Influence of environmental factors and farming technique on growth and health of farmed Kappaphycus alvarezii (cottonii) in south-west Madagascar

Influence of environmental factors and farming techniqueon growth and health of farmed Kappaphycus alvarezii (cottonii)in south-west Madagascar

Mebrahtu Ateweberhan & Antoine Rougier & Cicelin Rakotomahazo

Received: 24 February 2014 /Revised and accepted: 3 July 2014# Springer Science+Business Media Dordrecht 2014

Abstract A monitoring programme was established in orderto support community-based seaweed farming in south-westMadagascar by providing scientific information on the effectsof physico-chemical and health factors influencing the growthof Kappaphycus alvarezii (cottonii). Six aquaculture site con-figurations were studied. These consisted of high and lowflow locations, off-bottom and long-line farming techniquesand different benthic/substrate types. At each site, a number ofgrowth and health variables were monitored monthly betweenJanuary 2012 and March 2013 on 30 randomly selected thalli.Variables included thallus-level growth, intensity of epiphyteand disease infection, intensity of fish and sea urchin grazingand cover of pest seaweed and sediment. The following keyenvironmental variables were also monitored at the site level:water temperature, irradiance, salinity, water depth, waveaction, pH and oxygen content. Overall average relativegrowth rate of K. alvarezii in the region was 4.5±0.06 %day−1 and varied by site and season. Generally, growth ratewas higher during the winter season (April–August, 5.04±0.31 % day−1) than in the summer (3.90±0.28 % day−1). Thelong-line farming technique provided higher growth (5.46±0.09 % day−1) than the off-bottom technique (3.99±0.07 %day−1). Thallus-level analysis showed that fish grazing, epi-phyte cover, sediment cover and disease had significant neg-ative correlations with growth, and the four variables werepositively correlated. Site-level analysis on the effects ofphysico-chemical and health factors showed that sedimenta-tion, daily maximum of water temperature and variability, and

interactions between these factors were the main determinantsof growth. Growth was lower at high sedimentation levels andhigher values of maximum temperature and temperature var-iation. Our findings highlight that farming should focus main-ly in the cold season and long-line technique in order to limitthe major ecological constraints encountered and maintaingrowth and production at sustainable levels.

Keywords Alternate livelihoods .Marine conservation .

Environmental sustainability . Marine farming . Seaweedphysiology . Seasonality . Temperature variability . Thermalstress effects .Western IndianOcean

Introduction

As a primary source of the highly demanded cell wall poly-saccharide, κ-carrageenan, Kappaphycus alvarezii (cottonii)is one of the most commercially important seaweed species(Bixler 1996; Bixler and Porse 2010; Ask and Azanza 2002).The market for cottonii is continually expanding, resulting inincreased introduction of this species for experimentation andfarming (Munoz et al. 2004). Cottonii farming has become animportant livelihood in coastal areas of many countries (Asket al. 2003a; Munoz et al. 2004; Msuya 2006; Bindu andLevine 2011) and is expanding quickly as a source of alterna-tive income for poor coastal communities (Ask 2003; Hillet al. 2012; Pickering 2006). It is also seen as a major com-ponent of coastal management programmes aimed at decreas-ing pressure on over-exploited wild marine populations, im-proving habitat conditions (Munoz et al. 2004; Hill et al.2012) and promoting climate change adaptation by coastalcommunities. K. alvarezii has been introduced in parts of thePacific and Indian Oceans and the Caribbean-Atlantic regions.Tanzania and Madagascar were the first countries to introducethis species for commercial aquaculture in the western Indian

M. Ateweberhan (*) :A. Rougier :C. RakotomahazoBlue Ventures Conservation, Omnibus Business Centre, 39-41 NorthRoad, London N7 9DP, UKe-mail: [email protected]

M. AteweberhanDepartment of Life Science, University of Warwick, Coventry CV47AL, UK

J Appl PhycolDOI 10.1007/s10811-014-0378-3

Ocean region (Msuya et al 2014). While seaweed farming inZanzibar (Tanzania) contributes significantly to the economy,farming programmes in Madagascar have faced many chal-lenges, marked by several interruptions (Tsiresy, unpublisheddata).

InMadagascar, there has been a renewed interest by privatebusiness companies, as well as development and conservationNGOs, in reviving the country’s seaweed farming sector. Newcommunity farms have been established in the north andsouth-west regions of the country with initial investmentsmade by private businesses and conservation NGOs in part-nership with governmental development agencies. Seaweedfarming activities in the Velondriake region of south-westernMadagascar started in 2009 and are still at the trial stage. Thehigh poverty levels and the low level of income of thecommunities (US$0.83 person−1 day−1 in 2010; Barnes-Mauthe and Oleson 2013) that mostly originates from har-vest of low-value species, such as octopus, could makeseaweed farming an economically viable venture. In addi-tion, low initial input in money and material is required.However, K. alvarezii is very sensitive to small changes inenvironmental conditions, which can result in reduced pro-ductivity and eventually undermine the economic viabilityof farming projects, leading to decreased profitability, dis-couragement of farmers to continue farming and eventuallythreatening the long-term sustainability of the farming(Msuya 2006; Msuya and Porter 2014).

Understanding the environmental factors influencing sur-vival and growth of K. alvarezii and identifying technical,socio-economic and political bottlenecks are crucial for asuccessful farming and sustaining viability of a project. Inorder to identify sites that are best suited for community-basedfarming of K. alvarezii in south-west (SW) Madagascar, fieldexperiments were established in December 2011. Growth,health and environmental factors influencing growth and pro-duction of K. alvarezii were monitored for 1 year. The infor-mation gathered will be useful in understanding spatio-temporal dynamics in the growth of K. alvarezii and thefactors controlling production, providing important informa-tion to help guide management decisions concerning farmingpractices.

Methods

Study area and farming technique

Monitoring was conducted at six sites in Velondriake, SWMadagascar where community-based K. alvarezii farming istaking place (Fig. 1). Three sites were selected aroundLamboara, one at Ampampa and two at Ambolimoke.Lamboara site 2 was located in a channel (high water flow)while all the remaining sites were located in low water flow

sites (Table 1). Off-bottom lines were used at the threeLamboara sites and Ambolimoke site 2 while long lines wereused at Ampampa and Ambolimoke site 1. Hereafter,Ambolimoke site 1 is referred as Ambolimoke LL (long-line) and Ambolimoke site 2 as Ambolimoke OB (off-bottom). Off-bottom seaweed lines were attached to woodenstakes at each end and suspended about 20 cm from thebottom. Long lines were attached to a rock at both ends andsuspended with plastic bottle floats.

Monitoring of propagules

At each site, five seaweed lines of 10-m length wereestablished and populated with 50 seedlings (also referred toas propagules) of 40–275-g wet weight, each placed at equalintervals of 20 cm. Seedlings were checked thoroughly for thepresence of ice-ice disease, epiphyte infection and grazingscars; affected seedlings were discarded. Seedlings were at-tached to the main line using the ‘Made Loop’ method (Asket al. 2003b). On each line, six seedlings were selected ran-domly and tagged, their initial weights were taken and health

Fig. 1 Map of the study area (circle indicates off-bottom farming tech-nique; square indicates long-line farming technique)

J Appl Phycol

status was checked (presence of ice-ice, epiphytes and grazingscars). Thallus weight and health status were again determinedafter 4–5 weeks of growth. Monitoring of thalli continued for1 year to be able to test the full effects of season.

Monitoring of health factors

Epiphyte, pest seaweed and sediment cover The presence andseverity of epiphyte infection, pest seaweed and sedimentationwas noted and severity was estimated based on percent cover.Cover was converted to five scales of severity as follows: 0,no epiphytes, sediment or pest seaweed; 1, very low coverage(0–20 %); 2, low coverage (20–40 %); 3, moderate coverage(40–60 %); 4, high coverage (60–80 %); and 5, very highcoverage (80–100 %).

Ice-ice infection Because of the difficulty in applying a strictquantitative procedure of estimating disease severity, we useda semi-quantitative technique. This is mainly because severityvaries based on the section of the thallus affected. For in-stance, effects of disease on a main branch, secondary branchand branch tips are expected to be different. A small butnecrotic lesion occurring on a main branch could result inthe loss of a whole thallus while an infection that is thinlyspread and occurring on growing tips will probably have a lessdamaging impact. Hence, the presence and severity of ice-ice infection was estimated on a semi-quantitative scale of 0to 5: 0, no infection; 1, infection occurring on few branchesand limited to branch tips only; 2, infection occurring onmany branches but limited to branch tips only; 3, infectionreaching secondary branches and most branches infected;4, infection occurring on main branches but at a less ad-vanced stage; and 5, infection advanced and main branchesseverely infected. The use of semi-quantitative scales thathas been used in plant science, particularly crop science,has a long history (e.g. James 1971; Horsfall and Barratt1945). However, most of the techniques used differ fromthe one used in the present study as they are mainly basedon standard area diagrams and represent conversions of thepercent scale.

Grazing Preliminary surveys indicated that fish and sea ur-chins were the main grazers in the area. As in disease, it is

difficult to apply a strict quantitative procedure in estimatinggrazing intensity on growth. For instance, for the same amountof grazing, effects on a main branch, secondary branch andbranch tips are expected to be different. Similarly, grazing byurchin, fish and turtle is expected to be different. Grazing byfish and turtles could result in the loss of a large portion or awhole thallus while grazing on growing tips will result mainlyin a stunted growth. The presence of grazing scars and theirintensity was estimated again on a semi-quantitative scale of 0to 5: 0, no grazing scars; 1, thallus healthy and bushy andbranching with a few grazed branch tips; 2, thallus healthy andbushy with obvious signs of grazing on tips; 3, about half ofthe branch tips grazed; 4, most growing tips grazed with a fewgrowing tips remaining ungrazed; and 5, no growing tipsremaining ungrazed and signs of grazing on secondary andprimary branches. In addition, grazing was estimated byconducting visual surveys of major grazers along the seaweedlines by counting fish grazers within 2.5×2.5 m on either sideof a 10-m seaweed line (10-×5-m belt transect). Fish size wasestimated to the nearest cm according to the following cate-gories: 0–5, 5–10, 10–25 and 15–20 cm. The biomass of seaurchins in the area was used as a proxy to grazing by seaurchins. The number of sea urchins within 2.5×2.5 m oneither side of the seaweed line was counted, and biomasswas estimated in kg m−2 based on size-biomass conversionsof randomly selected individuals.

Monitoring of physico-chemical parameters

Water temperature was measured at hourly intervals using insitu water temperature loggers (HOBO, Onset ComputerCorporation, California). One data logger each was placed ateach site. One data logger was lost fromAmbolimoke LL, anddata were available only for January to May for AmbolimokeOB. Salinity was measured on-site using a handheld refrac-tometer with a minimum of three measurements taken at eachsite during monthly surveys. Water depth was measured inmetres using a graduated measuring stick at spring low tideand on a monthly basis. Dissolved oxygen (μmol L−1) wasdetermined with a portable digital oxygen/pHmeter with threereadings taken at each site during monthly surveys. Waveactivity was estimated using a modified Beaufort scale (0:calm–smooth condition, 1: slightly rough–light air to gentle

Table 1 Geographic location,characteristics of the study sitesand farming techniques used

Site Farming technique Location Wave exposure Substrate/benthos

Ambolimoke 1 Long-line Out of channel Low Silt/fine sand

Ambolimoke 2 Off-bottom Out of channel Low Silt/fine sand/seagrass

Ampampa Long-line Out of channel Low Sand/coral

Lamboara 1 Off-bottom Out of channel Low Silt/fine sand/seagrass

Lamboara 2 Off-bottom In channel Low Fossil reef/course sand

Lamboara 3 Off-bottom Out of channel Low Silt/fine sand

J Appl Phycol

breeze; 2: moderately rough, moderate to strong breeze; 3:strongly rough, gale to storm).

Substrate/benthic composition

Composition of the main substrate and benthic types wereestimated using a 0.5-×0.5-m quadrat, divided into twenty-five 10×10-cm grids. Three quadrat samples were taken 1–2 m near off-bottom or under a long line by throwing thequadrat frame haphazardly, and percent substrate and benthiccover were estimated. The following substrate and benthicgroups were considered and monitored twice a year (one eachduring the hot season, December–February, and cold season,June–August): substrate—fossil limestone, course sand, finesand and silt and benthos—hard coral, soft coral, sea anemo-ne, zoanthid, sponge, tunicate, blue green algal mat, turf algae,fleshy macroalgae, crustose coralline algae, Halimeda andseagrass.

Data analysis

Growth was calculated as a relative growth rate according tothe following formula: g (% day−1) = [(ln Wf − ln Wi) / t] ×100, where Wi = initial wet weight and Wf = final wet weightafter t number of days (Dawes et al. 1993). Univariate analysiswas used to compare differences in growth and health param-eters (epiphyte cover, pest seaweed cover, intensity of disease,intensity of grazing by fish and sea urchin) by site, farmingtechnique and period. First, we performed tests of normalityand equality of variance in order to test conformity with thecriteria of parametric tests. Both Bartlett’s test of normalityand Levene’s test of equality of variance showed that data ongrowth and health parameters were not normality distributednor homogeneous. Both log 10 and square root and fourth roottransformations of data failed to normalise or homogenise thedata. As a result, the non-parametric Wilcoxon signed-ranktest (Kruskal-Wallis test) was used to test differences ofgrowth by site and period. By period comparisons for eachsite and by site comparisons for each period were also con-ducted using Kruskal-Wallis test. Since site-farming techniqueinteraction was expected, we made separate comparisonsamong sites with the same farming technique (off-bottom:Ambolimoke OB and the three Lamboara sites; long-line:Ambolimoke LL vs. Ampampa) and between the two farmingtechniques for the two closely located sites at Ambolimoke.Similarly, univariate analysis was used in comparing between-site and between-period variations in physico-chemical fac-tors. Temperature data met the assumptions of parametricstatistics, and effects of site and month were analysed withgeneral linear model ANOVA, and between-site and between-month comparisons were conducted with the Tukey HSD test.

Thallus-level relationships between growth and health fac-tors (epiphyte and disease infection, sediment cover, intensityof fish and urchin grazing) were analysed with the non-parametric Spearman’s correlation analysis. Effects of healthand physico-chemical factors on growth were analysed usingthe non-parametric recursive partitioning regression (De'athand Fabricius 2000). Partition trees explain variation of asingle response variable by repetitively splitting the data untilhomogeneous groups are achieved. The procedure uses bothnumeric and categorical variables with hierarchical groupingsshowing effects of a single variable or interactive effects ofmore explanatory variables. It indicates cut-level points ofindependent variables, thereby suggesting threshold values.Despite some disadvantages (e.g. over-fitting of data), themethod has higher sensitivity and specificity, creating moreintuitive models (Strobl et al. 2009). The analysis was con-ducted at the site level by aggregating growth and health(measured at thallus level) and herbivore abundance (mea-sured at transect level) data in order to match physico-chemical factors (measured at site level). Only two tempera-ture variables, maximum (Tmax) and coefficient of variation(TCV), were included because average (Tavg), minimum (Tmin)and Tmax temperatures were strongly correlated with eachother. We also conducted Spearman’s correlation analysisbetween growth and physico-chemical variables and betweenimportant health and physico-chemical variables.

Since only two measurements were made for substrate/benthic composition, only during the cold and hot seasons(not at monthly intervals), relationships of substrate/benthoswith growth, health and physico-chemical variables were notanalysed.

Results

Patterns in the growth of K. alvarezii

Average growth of K. alvarezii in the region was 4.5±0.06 %day−1. Overall, there was a significant by-site difference(Kruskal-Wallis χ2=476.90, p<0.0001). Ampampa (longline) was the site with the highest growth (6.79±0.12 %day−1) while Ambolimoke OB had the lowest growth (2.69±0.10 % day−1), with the four remaining sites having inter-mediate growth (4.54 to 4.75 % day−1). There was also asignificant effect of period on growth (Kruskal-Wallis (K-W)χ2=135.51, p<0.0001) (Fig. 2); it was highest in May andlowest in January. Generally, it was higher during the coldseason (April–August: 5.12±0.08 % day−1) than that duringthe hot season (September–March: 4.12±0.08 % day−1) al-though it was exceptionally high in February (4.89±0.30 %day−1).

For the periods when all sites were monitored, a significantby-site difference was detected (K-W χ2=472.37, p<0.0001;

J Appl Phycol

Fig. 3). Generally, Ambolimoke LL and Lamboara site 3(off-bottom) were the least seasonal of the study sites.Ampampa had the highest growth in most months whileAmbolimoke OB had the lowest growth. Comparison be-tween the two closely located Ambolimoke sites showedthat except in November and December when no significantdifference was observed, Ambolimoke LL had a highergrowth in the remaining months. A significant differencewas also small in June (K-W χ2 = 4.07, p=0.04).Comparison between the two sites with long-line farmingshowed that Ampampa had higher growth than AmbolimokeLL in most months. There was no difference observed inOctober (K-W χ2=0.58, p=0.45). The three Lamboara siteshad similar growth patterns in most months (p>0.01).Lamboara site 2 had a lower growth in September andOctober (p<0.0001).

Relationship between growth of farmed K. alvareziiand health and physico-chemical factors

Growth negatively correlated with epiphyte cover, fish graz-ing, disease and sediment cover. All four health factors had apositive correlation with each other (Table 2). Growth showedno significant relationship with sea urchin grazing and pestseaweed cover.

Abundance of herbivorous fish and from belt transects,measured as a proxy to grazing intensity, had no significantrelationship with direct observation of thallus-level grazing(p>0.05). However, abundance of the three dominant urchinsin the area (Tripneustes gratilla: Spearman’s ρ=0.44, p=0.02;Echinometra mathaei: ρ=0.55, p=<0.0001; Echinothrixdiadema: ρ=0.50, p=0.0004) and total sea urchin abundance(ρ=0.44, p=0.002) significantly correlated with thallus-level

observation of grazing by urchins. This is probably associatedwith the high mobility of fish and low mobility of sea urchins.

Results of recursive partitioning analysis showed signifi-cant effects of sediment cover, maximum temperature (Tmax)and temperature variability (TCV) (Fig. 4) while the remainingthree physico-chemical variables (salinity, pH and oxygenlevel) were excluded by the model. Four homogeneous groupswere identified (Fig. 4) (R2=0.66). The main dichotomy rep-resented a separation between site-period combinations withsediment cover ≥48 and <48 %. Those with sediment cover

Re lativeg row

t hr ate

(%d-

1 )

4

5

6

a a

bb,c b,c b,c

cc c

dd

c,d

Month

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

thrate

(%d

0

1

2

3

Fig. 2 K. alvarezii. Monthly patterns in growth in south-westernMadagascar. Mean±SEM indicated. Letters different indicate signifi-cant differences

Fig. 3 K. alvarezii. By site comparison of monthly growth in south-western Madagascar. Mean±SEM indicated. Letters indicate significantdifferences

J Appl Phycol

≥48 % (group 1) had the lowest growth and were entirelyrepresented by sites with off-bottom farming in Ambolimokeand Lamboara. The clusters with sediment cover <48 % wasdivided based on Tmax. Those with Tmax≥27.91 °C formed adistinct cluster (group 2) and had the second-lowest growth.With the exception of Ampampa in April, October andNovember, this group was mostly composed of off-bottomsites in Ambolimoke and Lamboara during the hot summerseason. Group 3 represented sites with Tmax<27.91 °C andhigher temperature variability (TCV≥4.54 °C) and includedmostly sites in Lamboara during the cold season andAmbolimoke OB in July and October, with a relatively highgrowth rate. Group 4 (Tmax<27.91 °C and TCV<4.54 °C) wasmade entirely of Ampampa during the cold season andshowed the highest growth.

Growth negatively correlated with Tmax and TCV (Table 3).Of the four health variables that correlated significantly with

growth, epiphyte cover positively correlated with Tmax and TCV,while fish grazing had a significant correlation with TCV only(negative). Salinity, pH and oxygen level showed no correlationwith growth but pH had a positive correlation with Tmax, whileoxygen level showed a negative correlation with Tmax.

Spatio-temporal patterns in health factors

Sediment cover Sediment cover showed significant by-period(K-W χ2=457.80, p<0.0001) and by-site variations (K-Wχ2=106.69, p<0.0001; Fig. 5a). Generally, it was higher inMarch, May, August and November; it was lower in January,February and April. Overall, it was lower in Lamboara site 3(9.75±1.76) and Ampampa (5.20±1.59 %) than in the re-maining four sites. Comparison between the two sites inAmbolimoke showed that overall mean sediment cover wassimilar between the two sites (K-W χ2=0.18, p<0.67).However, there was a significant seasonal difference betweenthe two sites (p<0.05). It was higher in Ambolimoke LL thanin Ambolimoke OB in March and August whereas patternswere the opposite in May. Comparison between the two long-line farming sites showed that sedimentation was generallylower in Ampampa and in most months. Lamboara site 2 hadthe highest sediment cover among the off-bottom farmingsites while Lamboara site 3 had the lowest sedimentation(K-W χ2=88.37, p<0.0001). Both Lamboara site 1 and site2 had a higher sediment cover in most months.

Fish grazing There was significant by-period (K-W χ2=151.10, p<0.0001) and by-site (K-W χ2=428.21, p<0.0001)variation in fish grazing (Fig. 5b). Generally, grazing waslower during May–July and higher in March and December.It was highest at Lamboara sites 1 and 2 and Ambolimoke OBand lowest at Lamboara site 3. During months when all siteswere compared, between-site difference was highest in July(K-W χ2=115.31, p<0.0001) and September (K-W χ2=134.38, p<0.0001). In July, grazing was highest at Lamboarasite 2 and Ambolimoke OB. In September, it was highest atLamboara site 2, Ambolimoke OB and Ampampa. Monthlyvariation was highest at Ambolimoke OB (K-W χ2=236.69,

Fig. 4 K. alvarezii. Classification tree of homogeneous growth groupsidentified by recursive partitioning using growth as a dependent variableand health and physico-chemical factors as explanatory variables. Valueof environmental variable responsible for each classification node, meangrowth value±SD (% day−1) and number of samples of each cluster areindicated

Table 2 Correlation of growth and health factors

Growth %Epiphyte Fish grazing Urchin grazing Pest seaweed Disease

%Epiphyte −0.161***Fish grazing −0.248*** 0.148***

Sea urchin grazing −0.001 −0.029 -0.118***

Pest seaweed 0.048 −0.059* -0.127*** −0.013Disease −0.129 0.368*** 0.225*** −0.051 −0.049%Sediment −0.153 0.250*** 0.193*** −0.027 −0.029 0.131***

Spearman’s ρ (rho) and probability (p) are presented. Significant values are indicated in italics

*p=0.01–0.05; **p=0.001–0.01; ***p<0.001

J Appl Phycol

p<0.0001); grazing was higher in November–December andlower in March and October.

Comparison between the two Ambolimoke sites showed thatgrazing at Ambolimoke OB (1.73±0.06) was twice that ofAmbolimoke LL (0.90±0.07) (K-W χ2=92.86, p<0.0001).Ambolimoke OB had higher grazing in most months except inMarch when it showed lower grazing and in October when therewas no difference. Grazing was not observed in both sites inDecember. There was no significant difference between the twolong-line sites in overall mean grazing (K-W χ2=2.08, p=0.15).Grazing was significantly higher in Ambolimoke LL in June and

August and in Ampampa in September and October. Among thethree Lamboara sites, grazingwas highest in Lamboara site 2 andlowest in Lamboara site 3 in most months (p<0.05).

Epiphyte cover Epiphyte cover on K. alvarezii showed sig-nificant seasonal and by-site variations (Fig. 6a, p<0.0001).No epiphytes were observed in Ampampa and both sites inAmbolimoke at any period while all three Lamboara sitesshowed presence of epiphytes at some stage. Comparison ofthe three Lamboara sites showed a significant effect of site (K-W χ2=52.93, p<0.0001) and period (K-W χ2=440.48,p<0.0001). Overall epiphyte cover was higher in Lamboarasites 1 and 2 than in Lamboara site 3. It was highest in April

Fig. 6 Spatio-temporal patterns in a intensity of epiphyte infection and bdisease (ice-ice) infection on monitored thalli. Codes of site names andfarming technique as above

Fig. 5 Spatio-temporal patterns in a sedimentation and b intensity of fishgrazing on monitored thalli. Amb Ambolimoke, Amp Ampampa, LambLamboara, LL long-line farming technique, OB off-bottom farmingtechnique

Table 3 Correlation among growth, health and physico-chemical factors

Salinity pH Oxygen Growth Epiphyte Disease Fish-grazing Sediment

Tmax 0.052 0.44** −0.395* −0.583*** 0.453 0.284 0.258 0.22

TCV 0.284 −0.083 −0.202 −0.354* 0.439** −0.04 −0.358* 0.02

Salinity −0.078 −0.17 −0.196 0.242 0.153 −0.007 −0.17pH −0.178 0.061 −0.027 0.055 0.081 0.04

Oxygen 0.25 0.05 0.08 0.05 −0.1

Spearman’s ρ and probability (p) are indicated. Significant p values are in italics

*p=0.01–0.05; **p=0.001–0.01; ***p<0.001

J Appl Phycol

followed by September and November; no epiphytes wereobserved in May–July. Monthly variation was highest atLamboara site 2 (K-W χ2=243.10, p<0.0001) where epi-phyte cover was highest in April, August and September. Inthe remaining two Lamboara sites, it was highest in April,October and November. Lamboara site 3 had lower epiphytecover than Lamboara sites 1 and 2 in most months. Theepiphytes observed mainly belong to two genera of red fila-mentous algae: Neosiphonia and Herposiphonia.

Disease infection Ice-ice infection was observed at all sitesbut intensity was generally low, with Ambolimoke OB,Lamboara site 1 and site 2 having relatively higher levels ofinfection (Fig. 6b). Infection was highest in July atAmbolimoke OB, in March at Lamboara site 1 and inSeptember at Lamboara site 2. When disease was observed,it was consistently higher at Ambolimoke OB than atAmbolimoke LL (K-W χ2=55.83, p<0.0001). Overall, therewas no significant difference between the two long-line sites(K-W χ2=0.26, p=0.61). However, a significant differencewas found between the two sites in May when infection washigher in Ampampa than in Ambolimoke LL (K-W χ2=57.0,p<0.0001). Among the three Lamboara sites, it was highest inLamboara site 2 and lowest in Lamboara site 3 in most months(p<0.05). Infection was higher in site 1 than in site 2 and site 3in March (K-W χ2=77.03, p<0.0001).

Other health factors Sea urchin grazing was observed only atAmpampa (0.28±0.08) and Lamboara site 2 (0.10±0.08).However, seasonal variation was observed only at the latter

site (K-W χ2=27.40, p=0.001) with growth being higher inJuly (0.68±0.11) than in the remaining months (0 to 0.18).Pest seaweed cover was found only at Ambolimoke LL (0.08±0.02) and Lamboara site 3 (0.09±0.02). It was higher inMayat Ambolimoke LL (0.64±0.08) and in August at Lamboarasite 3 (0.64±0.08). Dictyota cervicornis, Hydroclathrusclathratus, Cystoseira myrica and Ulva reticulata were thedominant pest seaweed species.

Spatio-temporal patterns in seawater temperature

Of the physico-chemical factors considered, only seawatertemperature properties had a significant direct or interactiveeffect on growth (see below). There was a clear and strongseasonal pattern in average (ANOVA F=290.30, p<0.0001),minimum (F=229.22, p<0.0001) andmaximum temperatures(F=184.10, p<0.0001; Fig. 7). Values were highest inFebruary and lowest in July with sharp decreases observedbetween March and June and increases between July andOctober. In contrast, seasonal variation in temperature vari-ability was less pronounced (F=16.09, p<0.0001; Fig. 7d).Generally, January was the month with the lowest variabilitywhile November and December had the highest variability.

However, temperature variation (TCV) showed the highestbetween-site difference of the four water temperature proper-ties considered (F=70.35, p<0.0001; Fig. 7d). It was highestat Lamboara site 3 and lowest at Ambolimoke and Ampampahad the lowest TCV. There was no significant between-sitedifference in average temperature (F=1.61, p=0.17) whilethere was weak difference in Tmin (F=8.76, p<0.0001) and

Fig. 7 Spatio-temporal patternsin water temperature properties atfive study sites. Mean±SEM

J Appl Phycol

Tmax (F=5.94, p<0.0001). Tmin was highest at Ambolimokeand Ampampa and lowest at Lamboara site 3 whereas Tmax

was highest at Lamboara site 3 and lowest at Ampampa. Nosignificant interaction was observed between the effects of siteand month for all four water temperature properties (p>0.97).

Comparison of substrate and benthos

Comparison of substrate/benthos showed significant between-site differences for all substrate/benthic categories (Table 4).There was no difference between the cold and hot seasons.Most sites were located in sandy areas except Lamboara site 2,whichwas mostly rocky (fossil reef) (Table 4).Most sites had avery low cover of fleshy macroalgae and turf algae; Lamboarasite 2 and Ampampa had a relatively higher cover of the formerwhile Ampampa had a higher cover of the latter. AmbolimokeOB and Lamboara site 1 had a higher cover of the seagrassThalassia hemprichii. Halophila minor was observed inAmbolimoke LL during both seasons, but it was not recordedin our quadrats due to its sparse distribution in the area.

Discussion

Growth of farmed K. alvarezii in south-western Madagascarshowed a significant variation between sites and seasons.These differences were mainly associated with variations inenvironmental and health factors. Physico-chemical andhealth factors interact in complex ways in influencingspatio-temporal pattern in growth. Growth was higher duringthe cold season in most sites, and water temperature, sedimentand epiphyte cover and fish grazing were the main factorsinfluencing growth.

Effect of physico-chemical and health factors

Water temperature The study shows that Tmax and TCV areimportant environmental parameters influencing growth di-rectly (and probably indirectly through their influence onhealth factors such as epiphyte infection). The negative corre-lation between Tmax and oxygen content also suggests thepresence of stressful environmental conditions of high tem-perature and low oxygen content during the hot season. Thefact that growth is lower at high TCV also suggests that thisparameter could be an important variable influencing growthand health independent of Tmax. The lack of a significantcorrelation between Tmax and TCV indicates that their effecton growth is interactive. At high Tmax, TCV has no significanteffect, suggesting that stressful effects of TCV are felt at Tmax

below 27.9 °C. Thus, Tmax is the main stressor between thetwo water temperature properties while TCV plays a secondaryrole and operates more when Tmax is less stressful.Physiological and ecological effects of extremes in meanand maximum temperatures on growth and health ofK. alvarezii are widely covered in the phycological literature(Paula and Pereira 2003; Gerung and Ohno 1997; Largo et al.1995). In contrast, effects of temperature variability on sea-weed growth, rate and intensity of infection by epiphytes, ice-ice and other diseases are rarely considered. As far as we areconcerned, this is the first study to indicate negative effects oftemperature variation on the growth of farmed K. alvarezii.However, whether its effects are direct or indirect by influenc-ing health factors, such as epiphyte and disease, or both andthe mechanisms are not clear.

Sediment cover Partitioning regression identified sedimentas one of the most important factors influencing growth ofK. alvarezii in the study area. The study supports otherfindings showing strong decreases in growth associatedwith sedimentation (Paula and Pereira 2003). Thalli with

Table 4 Comparison of substrate/benthic cover in the study area

Site Sand/silt Rock/fossil reef Fleshy macroalgae Turf algae Seagrass

Mean SEM Mean SEM Mean SEM Mean SEM Mean SEM

Ambolimoke-LL 100.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Ambolimoke-OB 92.7 1.9 0.0 0.0 0.5 0.5 0.0 0.0 6.8 1.9

Ampampa-LL 90.3 2.9 0.0 0.0 5.6 2.4 2.6 1.2 1.5 0.8

Lamboara 1-OB 93.0 1.3 0.0 0.0 0.5 0.5 0.5 0.5 6.0 1.3

Lamboara 2-OB 15.5 3.0 80.0 3.3 4.0 1.6 0.5 0.5 0.0 0.0

Lamboara 3-OB 100.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Statistical comparison K-W χ2 p K-W χ2 p K-W χ2 p K-W χ2 p K-W χ2 p

By site 35.57 <0.0001 48.26 <0.0001 13.11 0.01 9.89 0.04 29.57 <0.0001

By month 0.54 0.46 0.005 0.94 1.40 0.24 0.83 0.36 1.18 0.28

Kruskal-Wallis χ2 and probability (p) are indicated

J Appl Phycol

low growth at Ambolimoke and Lamboara also had thehighest sediment cover. High sedimentation levels at thesesites could influence the availability of light for photo-syn thes i s and norma l phys io logy and hea l th .Accompanied by fish grazing, high sedimentation rate atAmbolimoke OB may have resulted in the lowest growthin the study area.

Fish grazing Fish grazing creates a significant variation ingrowth with site/period combinations of low grazing valueshaving one of the highest growth rates. Similar to infection,grazing was also excluded from the partitioning regressionmodel, probably due to a loss of degrees of freedom associatedwith reduction of data at site-level analysis. However, fishgrazing had a significant negative correlation with TCV, sug-gesting that grazing is higher at more thermally stable sitesand could result in significant reductions in production(Luxton 2003).

In comparison to other sites, the seasonal growth pattern atAmbolimoke OB was reversed; it was higher during the warmseason while seasonality was weak at Ambolimoke LL. Thereversed seasonality at Ambolimoke OB is possibly associat-ed with sediment cover and fish grazing that were higherduring the cold season. The difference in growth betweenthe two Ambolimoke sites could be attributed to reductionof grazing at Ambolimoke LL associated with farmingtechnique.

Epiphyte and disease infection Growth had a negative corre-lation with epiphyte cover with sites near Lamboara withhigher epiphyte cover having one of the lowest growths.Seasonality in epiphyte infection has been observed in otherregions and is often linked with changes in salinity or hightemperatures (Hurtado et al. 2006; Vairappan 2006; Hayashiet al. 2010). Variations in epiphyte infection observed in thisstudy are most likely related to seasonal changes and between-site differences in water temperature. There was little to noseasonal or by-site variation in salinity and pH observed in thestudy. The study area in SW Madagascar is very arid andprobably maintains constant salinity and pH throughout theyear.

The positive correlation of epiphyte with Tmax and TCVsuggests interactive effects of the two temperature propertieson epiphyte infection. For example, despite the similarity inTmax among the three Lamboara sites and Ampampa, the lattersite had the most stable temperature and no epiphytes wereobserved. However, epiphyte cover was excluded by thepartitioning model. The latter was sampled at the site levelwhile health parameters at the thallus level. The loss of detec-tion of effect of epiphytes could be due to the reduction of dataduring aggregation for site-level analysis.

Sites with the highest infection of epiphytes had eitherthe highest cover of fleshy algae (Lamboara site 2) or

seagrass (Lamboara site 1) or had no major benthic cover(Lamboara site 3). The main substrate type at Lamboara site1 and site 3 is sand while Lamboara site 2 is dominated byfossil rocky reef. Epiphyte infection is possibly site specificas the above three sites lie at close proximity despite thedifferences in their benthic and substrate types. Ampampa,the closest site with no epiphyte infection, is isolated by achannel from the three Lamboara sites. The distance ofseaweed lines to the bottom associated with farming tech-nique could not explain the difference. For instance, despitethe similar farming technique used and similar cover ofseagrass as in Lamboara site 1, no epiphytes were detectedat Ambolimoke OB. Other unspecified depth-related fac-tors could also be a reason for the difference as the twoAmbolimoke sites and Ampampa were deeper than thethree Lamboara sites. By comparing Zanzibar sites of highand low seagrass cover (Mtolera 2003) concluded that highseagrass cover could induce higher seaweed productivity.The low productivity at the two sites with high seagrasscover in this study suggests that other factors could alsoplay important roles in influencing health and growth ofK. alvarezii.

Epiphyte and disease (ice-ice) infections are becomingincreasingly recognised as a threat to seaweed farming glob-ally (Hurtado et al. 2006; Vairappan et al. 2008) and one of themain reasons behind the decline of seaweed farming in nearbyZanzibar (Bryceson 2002; Msuya 2006). In this study, ice-iceinfection was too low to cause a noticeable reduction ingrowth. However, it is not clear if disease prevalence wouldincrease as farming becomes intensified to meet projectedeconomic targets as high density farms often become moreprone to disease (Vairappan et al. 2008).

Comparison with other regions

The average growth observed in this study falls within themid-high range of values observed globally (Hayashi et al.2010). An exception is Ambolimoke OB where growth iswithin the lower range of the global values. Ampampa, thesite with the highest growth, has growth values comparable tothose in the high end of global figures. Compared to Zanzibarwhere seaweed farming has been practiced since the late1980s, the lowest growth value in this study is higher thanthe lowest value recorded there while the highest values aremore or less similar (0.5 to 7.6 % day−1) (Msuya et al. 2007).The maximum value from Zanzibar is higher for nutrient-fertilised thalli (fertilised, 0.5 to 5.6; unfertilised, 0.5 to7.6 % day−1). However, older reports from Zanzibar indicatehigher growth values of 4.4. to 8.9 % day−1 (Semesi andDawes 1984). Seaweed farming in Zanzibar has seen signif-icant declines in recent years due to increased temperaturesand associated increases in epiphyte and disease infection(Rice et al. 2006; Msuya et al. 2007).

J Appl Phycol

Conclusions and recommendations

The study shows the presence of a clear seasonal and by-sitevariation in growth of farmed K. alvarezii in south-westernMadagascar. Growth is higher in the cold season in most sitesand is mainly limited by sedimentation, higher values intemperature maxima and variability, fish grazing and epiphyteinfection. The presence of strong seasonal variation in growthand intensity of epiphyte infestation in particular highlightsthe need for careful selection of appropriate farming sites andseedlings.

Seaweed farming is low input and less challenging techni-cally than finfish and shellfish farming, and the success of thecommunity-based seaweed farming project will mainly de-pend on site/period selection and maintaining best growthconditions by managing the ecological constraints observedin the study. K. alvarezii is a crop with limited genetic vari-ability as cultivated thalli are propagated only vegetatively.This is particularly true for the Western Indian Ocean (WIO)where cultivated thalli originate from a single clone intro-duced to the region in the 1980s. To avoid the main problemcurrently faced by the community (e.g. reduced growth dueto epiphyte infection associated with higher temperature inthe hot season), best management practices need to beintroduced quickly. Farming should focus on the cold sea-son when productivity is highest, and on more environmen-tally stable sites, since both higher temperature maxima andvariability seem to interact in causing a stressful environ-ment for K. alvarezii. Farming should also focus on longline as thalli on off-bottom lines that are close to the bottomseem to face higher epiphyte infection, grazing and sedi-ment cover. In addition, farming in cooler and well-mixeddeeper and offshore areas should be encouraged althoughits profitability and social effects need to be assessed first asthey could be more expensive and have negative socialimplications through the exclusion of women.

Acknowledgments We are grateful for the assistance in data collectionby the Blue Ventures aquaculture field staff in Madagascar: Josvah,Bevick, Mamonjy and Haja; to many of the farmers supported by theproject, especially Rahlesa and Rambely; and to staff of the InstitutHalieutique et des Sciences Marines, IHSM, University of Toliara. Thestudy was supported by NORAD through Norges Vel.

References

Ask EI (2003) Creating a sustainable commercial Eucheuma cultivationindustry: the importance and necessity of the human factor. Proc IntSeaweed Symp 17:13–18

Ask EI, Azanza RV (2002) Advances in cultivation technology of com-mercial eucheumatoid species: a review with suggestions for futureresearch. Aquaculture 206:257–277

Ask EI, Batibasaga A, Zertuche-Gonzalez JA, De San M (2003a) Threedecades of Kappaphycus alvarezii (Rhodophyta) introduction tonon-endemic locations. Proc Int Seaweed Symp 17:49–57

Ask EI, Azanza R, Simbik M, Cay-An R, Lagahid J (2003b)Technological improvements in commercial Eucheuma cultivation(a short communication). Sci Diliman 15:47–51

Barnes-Mauthe M, Oleson KLL (2013) The total economic values ofsmall-scale fisheries with a characterization of post-landing trends:an application in Madagascar with global relevance. Fish Res 147:175–185

BinduMS, Levine IA (2011) The commercial red seaweed Kappaphycusalvarezii—an overview on farming and environment. J Appl Phycol23:1–8

Bixler HJ (1996) Recent developments in manufacturing and marketingcarrageenan. Hydrobiologia 326/327:35–37

Bixler HJ, Porse H (2010) A decade of change in the seaweed hydrocol-loids industry. J Appl Phycol 23:321–335

Bryceson I (2002) Coastal aquaculture developments in Tanzania: sus-tainable and non-sustainable experiences. West Indian Ocean J MarSci 1:1–19

Dawes CJ, Trono GC Jr, Lluisma AO (1993) Clonal propagation ofEucheuma denticulatum and Kappaphycus alvarezii for Philippineseaweed farms. Hydrobiologia 260/261:379–383

De'ath G, Fabricius KE (2000) Classification and regression trees: apowerful yet simple technique for ecological data analysis.Ecology 81:3178–3192

Gerung GS, Ohno M (1997) Growth rates of Eucheuma denticulatum(Burman) Collins et Harvey and Kappaphycus striatum (Schmitz)Doty under different conditions in warmwaters of Southern Japan. JAppl Phycol 9:413–415

Hayashi L, Hurtado AQ, Msuya FE, Bleicher-Lhonneur G, Critchley AT(2010) A review of Kappaphycus farming: prospects and con-straints. In: Seckbach J, Einav R, Israel A (eds) Seaweeds and theirrole in globally changing environments. Springer, Dordrecht,pp 251–283

Hill NAO, Rowcliffe OM, Koldewey H, Milner-Gulland EJ (2012) Theinteraction between seaweed farming as an alternative occupationand fisher numbers in the Central Philippines. Conserv Biol26:324–334

Horsfall JG, Barratt RW (1945) An improved grading system for mea-suring plant diseases. Phytopathology 35:655

Hurtado AQ, Critchley AT, Trespoey A, Lhonneur GB (2006) Occurrenceof Polysiphonia epiphytes in Kappaphycus farms at Calaguas Is.,Camarines Norte, Phillippines. J Appl Phycol 18:301–306

James WC (1971) An illustrated series of assessment keys for plantdiseases, their preparation and usage. Can Plant Dis Surv 51:39–65

Largo DB, Fukami K, Nishijima T, Ohno M (1995) Laboratory-induceddevelopment of the ice-ice disease of the farmed red algaeKappaphycus alvarezii and Eucheuma denticulatum (Solieriaceae,Gigartinales, Rhodophyta). J Appl Phycol 7:539–543

Luxton DM (2003) Kappaphycus agronomy in the Pacific Islands. ProcInt Seaweed Symp 17:41–47

Msuya FE (2006) Seaweed farming as a potential cluster. Proceedings ofinnovation systems and clusters programme in Tanzania (ISCP-Tz):cluster initiative launching workshop, Dar es Salaam, Tanzania, pp102-113

Msuya FE, Porter M (2014) Impact of environmental changes on farmedseaweed and farmers: the case of Songo Songo Island, Tanzania. JAppl Phycol. doi:10.1007/310811-014-0243-4:1-7

Msuya FE, Shalli MS, Sullivan K, Crawford B, Tobey J, Mmochi AJ(2007) A comparative economic analysis of two seaweed farmingmethods in Tanzania. The sustainable coastal communities andecosystems program. Coastal Resources Center, University ofRhode Island and the Western Indian Ocean Marine ScienceAssociation, 27pp

Msuya FE, Buriyo A, Omar I, Pascal B, Narrain K, Ravina JJM, MrabuE, Wakibia JG (2014) Cultivation and utilisation of red seaweeds inthe Western Indian Ocean (WIO) Region. J Appl Phycol 26:699–705

J Appl Phycol

Mtolera MSP (2003) Effect of seagrass cover and mineral content onKappaphycus and Eucheuma productivity in Zanzibar. West IndianOcean J Mar Sci 2:163–170

Munoz J, Freile-Pelegrin Y, Robledo D (2004) Mariculture ofKappaphycus alvarezii (Rhodophyta, Solieriaceae) colorstrains in tropical waters of Yucatan, Mexico. Aquaculture239:161–177

Paula EJ, Pereira RTL (2003) Factors affecting the growth rate ofKappaphycus alvarezii (Doty) Doty ex P. Silva (Rhodophyta,Solieriaceae) in sub-tropical waters of São Paulo State, Brazil.Proc In Seaweed Symp 17:381–388

Pickering T (2006) Advances in seaweed aquaculture among PacificIsland countries. J Appl Phycol 18:87–93

Rice MK, Mmochi AJ, Zuiberi L, Savoie RM (2006) Aquaculture inTanzania. World Aquacult 37:50–57

Semesi AK, Dawes CJ (1984) Carbohydrate, lipid and ash content ofselected seaweeds from Tanzania. In: Middleton J, Nikundiwe AM,Banyikwa FF, Mainoya, JR (eds) Proceedings of the symposium onthe role of biology in development, Dar es Salaam, September 1983,University of Dar es Salaam, p 85

Strobl C, Malley J, Tutz G (2009) An introduction to recursivepartitioning: rationale, application and characteristics of classifica-tion and regression trees, bagging and random forests. PsycholMethods 14:323–348

Vairappan CS (2006) Seasonal occurrences of epiphytic algae on the com-mercially cultivated red alga Kappaphycus alvarezii (Solieriaceae,Gigartinales, Rhodophyta). J Appl Phycol 18:611–617

Vairappan CS, Chung CS, Hurtado AQ, Soya FE, L’honneur GB, CritchleyA (2008) Distribution and symptoms of epiphyte infection in majorcarrageenophyte-producing farms. J Appl Phycol 20:477–483

J Appl Phycol