASpecial Report ofWorkingGroups I and II of theIntergovernmental Panel on Climate Change (Cambridge, UK:CambridgeUniversity Press)

[69] CaiW et al 2015 Increased frequency of extreme LaNiñaevents under greenhouse warmingNat. Clim. Change 5 132–7

[70] BPS 2011Trends of the Selected Socio-Economic Indicators ofIndonesia—Perkembangan Beberapa Indikator Utama Sosial-Ekonomi Indonesia, Nov 2011Katalog BPS: 3101015 (Jakarta,Indonesia: Badan Pusat Statistik BPS—Statistics Indonesia)

[71] Butt S 2014Disastermanagement law in Indonesia: fromresponse to preparedness?Asia-Pacific DisasterManagementSocio-legal Perspectives ed S Butt et al (Berlin: Springer)pp 183–96

15

Environ. Res. Lett. 11 (2016) 064016