UNIVERSIDADE TÉCNICA DE LISBOA INSTITUTO SUPERIOR DE AGRONOMIA

ESTRUTURA DOS SISTEMAS RADICAIS E DINÂMICA DA ÁGUA NO SOLO NUMA COMUNIDADE ARBUSTIVA DA

TAPADA NACIONAL DE MAFRA

JOAQUIM MANUEL SANDE SILVA

ORIENTADOR: Doutor Francisco Manuel Cardoso de Castro Rego

CO-ORIENTADOR: Doutor Stefano Mazzoleni

JURI: Presidente Reitor da Universidade Técnica de Lisboa.

Vogais Doutor Stefano Mazzoleni, full professor da Facoltà di Agraria da Università di Napoli Frederico II, Itália;

Doutora Maria Amélia Botelho de Paulo Martins Campos Loução, professora catedrática da Faculdade de Ciências da Universidade de Lisboa;

Doutor Manuel Armando Valeriano , professor catedrático do Instituto Superior de Agronomia da Universidade Técnica de Lisboa;

Doutor Francisco Manuel Cardoso de Castro Rego, professor associado do Instituto Superior de Agronomia da Universidade Técnica de Lisboa;

Doutor António Manuel Dorotêa Fabião, professor associado do Instituto Superior de Agronomia da Universidade Técnica de Lisboa;

Doutor José Miguel Oliveira Cardoso Pereira, professor associado convidado do Instituto Superior de Agronomia da Universidade Técnica de Lisboa.

DOUTORAMENTO EM ENGENHARIA FLORESTAL

LISBOA

2002 ii

AGRADECIMENTOS

Muito embora uma tese de doutoramento tenha apenas um autor, ela é normalmente o resultado de inúmeras colaborações. A tese que agora se apresenta não é, a esse respeito, uma excepção. Por isso esta secção de agradecimentos, muito mais que uma mera formalidade ou um dever moral, é sobretudo o resultado de uma vontade sincera de manifestar a minha gratidão a todas as pessoas que contribuíram para tornar possível este trabalho. O meu muito obrigado a todas elas! Em particular impõe-se um agradecimento especial: - Ao meu orientador Professor Francisco Castro Rego, companheiro constante deste e de outros desafios passados e esperemos que futuros também. - Ao meu co-orientador Professor Stefano Mazzoleni, pelo acompanhamento e conselhos prestados, apesar da distância a separar-nos. - À Professora Mariana Amato, pela co-orientação na fase incial desta tese. - À Professora Amélia Martins-Loução, pela colaboração prestada nos capítulos 2 e 3 desta tese e pelo interesse geral demonstrado. - Ao Francesco Giannino pela colaboração no capítulo 6 e pelo apoio logístico em Nápoles. - Ao Engenheiro Ricardo Paiva, pela autorização em realizar este estudo na Tapada Nacional de Mafra e ainda pelo apoio logístico prestado. - Aos funcionários da Tapada Nacional de Mafra: Sr. José Domingos, guarda Luis Filipe, guarda João Louro e a todos os outros não referidos aqui, pelo apoio prestado no trabalho de campo. - À Doutora Paula Gonçalves, pela colaboração relativamente aos dados da Serra da Malcata. - Aos alunos da ESAC: Ana Salomé David, Cristina Monteiro, Nuno Lobo e Inês Lopes, pela colaboração em várias fases do trabalho de campo. - Ao Victor Abadia, pela colaboração no trabalho de campo. - Aos estagiários do programa Sócrates: Ioana Chitu e Costel Petku, pela colaboração no trabalho de campo e de laboratório. - Ao técnico da DRABL Sérgio Correia, pelo precioso apoio no planeamento e na realização do fogo experimental. - Ao colega da UTAD Paulo Fernandes, pela bibliografia disponibilizada. - Aos colegas da ESAC: Carmo Magalhães, David Rodrigues, Hélia Marchante, José Maia, Manuel Nunes, Óscar Crispim e a outros colegas e funcionários não referidos, pela colaboração prestada em diversas fases deste trabalho. - À instituição a que pertenço, a Escola Superior Agrária de Coimbra, pela dispensa de serviço docente no âmbito de uma bolsa PRODEP durante a metade final da realização deste trabalho.

ii iii

RESUMO

São abordados diferentes aspectos da estrutura das raízes e da dinâmica da água no solo numa comunidade arbustiva da Tapada Nacional de Mafra. Numa amostra de plantas escavadas incluindo várias espécies, é realizada uma análise de diferentes variáveis estruturais revelando uma distribuição das plantas de acordo com grupos funcionais e estádios de desenvolvimento. São apresentadas relações alométricas consistentes entre a secção basal das plantas e as biomassas da raiz e da parte aérea. É desenvolvido e testado com sucesso um novo modelo de distribuição vertical de raízes. Este modelo é utilizado para caracterizar a distribuição das raízes ao nível do indivíduo e ao nível da comunidade. São estimadas a profundidade máxima de enraizamento, a distribuição vertical e os valores absolutos de biomassa e comprimento de raízes finas da comunidade arbustiva. É estudado o efeito de curto prazo do fogo na água do solo, sendo observado um aumento do teor de humidade na parcela queimada relativamente à parcela testemunha. É apresentado e testado com sucesso um novo modelo de simulação da dinâmica da água no solo, utilizando para tal medições de humidade no solo realizadas a diferentes profundidades ao longo de 18 meses.

Palavras chave: Vegetação mediterrânica, sistemas radicais, distribuição das raízes, efeitos do fogo, dinâmica da água no solo, modelação.

iii iv

ABSTRACT

(Root system structure and soil water dynamics in a shrub community at Tapada Nacional de Mafra) The present thesis approaches different aspects of the root structure and the soil water dynamics of a shrub community at Tapada Nacional de Mafra. Using a sample of excavated plants including different species, an analysis of different structural variables is performed, revealing a distribution of plants according to functional groups and developmental stages. Consistent allometric relationships between basal section and root and shoot biomass are presented. A new root distribution model is developed and successfully tested. This model is used to characterise root distributions at the individual and at the community level. The maximum rooting depth, the vertical distribution and the absolute values of fine root biomass and length are estimated at the community level. The short-term effect of fire on soil water dynamics is studied, revealing an increase on soil moisture at the burned plot when compared to the control plot. A new soil water dynamics model is presented and successfully tested, for which a set of soil water measurements taken at different depths during an 18-month period, is used.

Keywords: Mediterranean vegetation, root systems, root distribution, fire effects, soil water dynamics, modelling.

iv v

ÍNDICE

AGRADECIMENTOS...... ii RESUMO ...... iii ABSTRACT ...... iv ÍNDICE ...... v LISTA DE FIGURAS ...... vii LISTA DE TABELAS ...... xi 1 INTRODUÇÃO...... 1 1.1 As comunidades arbustivas das regiões mediterrânicas...... 1 1.2 O papel dos sistemas radicais nas estratégias evolutivas das plantas mediterrânicas ...... 3 1.3 O estudo das raízes das plantas mediterrânicas...... 7 1.4 Os efeitos do fogo na dinâmica da água do solo ...... 10 1.5 A modelação da dinâmica da água no solo ...... 12 1.6 Justificação, objectivos e estrutura da presente tese...... 16 Bibliografia...... 20 2 BELOWGROUND TRAITS OF MEDITERRANEAN WOODY PLANTS IN A PORTUGUESE SHRUBLAND...... 27 2.1 Introduction ...... 28 2.2 Methods...... 29 2.3 Results ...... 32 2.4 Discussion ...... 38 References ...... 40 3 ROOT DISTRIBUTION OF MEDITERRANEAN WOODY PLANTS; INTRODUCING A NEW EMPIRICAL MODEL ...... 43 3.1 Introduction ...... 44 3.2 Methods...... 45 3.3 Results ...... 49 3.4 Discussion ...... 54 References ...... 57

v vi

4 ROOT DISTRIBUTION OF A MEDITERRANEAN SHRUBLAND IN ...... 60 4.1 Introduction ...... 61 4.2 Methods...... 62 4.3 Results ...... 67 4.4 Discussion ...... 74 References ...... 77 5 FIRE EFFECTS ON SOIL WATER DYNAMICS IN A MEDITERRANEAN SHRUBLAND...... 81 3.1 Introduction ...... 82 3.2 Methods...... 83 3.3 Results ...... 85 3.4 Discussion ...... 92 References ...... 94 6 MODELLING SOIL WATER DYNAMICS IN A MEDITERRANEAN SHRUBLAND...... 96 6.1 Introduction ...... 97 6.2 Methods...... 98 6.3 Results ...... 105 6.4 Discussion ...... 111 References ...... 114 7 DISCUSSÃO E CONCLUSÕES ...... 117 7.1 Estudo das raízes de plantas individuais ...... 117 7.2 Estudo das raízes ao nível da comunidade de plantas...... 124 7.3 Estudo da dinâmica da água no solo ...... 128 Bibliografia...... 134

vi vii

LISTA DE FIGURAS

Fig. 1.1 Aspecto geral do matagal de urze (Erica scoparia e Erica lusitanica) na Tapada Nacional de Mafra, onde decorreram os trabalhos relatados nos capítulos 4, 5 e 6 e onde foram escavadas algumas das plantas estudadas nos capítulos 2 e 3...... 18 Fig. 2.1 PCA diagrams. A represents the components loadings for each variable and B represents the components scores for each plant individual. Legend for variables: DIAMETER – Average root diameter; BIOMASS – Root biomass; DEPTH – Maximum rooting depth; WIDTH – Root system width; LENGTH – Root length; R/S – Root-to-shoot ratio; SRL – Specific root length. Legend for species: Cc – Cistus crispus; Cs – Cistus salvifolius; Cm – Crataegus monogyna; Dg – Daphne gnidium; El – Erica lusitanica; Es – Erica scoparia; Ll – Lavandula luisieri; Mc – Myrtus communis; Pl – Pistacia lentiscus; Ru – Rubus ulmifolius; Uj – Ulex jussiaei. Symbols in bold correspond to obligate seeders. The developmental stage, as obtained by the respective basal section, is indicated by the number following the species symbol. Stage 1: 0 to 5 mm2; stage 2: 5 to 25 mm2; stage 3: 25 to 125 mm2; stage 4: > 125 mm2...... 34 Fig. 2.2 Relationships between Basal Section and two root system indices: Specific Root Length (SRL) and Root-to-Shoot ratio (Root-Shoot), for two obligate seeders ( Cistus crispus and Lavandula luisieri; represented in bold) and two resprouters (Daphne gnidium and Crataegus monogyna; normal lettering)...... 36 Fig. 2.3 Images of 16 plants representing different developmental stages (as defined in Figure 1) of two resprouters: A – Daphne gnidium (stages 1,2,3 and 4), B – Crataegus monogyna (stages 1,2,3 and 4); and two obligate seeders: C – Lavandula luisieri (stages 1,2,3 and 4), D – Cistus crispus (stages 1,2,2 and 3). The vertical bars represent 0.5 m. Arrows indicate the ground surface ...... 37 Fig. 3.1 Root distribution of nine plants excavated at Tapada Nacional de Mafra as represented by the fitted MLDR model (see text and equation 4). D and c are the model parameters. The solid line represents the cumulative root biomass distribution and the dotted line represents the cumulative root length distribution...... 52

vii viii

Fig. 3.2 Root systems of nine plants excavated at Tapada Nacional de Mafra. The corresponding cumulative root distributions are shown on Fig. 3.1. The vertical bar represents 0.5 m...... 53 Fig. 4.1 Relationship between root diameter and the vertical distance to the root tip for Erica ...... 66 Fig. 4.2 Average density of root counts for each species including all diameters. (mean±SE, n=6). Insets represent the MLDR model fitted to the cumulative root fraction.

Horizontal lines and corresponding depths indicate the value of D50...... 70 Fig. 4.3 Average density of root counts for each diameter, including all species (mean±SE, n=6). Insets represent the MLDR model fitted to the cumulative root fraction.

Horizontal lines and corresponding depths indicate the value of D50...... 71 Fig. 4.4 Schematic representation of the estimated average maximum rooting depth (mean of the deepest 28 roots) of Erica plants at each trench. The solid line represents the soil surface. The broken straight line represents the maximum depth of excavation (bottom of the trenches). Both the above and the belowground parts of each plant have been drawn to scale in order to represent the average height and the average maximum rooting depth respectively. Each trench is represented by an image obtained from an Erica scoparia individual...... 72 Fig. 4.5 Distribution of the estimated maximum rooting depths of Erica (mean±SE; n=2, n=3 and n=4 for the first, second and third depth classes, respectively; n=6 for the remaining classes)...... 72 Fig. 4.6 Biomass and length of fine roots per unit of soil volume (mean±SE, n=10) as obtained from core samples. Gaps at 50 cm, 70 cm and 90 cm on the y axis correspond to non-sampled depths. Insets represent the MLDR model fitted to the cumulative root fraction. Horizontal lines and corresponding depths indicate the

value of D50...... 73 Fig. 5.1 Evolution of plant cover during the treatment period. Triangles and diamonds represent Erica and P. aquilinum respectively. Open symbols/dashed lines refer to plant height whereas closed symbols/solid lines are the density of P. aquilinum fronds or resprouting of Erica plants...... 86 Fig. 5.2 Root density (number of root counts/dm2) at the control and the burned plots (mean±SE). Data collected before fire...... 87 viii ix

Fig. 5.3 Values of soil moisture for each plot in the two study periods with the corresponding meteorological data (bars and solid line represent rainfall and average daily temperature, respectively)...... 87 Fig. 5.4 Average soil moisture of the control and the burned plots during the treatment period and two weeks before fire...... 90 Fig. 5.5 Differences in soil water storage between the treatment and the reference periods in

the burned (Dbi) and the control (Dci) plots. Si = Dbi - Dci represents the effect of

fire on soil water. Si is represented by a moving average (n=7). Estimation for 0-180 cm is an integration using 20 cm increments...... 91 Fig. 6.1 Schematic diagram of model SWADY drawn using the modelling environment Simile. The compartment SWC represents the Soil Water Content, circles with a cross represent variables, thick arrows represent flows to and from the compartment and the curved thin arrows represent influences between the different model entities...... 100 Fig. 6.2 Comparison between soil water content values obtained from laboratory- determined water retention relationships (measured) and the corresponding values obtained using the soil water retention sub-model (modelled) based in the method proposed by Wösten et al., (1999) for computing the parameters of the Mualem- Van-Genuchten equation. Soil water content values correspond to three pressure head levels applied to samples from six different depths, from both plots. The solid line represents the linear regression (n=36) and the broken line represents a reference y=x relationship...... 108 Fig. 6.3 Water content at six different depths at the control plot. Circles represent actual measurements and the continuous lines represent model simulations. The histogram shows the rainfall during the two years (separated by a broken line) of measurements. Horizontal broken lines represent saturation, field capacity and wilting point, respectively, as obtained by the model...... 109 Fig. 6.4 Water content at six different depths at the burned plot. Circles represent actual measurements and the continuous lines represent model simulations. The histogram shows the rainfall during the two years (separated by a broken line) of measurements. Horizontal broken lines represent saturation, field capacity and

ix x

wilting point, respectively, as obtained by the model. The arrow indicates the date of fire...... 110 Fig. 6.5 Modelled (line) and measured (circles) net effect of fire in terms of soil water

storage (Si). Si=Dbi- Dci, where Dbi represents the soil water storage difference for each day (i) between the treatment period and the reference period in the burned

plot and (Dci) represents the same difference for the control plot. The estimation refers to the layer 0-180 cm and results from an integration using 20 cm increments. Fire occurred at the beginning of the time series (Julian day 155, 4th of June)..111

x xi

LISTA DE TABELAS

Table 2.1 Descriptive parameters (mean ± SE) of the root systems of 33 plants excavated at Tapada Nacional de Mafra, distributed by species. Legend for abbreviations: n – number of plants; Reg. strat. – regenerative strategy; Maxim. root. depth - maximum rooting depth; Aver. root diam. – average root diameter; R/S – root-to- shoot ratio; SRL – specific root length; s – obligate seeder; r – resprouter...... 33 Table 2.2 Allometric relationships obtained by linear regression between basal section (mm2) and four different root variables: root biomass, shoot biomass, root system length and root system width. All variables were log-transformed. Biomass data is indicated in decigrams in order to obtain only positive values. For each linear regression it is indicated the intercept (a), the slope (b), the coefficient of determination (r2) and the associated probability (p)...... 35 Table 3.1 Comparison of average ranks (1 to 4) and mean r2 of four models fitted to root biomass and root length data, as obtained by the Friedman Anova (p<0.001 for both root biomass and root length distributions). Values of mean r2 followed by the same letter did not present significant differences (p>0.05) as obtained by paired comparisons using the Mann-Whitney U test...... 49 Table 3.2 Maximum rooting depth averaged by species and basal section class. Maximum rooting depth represents the depth achieved by the deepest root. Values are averages and the number of plants is shown in brackets. The range was obtained as the difference between the maximum and the minimum values observed. Basal section classes are defined as: BS1 ≤ 12.6 mm2, 12.6 mm2< BS2 ≤ 50.3 mm2, BS3 > 50.3 mm2...... 50

Table 3.3 Values of db50 and dl50 averaged by species and by basal section (BS) class. Basal section classes are defined as: BS1 ≤ 12.6 mm2, 12.6 mm2< BS2 ≤ 50.3 mm2, BS3 > 50.3 mm2. The range was obtained as the difference between the maximum and the minimum values observed within all plants from each species...... 51 Table 4.1 General soil characteristics (mean±SE, n=6)...... 67 Table 4.2 Aboveground plant cover before trench excavation. Basal area is the sum of individual cross sections measured at the stem base per m2 (mean±SE, n=6)...... 68

xi xii

Table 4.3 Average root density (number of root counts/dm2) including all depths (mean±SE, n=6). Means followed by the same letter are not significantly different (p<0.05) according to the Mann-Whitney U test. The first letter refers to differences among diameter classes (columns) and the second letter refers to differences among species (rows). Symbol “_” represents no data...... 69

Table 5.1 Mean ± SE values of Di (soil moisture differences in terms of % volume, between the treatment and the reference periods). Means sharing the same letter are not statistically different (p >0.05, n=4). Letters in the first column refer to comparisons between plots within each season (t tests). Letters in the second column refer to multiple comparisons between depths (Tuckey tests)...... 90 Table 6.1 List of input variables...... 101 Table 6.2 Descriptive statistics of soil water measurements (% vol.) representing 318 days between May 2000 to December 2001, for the control and the burned plots. SD – Standard Deviation; SE – Standard Error...... 106 Table 6.3 Results of linear regressions (n=318) for comparison between measured (abscissa) vs. modelled (ordinate) soil water data for the two plots and each depth...... 108

xii 1

1 INTRODUÇÃO

1.1 As comunidades arbustivas das regiões mediterrânicas A designação de ecossistema mediterrânico, apesar de utilizada recorrentemente, nem sempre recolhe unanimidade quando aplicada aos ecossistemas existentes em Portugal. Em parte devido à situação geográfica do nosso país, exterior à Bacia Mediterrânea, há frequentemente dificuldade em classificar clara e inequivocamente um determinado tipo de clima e de comunidade vegetal como possuindo características mediterrânicas. Esta dificuldade é tanto maior quanto maior é a influência dita atlântica, que em Portugal se faz sentir sobretudo junto ao litoral e em particular na Região Norte. Esta dicotomia Mediterrâneo - Atlântico, embora tendo na sua essência causas de natureza climática, reflecte-se não só ao nível da vegetação mas num leque alargado de características físicas e humanas, tal como é retratado em Ribeiro (1991). Apesar da situação geográfica do território português, o facto de o autor apenas dedicar 28 páginas ao Portugal Atlântico e 99 páginas ao Portugal Mediterrâneo, diz bem da predominância do segundo sobre o primeiro. No entanto não se pense que esta indefinição constitui uma dificuldade apenas relativa ao território português. Tendo em conta a necessidade de delimitar as zonas de influência mediterrânica, diversos autores tentaram utilizar critérios objectivos de forma a estabelecer limites claros e definidos, para a Região Mediterrânea e/ou para outras regiões do Planeta com clima mediterrânico (Aschmann, 1973; Daget, 1977; Quézel, 1985). Para além da Região Mediterrânea existem quatro outras regiões no Mundo onde é possível encontrar um clima e uma vegetação com características semelhantes (Walter, 1979): a California nos EUA, a região litoral Centro do Chile, a costa sudoeste da África do Sul e a costa sul e sudoeste da Austrália. No caso concreto da Região Mediterrânea, Quézel (1985) compara a utilização de critérios florísticos, critérios fitossociológicos, critérios climáticos e critérios bioclimáticos para definir os seus limites. Através dos mapas resultantes da aplicação destes critérios constata-se que todos eles resultam na inclusão de todo ou quase todo o território português, na Região Mediterrânea. Também Le Houérou (1992) refere como possuindo clima mediterrânico os dois terços sul do território português. Vale enfim a pena reter a definição mais ou menos consensual de bioclima mediterrânico avançada por Daget (1977). Segundo este autor um clima é mediterrânico nas regiões em que o Verão é a estação mais seca e em 1 2 que existe um período de stress fisiológico para as plantas. Uma referência final para o trabalho de Costa et al. (1998) onde o território português é, em termos fitogeográficos, quase todo incluído na Região Mediterrânica. Exceptua-se apenas a região situada sensivelmente a norte de Aveiro-Viseu e a oeste de Vila Real, a qual é incluída na Região Eurosiberiana. Parecendo assim clara a marcada influência mediterrânica no território Português, há no entanto que ter em conta as especificidades dos nossos ecossistemas e em particular das formações arbustivas quando comparadas com as de outras zonas da Região Mediterrânea. Aqui há que entrar sem dúvida em conta com o efeito moderador da vizinhança do oceano sobre o nosso território, o qual se traduz essencialmente em precipitações mais elevadas e em menores amplitudes térmicas que noutras zonas da Região Mediterrânea com uma influência mais continental. Este aspecto é responsável em boa medida pelo carácter mais mésico e menos xérico da vegetação que compõe uma boa parte das comunidades arbustivas em Portugal. De acordo com a simplificação da classificação climática de Emberger adoptada por Di Castri (1981), estas regiões são classificadas como sub-húmidas (5-6 meses secos) ou húmidas (3-4 meses secos) dentro do universo de climas mediterrânicos. A este carácter mésico estão normalmente associadas comunidades arbustivas mais altas e mais densas, normalmente englobadas na designação francesa de maquis a qual não tem um equivalente claro no vocabulário Português. De entre as diferentes designações aplicadas em Portugal podemos referir as de mato alto ou matagal. Este tipo de formações tem equivalente noutras regiões mediterrânicas do Globo, recebendo a designação de fynbos na África do Sul, de mallee na Austrália, de chaparral na California e de matorral no Chile. Na Bacia Mediterrânea estas comunidades são geralmente consideradas etapas intermédias da sucessão, ou antes como etapas intermédias de degradação relativamente a formações climácicas dominadas por espécies arbóreas do género Quercus (Le Houérou, 1981). No outro extremo encontram-se as formações altamente degradadas de vegetação arbustiva esparsa e baixa, designadas em França como garrigue, na Grécia como phrygana e em Espanha como tomillares. A sua ocorrência no nosso país está normalmente associada a zonas do Sul com um grau elevado de aridez ou a zonas de serra com uma frequência elevada de fogo e pastoreio. Ao contrário, os matagais altos resultam inevitavelmente da ocorrência de períodos relativamente prolongados com ausência de perturbações. De um modo geral caracterizam-se por uma baixa diversidade florística sendo normalmente dominados por uma ou duas espécies arbustivas constituindo populações com elevada

2 3 densidade e também altamente susceptíveis ao fogo (Naveh, 1994). De entre as diferentes espécies que se podem apresentar como dominantes neste tipo de formações arbustivas importa referir, devido à sua importância em Portugal, as dos géneros Erica, Quercus (sobretudo Q. coccifera), Pistacia, Arbutus, Cytisus e Cistus (sobretudo C. ladanifer). O local escolhido para a realização do presente estudo, a Tapada Nacional de Mafra, encontra-se maioritariamente coberto por matagais dominados por Erica scoparia e Erica lusitanica sendo igualmente comum o aparecimento de plantas do género Ulex. Em termos fitogeográficos, Costa et al. (1998) inclui esta região no Superdistrito Olissiponense apontando como formações climácicas para esta região, bosques de Quercus suber e de Quercus faginea. e como etapa sucessional de substituição, matagais dominados por Ulex.

1.2 O papel dos sistemas radicais nas estratégias evolutivas das plantas mediterrânicas De entre os diferentes critérios utilizados para definir o clima mediterrânico, resulta das considerações anteriores que o traço fundamental é a existência de um período de secura estival. A ocorrência de uma estação quente e seca tem duas consequências principais as quais se constituem como forças modeladoras, essenciais para a compreensão das características evolutivas das plantas mediterrânicas: a ocorrência do fogo e a deficiência de água no solo durante uma parte do ano. Em relação ao primeiro aspecto é comum agrupar as plantas dos ecossistemas mediterrânicos em tipos funcionais resultantes dos mecanismos de adaptação ao fogo. A este respeito são numerosas as tentativas de encontrar sistemas de classificação que permitam agrupar as plantas de forma a reflectir características comuns de adaptação ao fogo (Le Houérou, 1973; Naveh, 1975; Noble & Slatyer, 1980; Gill, 1981; Keeley, 1991; Trabaud, 1992, Keeley, 1995; Noble & Gitay, 1996; Pausas, 1999). De todos estes sistemas ressalta um aspecto comum fundamental que consiste na separação das diferentes espécies em dois grandes grupos funcionais de acordo com a sua capacidade para regenerar ou não vegetativamente. Deste modo, o estudo dos mecanismos ecológicos que envolvem as espécies mediterrânicas passa assim em boa parte pela abordagem das suas características regenerativas enquanto adaptações ao fogo. Podemos assim de forma simplista considerar um grupo constituído pelas espécies com aptidão para regenerar vegetativamente (resprouters) e um outro constituído pelas

3 4 espécies que apenas podem regenerar por semente (seeders). O primeiro grupo inclui todas as espécies cuja regeneração imediatamente após o fogo é garantida através do lançamento de novos rebentos com origem em tecidos que resistiram, ou não foram atingidos, pelas altas temperaturas. Estes tecidos localizam-se no caule ou em órgãos subterrâneos tais como rizomas, bolbos ou tubérculos. No segundo grupo incluem-se todas as espécies cujos indivíduos morrem após a ocorrência do fogo e que, como tal, estão completamente dependentes da regeneração por via seminal para poder assegurar a sua continuidade. Apesar de se tratar apenas de uma característica regenerativa, esta diferença de estratégias tem sido associada a uma multiplicidade de outros aspectos da morfologia, da fisiologia e da ecologia das respectivas espécies (Margaris, 1981; Ferreira, 1996; Bell, 2001). Em termos gerais as espécies de regeneração vegetativa podem atingir um porte mais elevado e possuem um ciclo de vida mais longo, têm uma taxa de crescimento mais baixa, têm uma menor produção de sementes e têm um sistema radical mais extenso. Estão normalmente associadas a etapas mais avançadas da sucessão, nomeadamente a matagais altos e apenas regeneram por semente quando estão reunidas condições ambientais mais favoráveis à germinação e ao crescimento das plântulas (Silva & Rego, 1998a). O recrutamento é feito de forma gradual dando origem a baixas densidades de plântulas (Silva & Rego, 1998b; Clemente et al., 1994). Dadas estas características, é normal referir estas espécies como estrategas K (Ferreira, 1996; Díaz Barradas et al., 1999) no âmbito das definições para as estratégias populacionais propostas por McArthur & Wilson (1967). Ao nível das espécies existentes em Portugal são incluídas neste grupo quase todas as espécies normalmente apontadas como dominantes das formações climácicas referidas para o nosso país, nomeadamente os géneros Quercus, Arbutus, Pistacia, Rhamnus, Viburnum, Phyllirea e Laurus, entre outros. As espécies de regeneração obrigatória por semente têm de um modo geral as características opostas às anteriores e exibem frequentemente adaptações morfológicas e anatómicas à secura tais como indumento, menor área foliar, produção de óleos voláteis, cutícula e mesófilo mais espessos e um mais eficiente controlo estomático (Keeley, 1986). Estas espécies têm tendência a dominar em zonas mais secas e menos férteis e em fases pouco evoluídas da vegetação, sendo frequentemente associadas a matos baixos e dispersos do tipo phrygana (Margaris, 1981). Muitas delas dão origem a um banco de sementes que se podem manter no solo em estado de dormência durante vários anos. Em oposição ao grupo anterior e dado o intenso recrutamento que ocorre sempre que estejam reunidas as condições necessárias à germinação, nomeadamente o calor

4 5 proporcionado pelo fogo (Christensen & Muller, 1974; Papanastasis & Romanas, 1977; Arianoutsou & Margaris, 1982; Mazzoleni, 1989; Valbuena et al., 1990; Gonzalez- Rabanal & Casal, 1993) ou a simples abertura de clareiras, estas espécies são normalmente conotadas com a estratégia r. De entre as espécies que ocorrem no nosso país são importantes representantes deste grupo as plantas dos géneros Cistus, Lavandula, Rosmarinus e Halimium. No entanto existem numerosas espécies que, muito embora estando obrigatoriamente incluídas numa das duas categorias referidas, não correspondem minimamente ao padrão de características descrito. Daí a necessidade constatada por vários autores de identificar sub-grupos que tivessem um número mínimo de semelhanças funcionais nomeadamente no que toca às adaptações ao fogo. De entre os trabalhos já citados parece-nos interessante o modelo proposto por Pausas (1999) o qual simplesmente sub-divide os dois grupos de acordo com a existência ou não de um recrutamento de plântulas directamente favorecido pelo fogo, dando assim origem a quatro categorias distintas. No que toca aos mecanismos de sobrevivência ao stress hídrico ou tolerância à secura (drought tolerance) também diferentes autores têm proposto diferentes sistemas de agrupamento das estratégias adoptadas pelas várias espécies. Muito embora exista alguma inconsistência em torno da terminologia adoptada (Kozlowski et al., 1991), é mais ou menos clássica a distinção entre mecanismos para evitar a secura (drought/dessication avoidance) e mecanismos de resistência/tolerância à secura (drought/dessication resistance/tolerance) (Larcher, 1975; Kozlowski, 1982, Kozlowski et al., 1991). O primeiro tipo de mecanismos tem a ver com adaptações estruturais ao passo que o segundo tem a ver com adaptações fisiológicas ao nível celular. Enquanto que no segundo caso se tratam de mecanismos que permitem à planta resistir a baixos potenciais hídricos, o primeiro tipo de estratégia consiste essencialmente em evitar a ocorrência desses níveis de potencial hídrico. Jones (1992) distingue ainda num grupo à parte os mecanismos de eficiência directa e indirecta do uso da água e Fitter (2002) divide os mecanismos para evitar a secura, em adaptações para a aquisição da máxima quantidade de água e adaptações para a conservação e uso eficiente da água. Esta última interpretação corresponde a considerar a existência de espécies que resistem à secura essencialmente gastando água (water spenders) e um outro grupo de espécies que essencialmente resistem à secura poupando água (water savers). São inúmeras as descrições das adaptações estruturais desenvolvidas pelas plantas para evitarem o stress hídrico. Algumas delas foram atrás referidas dado existir

5 6 alguma correspondência entre as estratégias regenerativas das espécies mediterrânicas após o fogo e as estratégias de adaptação das mesmas espécies ao stress hídrico. Um dos principais aspectos a ter em conta a este respeito tem a ver com a estrutura do sistema radical dado esta estar intimamente ligada à possibilidade de regeneração vegetativa e simultaneamente, à capacidade de exploração do solo e de extracção de água e nutrientes. No caso das espécies que possuem regeneração vegetativa é fundamental a existência de um sistema de raízes extenso e profundo que permita responder às necessidades em água e nutrientes dos rebentos que surgem após o fogo. A este nível algumas espécies como as ericáceas desenvolveram paralelamente a formação de órgãos tuberosos de reserva onde são armazenados hidratos de carbono sob a forma de amido (lignotubers). Nas espécies de regeneração vegetativa é frequente a formação de raízes pivotantes (tap roots) as quais podem atingir profundidades de dezenas de metros (Kummerow, 1981; Canadell et al., 1996). Estas raízes têm um papel fundamental durante a estação seca dado permitirem à planta aceder a camadas de solo com teores de humidade bastante superiores aos da superfície, onde existe uma muito maior concentração de raízes finas e onde se fazem sentir os efeitos directos da evaporação (Nepstad et al., 1994; Canadell & Zedler, 1995). Esta possibilidade de aceder a camadas de solo mais húmido permite que estas espécies não necessitem das adaptações estruturais já referidas (indumento, folhas reduzidas, cutícula espessa, entre muitas outras) normalmente associadas à poupança de água, típicas das plantas mais xerofíticas. No caso das espécies de regeneração por semente tudo se passa de modo a permitir que durante o seu relativamente curto ciclo de vida as plantas consigam crescer e produzir sementes, o seu único meio de assegurar a continuidade da espécie. Deste modo o investimento faz-se preferencialmente na parte aérea em detrimento do desenvolvimento radical, dando normalmente origem a relações raíz-parte aérea mais baixas (Kummerow, 1981). Dado apenas poderem aceder a camadas de solo sujeitas a uma intensa dessecação durante os meses de Verão, estas plantas necessitam de adaptações estruturais de defesa contra a secura no sentido de evitar ao máximo as percas de água de forma a manter o seu equilíbrio hídrico. A tipificação de sistemas radicais que foi referida para os dois grandes grupos funcionais em que se podem incluir as plantas mediterrânicas, tem no entanto inúmeras variantes. Estas variantes têm a ver não só com as características próprias de espécies “menos típicas” mas bastante com a influência dos factores ambientais, nomeadamente do solo (Fitter, 1996). A existência de diferentes condições ambientais pode ser responsável por importantes diferenças no

6 7 sistema radical de plantas da mesma espécie (Kummerow, 1981, Canadell & Zedler, 1995). Deste modo o estudo comparativo dos sistemas radicais de espécies com diferentes estratégias adaptativas apenas pode ser completamente conclusivo quando as condições de solo, nomeadamente em termos de profundidade e distribuição de horizontes, são semelhantes. As características da distribuição das raízes das plantas individuais reflectem-se directamente ao nível das comunidades de plantas. O estudo da distribuição das raízes no solo de uma comunidade de plantas permite assim conhecer o potencial de absorção para a água e para os nutrientes ao longo do perfil do solo e de certa forma enquadrar dentro das estratégias referidas as espécies que a compõem. O conhecimento da distribuição das raízes no solo reveste-se de uma importância fundamental sempre que se pretendem representar através de modelos, os fluxos de água, carbono ou nutrientes numa comunidade de plantas (Feddes et al., 2001). A este nível, quando não é possível conhecer a distribuição vertical das raízes, é pelo menos fundamental o conhecimento da profundidade máxima de enraizamento dado que tal permite avaliar a extensão do perfil de solo no qual ocorrem os processos de absorção de água e nutrientes (Schenk & Jackson, 2002b). Quando estudamos uma comunidade de plantas há ainda que ter em conta o factor competição o qual pode ser responsável por importantes modificações ao nível da distribuição das raízes no solo (Atkinson, 1978; Casper & Jackson, 1997; Casper et al., no prelo).

1.3 O estudo das raízes das plantas mediterrânicas Quase todos os autores são unânimes em reconhecer as dificuldades associadas ao estudo das raízes no solo (e.g. Kummerow, 1981; Box, 1996; Atkinson, 2000). Mesmo tendo em conta as inovações tecnológicas que podem ser utilizadas actualmente, o estudo das raízes das plantas continua a ser um desafio muitas vezes recusado, dada a baixíssima relação entre os resultados obtidos e o esforço investido. De acordo com Jackson et al. (1996), os primeiros estudos relatados sobre raízes terão começado há mais de 250 anos através das investigações levadas a cabo por S. Hales em 1727 com plantas cultivadas. Desde cedo foram tentadas técnicas várias para observar as raízes das plantas. Um dos primeiros relatos sobre a utilização de água sob pressão para a escavação de raízes data de 1857 e a utilização de paneis de vidro para a observação do crescimento radicular data de 1873. No entanto foi já no século XX que foram estabelecidas metodologias consistentes para o estudo das raízes das plantas. A este

7 8 respeito merecem referência os trabalhos pioneiros de alguns ecologistas notáveis do início do século passado tais como J. Weaver e W. Cannon (Fitter, 1996; Guowei et al., 1997). Dada a panóplia de técnicas e métodos entretanto especificamente desenvolvidos para o estudo das raízes das plantas durante o século XX, alguns autores dedicaram-se a reunir essa informação de forma sistematizada. Consultando alguma dessa bibliografia é curioso verificar a evolução sofrida ao nível dos métodos para o estudo de raízes, através dos trabalhos de Schuurman & Goedewaagen (1971), Böhm (1979), Caldwell & Virginia (1989) e do recente trabalho editado por Smit et al. (2000). Essa evolução está sobretudo relacionada com os avanços tecnológicos verificados entre cada uma das publicações referidas. Estes avanços têm-se verificado em diferentes direcções no sentido de facilitar a recolha de informações sobre a estrutura e o funcionamento das raízes. De entre as diferentes inovações tecnológicas recentes podemos referir a título de exemplo: - A utilização de técnicas até há pouco reservadas a outros campos da ciência como a Tomgografia Assistida por Computador e a Ressonância Magnética (Asseng et al., 2000) ou o uso de detecção através de radar (Butnor et al., 2001). - O melhoramento e a generalização do uso de sistemas ópticos para observação e registo de imagens destinadas a estudar o crescimento de raízes através de mini-rizotrões e outros sistemas afins (Smit et al., 2000). - A utilização de radio-isótopos permitindo o estudo dos fluxos de água nutrientes e dióxido de carbono ou mesmo da distribuição das raízes no solo (Milchunas et al., 1992; Bingham et al., 2000, Casper et al., no prelo). - A utilização de técnicas de identificação de DNA (Linder et al., 2000). Esta técnica permite associar as raízes observadas em cavernas ou outros locais abaixo da superfície do solo às as respectivas plantas. - A utilização de técnicas de aquisição e tratamento de imagens digitais incluindo o desenvolvimento de novas ferramentas informáticas especificamente direccionadas para o estudo de raízes (Richner et al., 2000). - O desenvolvimento de novos sistemas possibilitando a escavação de raízes in situ, como o Air Spade® (Concept Engineering Group, Inc. Verona, E.U.A.) ou o registo tridimensional da sua arquitectura como o Fastrak® (Polhemus Inc. Colchester), (Danjon et al., 1998).

8 9

- De referir ainda de um modo geral as enormes capacidades computacionais existentes hoje em dia quando comparadas por exemplo com aquelas existentes na altura da publicação do clássico (ainda hoje seguido) trabalho sobre métodos de estudo dos sistemas radicais por W. Böhm em 1979. No entanto apesar de todas as inovações tecnológicas referidas, uma grande parte dos estudos de raízes de plantas desenvolvidas em condições naturais continua, e continuará (veja-se o presente trabalho por exemplo) provavelmente durante muito tempo, a ser realizada com base em métodos que na sua essência pouco diferem dos métodos clássicos utilizados desde há mais de cem anos. Talvez por esse motivo é bastante elevado o número de trabalhos de investigação utilizando plantas desenvolvidas em condições controladas, dada a maior facilidade do estudo das suas raízes. No entanto este tipo de trabalhos tem importantes limitações, quer ao nível da dificuldade em reproduzir as condições naturais, quer em relação à possibilidade de estudar plantas lenhosas adultas. Por outro lado são conhecidas as dificuldades do estudo de raízes em condições naturais em particular quando é necessário estudar plantas com sistemas radicais profundos. Tais limitações fazem com que não sejam abundantes, em termos relativos, os exemplos de estudos sobre as raízes de plantas lenhosas adultas. Devido a todas estas limitações, a esmagadora maioria dos trabalhos diz respeito a espécies com interesse agrícola devido à sua maior importância económica, e uma boa parte da metodologia para o estudo de raízes foi desenvolvida com este tipo de plantas. No caso concreto dos estudos de raízes realizados em regiões com características mediterrânicas, é normal contar com dificuldades acrescidas. Entre essas dificuldades contam-se as características de uma boa parte dos solos destas regiões, frequentemente pouco profundos e com elevada pedregosidade. Por outro lado as características das próprias plantas e das comunidades que constituem são igualmente pouco favoráveis, nomeadamente no caso de espécies com raízes profundas e no caso de formações de matagal denso tipo maquis onde é difícil separar as raízes de diferentes indivíduos ou mesmo de diferentes espécies. Estes são alguns dos motivos apontados por Kummerow (1981) para justificar a escassez de estudos destinados ao conhecimento da estrutura dos sistemas radicais das espécies que compõem os ecossistemas mediterrânicos. Se é verdade que alguns trabalhos têm sido desenvolvidos noutras regiões com características mediterrânicas, em particular na California, no que toca à Região Mediterrânea têm sido muito poucos os estudos com plantas não cultivadas. Nas

9 10 compilações sobre trabalhos relativos à distribuição das raízes no solo constantes em Canadell et al. (1996), Jackson et al. (1996), Schenk & Jackson (2002a) e Schenk & Jackson (2002b) é notável a escassez de trabalhos relativos aos países da Região Mediterrânea. Restringindo-nos em particular à Peninsula Ibérica, as espécies Quercus ilex (Canadell & Rodà, 1991; Djema, 1995) e Quercus coccifera (Kummerow et al., 1990; Cañellas & Ayanz, 2000) têm merecido especial atenção. Vale também a pena mencionar os trabalhos de Martinez & Rodriguez (1988) e Martinez et al. (1998) em comunidades arbustivas do Sul de Espanha. O conhecimento sobre as características estruturais das raízes das comunidades arbustivas mediterrânicas em Portugal e sobre as espécies que as constituem é ainda mais escasso na medida em que deverão ser, tanto quanto sabemos, praticamente inexistentes os estudos a este respeito. Em linguagem comum podemos dizer que conhecemos razoavelmente o que está à vista mas desconhecemos quase completamente o que está escondido abaixo da superfície do solo. Se bem que contemplando outros aspectos e outro tipo de espécies, devemos no entanto referir alguns trabalhos pioneiros sobre raízes realizados no nosso país. É esse o caso dos trabalhos realizados por A. Fabião (e.g. Fabião et al., 1985; Fabião et al., 1987; Fabião et al., 1991) sobre Eucalyptus globulus e ainda o trabalho realizado por M. Tavares (Tavares, 1989) sobre Pinus pinaster. Porventura outros trabalhos existirão mas apesar dos nossos esforços, deles não tivemos conhecimento. Já no que toca ao estudo dos sistemas radicais das plantas agrícolas, o número de trabalhos realizados em Portugal é bastante superior podendo servir de exemplo os estudos realizados por C.A. Portas e M.C. Oliveira (e.g. Portas, 1973; Portas & Taylor, 1976; Oliveira & Portas, 1987), e ainda por C.A. Pacheco (Pacheco, 1989; Rodrigues et al., 1995).

1.4 Os efeitos do fogo na dinâmica da água do solo Os efeitos do fogo nos ecossistemas mediterrânicos são estudados desde há bastante tempo e em múltiplas vertentes. Dado o acumular de conhecimentos nesta área, o estudo dos efeitos do fogo transformou-se numa espécie de sub-disciplina dentro do âmbito mais alargado da investigação sobre incêndios florestais. É normal dividir as questões ligadas aos efeitos do fogo em: efeitos na vegetação, efeitos na fauna e efeitos no solo. Dada a crescente importância das questões ligadas às previsões de aquecimento do planeta devido às emissões de CO2, o estudo dos efeitos dos incêndios na atmosfera tem também vindo a ganhar um peso crescente. De entre as várias categorias de efeitos,

10 11 o estudo dos efeitos na vegetação tem merecido especial atenção por parte dos investigadores, tal como se pode por exemplo constatar pelo número de trabalhos apresentados nas quatro edições da International Conference on Forest Fire Research. No entanto para uma compreensão integrada dos mecanismos envolvidos na dinâmica da vegetação após o fogo, é fundamental saber o que se passa abaixo da superfície do solo. De entre os estudos sobre os efeitos do fogo no solo, muitos têm sido aqueles dedicados às alterações verificadas ao nível dos nutrientes ou da erosão superficial mas muito poucos têm sido os trabalhos publicados sobre as alterações verificadas na dinâmica da água no solo. Tal lacuna no conhecimento da dinâmica dos ecossistemas mediterrânicos e da chamada ecologia do fogo, poderá dever-se ao facto de só desde há relativamente poucos anos estarem disponíveis meios técnicos economicamente acessíveis, fiáveis e inócuos para permitir a monitorização da humidade do solo ao longo do tempo e a diferentes profundidades. Este conhecimento é fundamental na medida em que o teor de água no solo funciona como factor limitante relativamente ao crescimento das plantas em clima mediterrânico. Deste modo podemos afirmar que todo o processo de recuperação da vegetação após o fogo está estreitamente associado ao teor de água no solo, quer enquanto causa quer enquanto consequência. Assim o desaparecimento dos órgãos aéreos da vegetação devido ao fogo tem como efeito imediato uma eliminação temporária da transpiração através das folhas e, consequentemente, uma redução drástica na absorção de água pelas raízes. No entanto o conjunto de mecanismos que se verificam após a ocorrência de um incêndio é bem mais complexo na medida em que a evaporação à superfície do solo aumenta devido ao desaparecimento do coberto vegetal (Pyne et al., 1996; Zwolinsky, 2000). Por outro lado a infiltração no solo, da água proveniente da precipitação, pode sofrer uma diminuição associada a um aumento do escoamento superficial. Estes fenómenos poderão ser devidos à eliminação do coberto vegetal e ainda à frequentemente relatada formação no solo de uma camada hidrófoba em incêndios de grande intensidade (DeBano, 1966; Ferreira, 1990; Midoun et al., 1998; DeBano, 2000). Finalmente há ainda que ter em conta a diminuição da intercepção da precipitação pelas copas o que faz com que a quantidade de água que chega à superfície do solo seja maior. O peso relativo de cada um destes mecanismos depende de vários factores, nomeadamente: da intensidade do fogo, do tipo de vegetação, dos factores meteorológicos, das características do solo e da topografia local.

11 12

Para além de escassa, a bibliografia existente relatando o efeito do fogo no teor de água no solo é contraditória. Na origem dos diferentes resultados obtidos estão diferenças ao nível dos métodos aplicados mas também diferenças em termos das condições estudadas. Uma boa parte dos estudos existentes neste domínio, como em muitos outros relacionados com a ecologia do fogo, foram realizados pelos Serviços Florestais dos Estados Unidos, em povoamentos florestais. Alguns dos trabalhos publicados relatam uma diminuição do teor de água no solo após o fogo quando comparado com o de parcelas testemunha. É esse o caso do estudo relatado por Campbell et al. (1977) num povoamento de Pinus ponderosa e também o caso de alguns dos estudos relatados na revisão feita por Wells et al. (1979). Resultados opostos foram relatados por Klock & Helvey (1976) numa experiência levada a cabo num povoamento misto de resinosas e por Soto & Diaz-Fierros (1997) num matagal dominado por Ulex europaeus. Também em Portugal, Rego & Botelho (1992) chegaram a resultados semelhantes ao verificar a existência de teores ligeiramente mais elevados de humidade no solo durante o Verão após a realização de um fogo controlado num povoamento jovem de Pinus pinaster. Grande parte dos trabalhos constituem apenas uma abordagem parcial do problema, quer por não fazerem o acompanhamento da situação durante um período suficientemente alargado quer, mais frequentemente, por não efectuarem o estudo ao longo de todo o perfil explorado pelas raízes. Deste modo para uma abordagem completa dos fenómenos envolvidos é importante ter em conta as variações da água no solo quer em termos temporais quer em termos espaciais. Este segundo aspecto deverá ser analisado em paralelo com a distribuição das raízes no solo de forma a poder ser avaliado o peso dos mecanismos de absorção de água pelas plantas no balanço de percas e ganhos.

1.5 A modelação da dinâmica da água no solo O estudo dos ecossistemas passa cada vez mais pela criação e utilização de modelos matemáticos de simulação dos diferentes processos associados ao funcionamento das comunidades vegetais e animais. Tal deve-se, entre outros aspectos, a uma crescente facilidade associada à criação de modelos e à cada vez maior necessidade de avaliar o impacto das actividades humanas nos processos naturais. Em relação ao primeiro aspecto tem sido vertiginosa a evolução verificada ao nível do desempenho quer dos computadores quer dos programas informáticos. No tocante a este último aspecto importa referir o aparecimento desde há alguns anos, de programas

12 13 interactivos destinados a facilitar a tarefa da construção de um modelo. Tratam-se de programas gráficos de modelação que permitem ao utilizador a construção intuitiva de modelos através da simples utilização de ícones gráficos. Estes ícones podem ser arrastados num diagrama e interligados de forma a representar por exemplo, fluxos entre compartimentos e relações entre variáveis. Tal permite reduzir bastante a morosidade do trabalho de modelação relativamente ao que acontecia anteriormente devido à necessidade de utilização de linguagens de programação. Exemplos deste tipo de ferramentas de modelação são os programas STELLA® (High Performance Systems Inc., Hanover, E.U.A.), SIMILE (Simulistics Ltd., Edinburgh, R.U.), SIMULINK® (The MathWorks, Inc., Natick, E.U.A) ou ainda DYMEX (CSIRO Publishing). Pormenores sobre os dois primeiros podem ser encontrados respectivamente em Hannon & Ruth (2001) e Muetzelfeldt & Taylor (2001). Relativamente ao segundo aspecto é de salientar a importância das alterações climáticas globais devido ao chamado “efeito de estufa”. Este tipo de alterações globais atribuídas ao aumento do teor de CO2 na atmosfera tem sido um dos principais incentivos à criação de modelos de circulação global (GCM – Global Circulation Models) os quais por sua vez são utilizados em modelos de simulação do funcionamento dos ecossistemas, de forma a prever alterações futuras na vegetação e restantes organismos. Para além deste, existem muitos outros efeitos das actividades humanas nos ecossistemas que têm sido estudados através de modelos tais como o impacto da poluição nos solos e no crescimento das plantas ou o impacto dos incêndios florestais. No entanto é importante referir que a utilidade da criação de modelos ultrapassa bastante a possibilidade de previsão de novos cenários e situações (Ford, 1999). Na verdade a utilização de modelos no estudo dos ecossistemas tem sobretudo a grande virtualidade de permitir uma melhor compreensão dos fenómenos associados ao funcionamento desses mesmos ecossistemas. Esta compreensão é frequentemente necessária a diferentes escalas quer temporais quer espaciais, pelo que a criação de modelos tem muitas vezes em conta essa necessidade. Um exemplo largamente difundido de um modelo de dinâmica de ecossistemas delineado para funcionar a diferentes escalas no tempo e no espaço é o modelo FOREST-BGC (Running & Cougham, 1988; Running & Gower, 1991). Trata-se de um modelo geral de funcionamento dos ecossistemas florestais baseado nos três ciclos biogeoquímicos básicos associados ao desenvolvimento das plantas: o ciclo do carbono, o ciclo dos nutrientes e o ciclo da água. Este último ciclo tem uma importância fundamental a nível de todos os ecossistemas terrestres e em particular nos ecossistemas mediterrânicos, já 13 14 que a disponibilidade de água no solo é um factor limitante relativamente ao crescimento das plantas em clima mediterrânico (Daget, 1977). O modelo de crescimento GOTILWA - Growth of Trees is Limited by Water (Gracia et al., 1999) é um exemplo da importância atribuída ao ciclo da água no funcionamento dos ecossistemas mediterrânicos dado ter sido desenvolvido e validado em particular para este tipo de condições. Dada a importância da água para as plantas, para além dos sub- modelos incorporados em modelos mais abrangentes como os dois exemplos referidos, existe uma profusão de modelos estritamente destinados à simulação da dinâmica da água no solo e nas interfaces solo-planta-atmosfera. Para além de todos os modelos desenvolvidos especificamente para culturas agrícolas, existem ainda alguns modelos particularmente vocacionados para a simulação da dinâmica da água nos ecossistemas florestais. De um modo geral o solo é encarado como um compartimento sujeito a entradas e saídas de água. Nos modelos mais desenvolvidos este compartimento pode ser dividido em sub-compartimentos homogéneos (horizontes ou simplesmente camadas de solo) em que as entradas e saídas de água afectam e são afectadas pelos sub- compartimentos vizinhos. A partir desta estrutura geral, diferentes soluções têm sido encontradas para a representação dos diferentes fluxos de água nas interfaces solo- planta-atmosfera. Tal implica uma selecção criteriosa das variáveis com maior influência no processo a simular e das respectivas relações, sendo este sem dúvida o aspecto mais crítico de qualquer trabalho de modelação de sistemas naturais (Jones, 1992). A simulação da dinâmica da água no solo implica a resolução de diferentes questões relacionadas por um lado, com os processos de evapotranspiração e por outro com os fluxos de água no solo. Para além destes, outros aspectos podem ou não ser tidos em conta tais como a intercepção da precipitação pelas copas e pela folhada a avaliação do escorrimento superficial, ou a libertação gradual de água pelo manto de neve no caso de climas onde tal ocorre. Existem ainda mecanismos que, tanto quanto sabemos, nunca foram contabilizados neste tipo de simulações devido à escassez de dados a este nível. É esse o caso, por exemplo, dos mecanismos de absorção e posterior libertação de água pelas raízes, permitindo o transporte de água entre diferentes profundidades no solo (Caldwell et al., 1991; Burgess et al., 1998). No tocante à evapotranspiração esta pode ser calculada de diferentes formas, sendo a mais generalizada a equação de Penman-Monteith (Monteith, 1965; Tiktak & Bouten, 1992; Heidmann et al., 2000; Williams et al, 2001). Outros autores optam por

14 15 soluções mais simples como a equação de Makkink (Makkink, 1957; Tiktak & Bouten, 1994) ou a equação de Priestley-Taylor (Priestley & Taylor, 1972; Krysanova et al., 1998). A componente transpiração pode ser calculada como uma função de Michaelis- Menten relativamente à radiação fotossintéticamente activa (e.g. Gracia et al., 1999). No tocante aos processos de evaporação da água do solo e de superfícies húmidas é frequentemente utilizada a aproximação proposta por Ritchie (1972) para culturas agrícolas, a qual se baseia numa relação empírica entre a transpiração potencial e o índice de área foliar (e.g. Paruelo e Sala, 1995; Krysanova et al., 1998). Para a simulação dos fluxos de água no solo é necessário determinar as propriedades hidráulicas do solo. Trata-se aqui fundamentalmente de determinar as curvas de retenção de humidade e a condutividade hidráulica não saturada em função do teor de humidade para cada horizonte ou tipo de solo (Hillel, 1998; Porta et al., 1999). Com este objectivo diversos autores desenvolveram relações empíricas baseadas na textura do solo e em alguns casos também no teor de matéria orgânica (Mualem, 1976; Gupta & Larson, 1979; Van Genuchten, 1980; Saxton et al., 1986; Vereecken et al., 1989). A absorção de água pelas plantas é uma componente com uma abordagem normalmente simplificada dada a frequente inexistência de dados sobre a distribuição das raízes no solo. Deste modo alguns autores optam por utilizar simplesmente uma densidade homogénea (Feddes et al., 1978), outros representam o decréscimo da densidade de raízes ao longo do perfil através de uma função linear (Heidmann et al., 2000), de uma função potência (Monteith et al., 1989), ou ainda de uma função exponencial (Williams et al., 2001). Outros autores utilizam valores específicos de densidade relativa de raízes para cada camada de solo considerada (e.g. Tiktak & Bouten, 1992; Sala & Paruelo, 1995). O fluxo de água entre camadas de solo é calculado para condições de saturação através da lei de Darcy (e.g. Gracia et al., 1999) e para o solo não saturado através da equação de Richards (Feddes e Koopmans, 1995; Tiktak e Bouten, 1992). Dada a profusão de modelos que foram sendo elaborados ao longo dos anos, podem também ser encontradas revisões comparativas incluindo diferentes modelos (e.g. Feddes et al., 1988; Tiktak & Grinsven, 1995). No entanto, apesar de toda a actividade de modelação desenvolvida, trata-se de um esforço direccionado sobretudo para servir objectivos relacionados com a produção agrícola ou florestal. Tanto quanto sabemos nenhum dos modelos descritos na literatura disponível foi delineado ou validado com dados obtidos a partir de matagais mediterrânicos. Deste modo o desenvolvimento de modelos com esta vocação poderá

15 16 constituir-se como uma mais-valia, nomeadamente tendo em conta a milenar relação entre este tipo de ecossistemas e os factores de perturbação de natureza antrópica como o fogo, o corte e o pastoreio.

1.6 Justificação, objectivos e estrutura da presente tese Todos os trabalhos de investigação desenvolvidos no âmbito da presente tese tiveram como origem a participação portuguesa no projecto ModMED III-Modelling Vegetation Dynamics in Mediterranean Ecosystems (contrato ENV4-CT97-0680), através da Estação Florestal Nacional (EFN). Este projecto Europeu (DG XII) foi coordenado pelo Professor Stefano Mazzoleni da Universidade de Nápoles e teve o seu funcionamento entre Janeiro de 1998 e Março de 2001, dando sequência aos projectos ModMED I (1995-1996) e ModMED II (1996-1997). As outras instituições participantes através de contrato directo foram as Universidades de Edimburgo, de Atenas e de Pisa, sendo a participação portuguesa realizada em associação com a Universidade de Edimburgo e coordenada pelo Professor Francisco de Castro Rego. Em termos gerais, os projectos ModMED tiveram como objectivos o estudo dos ecossistemas mediterrânicos a diferentes escalas (a paisagem, a comunidade de plantas e o indivíduo) e a modelação do seu funcionamento através da utilização de ferramentas de modelação desenvolvidas no âmbito do próprio projecto. A este nível foi dada particular ênfase ao efeito das perturbações (fogo, pastoreio, corte) nos ecossistemas, dada a importância deste tipo de mecanismos na Região Mediterrânea. Os projectos ModMED I e II permitiram identificar a existência de lacunas importantes de conhecimento ao nível do funcionamento dos ecossistemas, em particular naquilo que se refere aos fenómenos que ocorrem abaixo da superfície do solo. Deste modo um dos grandes desígnios do projecto ModMED III foi o de tentar colmatar algumas dessas lacunas através do estudo de aspectos aparentemente tão elementares como a distribuição das raízes no solo, as características estruturais das raízes das diferentes espécies ou a dinâmica da água no solo. Os temas e a estrutura da presente tese surgem assim como uma consequência directa da tentativa de responder aos desafios colocados a este projecto de investigação. Dado o enorme vazio de conhecimento sobre a estrutura e o funcionamento das raízes dos ecossistemas mediterrânicos, foi necessário estabelecer prioridades e deixar de lado muitos aspectos de inegável interesse e importância. Essas prioridades passaram pela identificação dos factores cuja importância se poderia vir a tornar fundamental para o desenvolvimento de modelos de

16 17 simulação de funcionamento dos ecossistemas, tendo em conta as diferentes escalas a abordar. Os trabalhos de investigação foram assim focados em dois aspectos distintos se bem que complementares: as características estruturais das raízes e a dinâmica da água no solo. Em relação ao primeiro aspecto foi decidido trabalhar à escala da comunidade e à escala da planta individual. Em particular foram tidos em conta os aspectos relativos à distribuição vertical das raízes no solo e à identificação de características que permitissem relacionar as diferentes espécies com as respectivas estratégias em termos ecológicos. Dado o enorme esforço que sabíamos ser necessário investir na recolha e preparação das plantas para a abordagem à escala individual, foi necessário optar entre um estudo mais intensivo de uma ou duas espécies e um estudo mais exploratório incluindo um leque mais alargado de espécies. Tendo em conta o carácter, tanto quanto sabemos inédito, deste tipo de estudos em Portugal, optou-se pela segunda hipótese, apesar dos riscos que acarretava em termos das dificuldades associadas ao tratamento estatístico dos dados e à posterior obtenção de resultados conclusivos. Por outro lado foi necessário alargar a amostragem para além da comunidade de plantas inicialmente escolhida como objecto de estudo de forma a incluir-mos um leque mais alargado de espécies. Neste leque de espécies apenas foram incluídas espécies lenhosas dado as espécies herbáceas estarem quase ausentes ou terem uma importância funcional reduzida nos matagais mediterrânicos. De forma a rentabilizar o esforço investido, as mesmas plantas foram utilizadas para dois estudos distintos, um dedicado à tipificação das principais características estruturais das respectivas raízes e um outro dedicado a estudar a distribuição vertical das raízes no solo. Ao nível da comunidade foi preocupação fundamental a obtenção de dados também sobre a distribuição vertical das raízes de forma a apoiar os estudos de água no solo. A este respeito foi igualmente tida em conta a profundidade máxima de enraizamento da comunidade estudada dada a importância deste parâmetro na avaliação das possibilidades de extracção de água das camadas mais profundas do solo. Em termos do estudo das raízes foi necessário deixar de fora aspectos fundamentais que mereceriam sem dúvida ter sido estudados tais como a dinâmica das raízes no solo, a influência dos factores ambientais na diferenciação dos sistemas radicais ou as relações hídricas das plantas. As razões para não o termos feito prenderam-se com óbvias limitações materiais e de tempo.

17 18

Fig. 1.1 Aspecto geral do matagal de urze (Erica scoparia e Erica lusitanica) na Tapada Nacional de Mafra, onde decorreram os trabalhos relatados nos capítulos 4, 5 e 6 e onde foram escavadas algumas das plantas estudadas nos capítulos 2 e 3.

Em relação aos aspectos ligados à dinâmica da água no solo, foram tidas sobretudo em conta as necessidades associadas ao desenvolvimento de um modelo de simulação integrando informações climáticas, pedológicas, e de caracterização da comunidade de plantas. A este respeito pareceu-nos óbvia a inclusão do factor fogo enquanto variável a estudar, dada a sua importância neste tipo de ecossistemas. A escolha da Tapada Nacional de Mafra (Fig. 1.1) para a realização dos estudos, teve como origem razões de natureza institucional (a ligação da EFN à Tapada Nacional de Mafra), razões de natureza prática (zona protegida com apoios ao nível de infra- estruturas e equipamento) e ainda razões de natureza científica (comunidades arbustivas mediterrânicas do tipo matagal ou maquis). Tendo em conta a linha geral de investigação traçada em função dos objectivos do projecto europeu a que já fizemos referência, os objectivos da presente tese foram os seguintes: - Contribuir para um melhor conhecimento dos sistemas radicais das plantas lenhosas mediterrânicas, nomeadamente no que se refere: ao relacionamento entre diferentes parâmetros caracterizadores das raízes, às alterações sofridas por esses parâmetros ao longo de diferentes estádios de desenvolvimento e à relação entre esses parâmetros e as estratégias regenerativas das plantas estudadas. 18 19

- Caracterizar a distribuição vertical de uma série de plantas lenhosas mediterrânicas utilizando um modelo inédito e desenvolvido com esse objectivo. Em particular pretendeu-se: comparar o desempenho do modelo proposto com outros modelos existentes e utilizar esse modelo para descrever e interpretar a distribuição vertical das raízes de diferentes plantas lenhosas mediterrânicas. - Caracterizar a distribuição vertical das raízes da comunidade arbustiva estudada nomeadamente no que diz respeito: às diferentes espécies, às diferentes classes de diâmetro, à profundidade máxima de enraizamento e à biomassa e comprimento total das raízes finas. - Determinar o efeito de curto prazo de um fogo experimental na dinâmica da água do solo a diferentes profundidades, na comunidade arbustiva. - Apresentar e testar através de duas séries de dados reais, um novo modelo de simulação da dinâmica da água no solo, assim como interpretar os resultados obtidos no sentido de conhecer melhor os mecanismos de funcionamento do ecossistema estudado. Os cinco capítulos base que compõem a tese (para além do presente capítulo introdutório e de um capítulo de conclusões) foram redigidos em Inglês, ou seja na língua original em que foram submetidos para publicação. A referência a cada uma das publicações encontra-se incluída numa nota de rodapé na primeira página de cada capítulo. As razões que nos levaram a optar por esta modalidade tiveram a ver com dois aspectos. Por um lado, o facto de a Língua Inglesa ser cada vez mais a forma universal de expressão nos meios científicos, possibilitando assim uma maior garantia de divulgação do trabalho realizado. O outro aspecto prendeu-se com a possibilidade de poder contar com uma revisão adicional dos diferentes capítulos por especialistas ligados às revistas ou comissões cientificas às quais os trabalhos foram submetidos. Tal possibilidade constitui à partida um potencial contributo para aumentar a qualidade do trabalho apresentado, enquanto tese de doutoramento. Deste modo, à parte pequenas correcções introduzidas posteriormente, o conteúdo de cada capítulo reflecte no seu essencial o conteúdo dos trabalhos submetidos e publicados. O preço a pagar por esta fidelidade ao trabalho original traduz-se sobretudo na constatação de pequenas incoerências e na repetição de aspectos metodológicos, quando se comparam os diferentes capítulos entre si. Algumas incoerências devem-se sobretudo ao facto de cada capítulo ter incluído, à data a que foi

19 20 submetido ou revisto (pela ordem apresentada nesta tese), novos elementos ou soluções que não estavam ainda presentes nos capítulos anteriores. Tal é o caso do presente capítulo de introdução, o qual inclui uma revisão bibliográfica mais actualizada e em alguns casos mais aprofundada que nos capítulos submetidos a publicação e terminados há bastante mais tempo. No tocante ao capítulo final trata-se fundamentalmente de uma tradução para Português da secção Discussion dos capítulos submetidos a publicação. Não podemos aqui deixar de referir as dificuldades enfrentadas para conseguir uma tradução correcta para Português, nos capítulos inicial e final desta tese, de alguma terminologia utilizada nos capítulos publicados em Inglês e pouco utilizada fora dos meios científicos.

Bibliografia

Arianoutsou-Faragitaki M. & Margaris N.S. (1982). Phryganic (East Mediterranean) ecosystems and fire. Ecologia Mediterranea 8: 473-480. Aschmann H. (1973). Distribution and peculiarity of Mediterranean ecosystems. In Di Castri F. & Mooney H.A. (eds.), Mediterranean-type ecossystems: origin and structure. Springer-Verlag, New York, 225-283 pp. Asseng S., Aylmore L.A.G., MacFall J.S., Hopmans J.W. & Gregory P.J. (2000). Computer-assisted tomography and magnetic resonance imaging. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 1-32 pp. Atkinson D. (1978). The use of soil resources in high density planting systems. Acta Horticulturae 65: 75-90. Atkinson D. (2000). Root characteristics: Why and what to measure. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 1-32 pp. Bell D.T. (2001). Ecological response syndromes in the flora of southwestern Western Australia: Fire resprouters versus reseeders. The Botanical Review 67: 417-440. Bingham I.J., Glass A.D.M., Kronsucker H.J., Robinson D. & Scrimgeour C.M. (2000). Isotope techniques. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 365-402 pp. Böhm W. (1979). Methods of studying root systems. Springer Verlag, Berlin. Box J.E. (1996). Modern methods for root investigation. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 193-238 pp. Burgess S.O., Adams M.A., Turner N.C. & Ong C.K. (1998). The redistribution of soil water by tree root systems. Oecologia 115: 306-311. Butnor J.R., Doolittle J.A., Kress L., Cohen S. & Johnsen K.H. (2001). Use of ground-penetrating radar to study tree roots in the south-eastern United States. Tree Physiology 21: 1269-1278. Caldwell M.M. & Virginia R.A. (1989). Root systems. In Pearcy R.W., Ehleringer J.R., Mooney H.A. & Rundel P. (eds.), Physiological plant ecology. Field methods and instrumentation. Chapman and Hall, London, 367-398 pp. Caldwell M.M., Richards J.H. & Beyschlag W. (1991). Hydraulic lift: ecological implications of water efflux from roots. In Atkinson D. (ed.), Plant Root Growth, an Ecological Perspective. Blackwell Scientific Publications, Oxford, pp. 423-436.

20 21

Campbell R.E., Baker P.F., Ffolliot P.F., Larson F.R. & Avery C.C. (1977). Wildfire effects on a ponderosa pine ecosystem. An Arizona case study. U.S.D.A. Forest Service Research Paper RM- 191. Rocky Mountains Forest and Range Experimental Station. Fort Colins. Canadell J. & Rodà F. (1989). Root biomass of Quercus ilex in a montane mediterranean forest. Canadian Journal of Forest Research 21: 1771-1778. Canadell J. & Zedler P.H. (1995). Underground structures of woody plants in mediterranean ecosystems of Australia, California and Chile. In Arroyo M.T., Zedler P.H. & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecosystems in Chile, California and Australia. Springer Verlag, Berlin, 177-210 pp. Canadell J., Jackson R.B., Ehleringer J.R., Mooney H.A., Sala O.E. & Schulze E.-D. (1996). Maximum root depth of vegetation types at the global scale. Oecologia 108: 583-595. Cañellas I. & Ayanz A. (2000). Biomasss of root and shoot systems of Quercus coccifera shrublands in Eastern . Annals of Forest Sciences 57: 803-810. Casper B.B. & Jackson R.B. (1997). Plant competition underground. Annual Review of Ecology and Systematics 28: 545-570. Casper B.B., Schenk H.J. & Jackson R.B. Defining a plant's belowground zone of influence. Ecology, no prelo. Christensen N.L. & Muller C.H. (1974). Relative importance of factors controling germination and seedling survival in Adenostoma chaparral. The American Midland Naturalist 93: 71 -78. Clemente A.S., Rego F.C. & Correia O.A. (1994). Demographic patterns and productivity of post-fire regeneration in Portuguese mediterranean maquis. In Proceedings of the 2nd International Conference on Forest Fire Research, Coimbra, Portugal, 937-950 pp. Costa J.C., Aguiar C., Capelo J.H. & Neto C. (1998/99). Aproximação à biogeografia de Portugal Continental. Quercetea 0: 56 pp. Daget P. (1977). Le bioclimat mediterranean: Characteres generaux, modes de characterisation. Vegetatio 34: 1-20. Danjon F., Sinoquet H., Godin C., Colin F. & Drexhage M. (1998). Characterisation of structural tree root architecture using 3D digitising and AMAPmod software. Plant and Soil 211: 241-258. DeBano L.F. (1966). Formation of a non wettable soils involves heat transfer mechanism. General Technical Report PSW-132. U.S.D.A. Forest Service Research Notes PSW-46. Pacific Southwest Forest and Range Experimental Station, Berkeley. DeBano L.F. (2000). Fire-Induced Water Repelency: An Erosional Factor in Wildland Environments. In Proceedings of Conference Land Stewardship in the 21st century. The Contributions of Watershed Management. Tucson, U.S.A., 307-310 pp. Di Castri F. (1981). Mediterranean-type shrublands of the World. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 1-52 pp. Díaz Barradas M.C., Zunzunegui M., Tirado R., Ain-Llout F. & Garcia Novo, F. (1999). Plant functional types and ecosystem function in Mediterranean shrubland. Journal of Vegetation Science 10: 709-716. Djema A. (1995). Cuantification de la biomassa e mineralomassa subterranea de un bosque de Quercus ilex. Tese de mestrado. Instituto Agronomico Mediterraneo de Zaragoza. Fabião A., Madeira M. & Steen, E. (1987). Root mass in plantations of Eucalyptus globulus in Portugal in relation to soil characteristics. Arid Soil Research and Rehabilitation 1: 185-194. Fabião A., Madeira M. & Steen, E. (1991). Effect of water and nutrient supply on root distribution in an Eucalyptus globulus plantation. Water Air and Soil Pollution 54: 635-640. Fabião A., Persson H.A. & Steen E. (1985). Growth dynamics of superficial roots in Portuguese plantations of Eucalyptus globulus Labill. studied with a mesh bag technique. Plant and Soil 83: 233-242.

21 22

Feddes R., Kowalik P. & Zaradny H. (1978) Simulation of field water use and crop yield. Pudoc, Wageningen. Feddes R.A., Hoff H., Bruen M., Dawson T.E., de Rosnay P. Dirmeyer P., Jackson R.B., Kabat P., Kleidon A., Lilly A., & Pitman, A.J. (2001). Modeling root water uptake in hydrological and climate models. Bulletin of the American Meteorological Society 82: 2797-2809. Feddes R.A., Kabat J.T., Bronswijk J.J., & Halbertsma J. (1988). Modelling soil water dynamics in the unsaturated zone: state of the art. Journal of Hydrology 100: 69-111. Feddes, R.A. & Koopmans, R.W. (1995). Agrohydrology. Department of Water Resourses. University Wageningen, Wageningen. Ferreira A.D. (1990). Fire Effect on Soil Water Dynamics. Proceedings of the 1st International Conference on Forest Fire Research. Coimbra, Portugal, C07-1. Ferreira O. (1996). Estrategias regenerativas de especies arboreas de ecosistemas forestales de Galicia en relacion con incendios: analisis del comportamiento germinativo y de la demografia de plantulas. Tese de doutoramento. Universidad de Santiago de Compostela. Fitter A.H. & Hay K.M. (2002). Environmental physiology of plants. Academic Press. San Diego. 3rd ed. Fitter A.H. (1996). Characteristics and functions of root systems. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 1-20 pp. Ford A. (1999). Modeling the environment. Island Press, Washington. Gill M. (1981). Adaptative response of Australian vascular plant species to fires. In Gill A.M., Groves R.H. & Noble I.R. (eds.), Fire and the australian biota. Australian Academy of Sciences, Camberra, 243-271 pp. Gonzalez-Rabanal F. & Casal M. (1993). Effects of thermal shock on germination of Ulex europaeus L. in wildfire-affected and unburnt soils. In Trabaud L. & Prodon R. (eds.), Fire in mediterranean ecosystems. Commission of the European Communities, Brussels, 201-208 pp. Gracia C.A., Tello E., Sabaté S. & Bellot, J. (1999) Gotilwa: an integrated model of water dynamics and forest growth. In Rodà F., Retana J., Gracia C.A. & Bellot J. (eds.), Ecology of mediterranean evergreen oak forests. Springer-Verlag, Berlin, 163-178 pp. Guowei S., Coffin D.P. & Lauenroth W.K. (1997). Comparison of root distribution of species in North American grasslands using GIS. Journal of Vegetation Science 8: 587-596. Gupta S.C. & Larson W.E. (1979). Estimating soil water characteristics from particle-size distribution, organic matter percent and bulk density. Water Resources Research 15: 1633-1635. Hannon B. & Ruth M. (2001). Dynamic modeling. Springer-Verlag, New York. 2nd ed. Heidmann T., Thomsen A. & Shelde K. (2000). Modelling soil water dynamics in winter wheat using different estimates of canopy development. Ecological Modelling 129: 229-243. Hillel D. (1998). Environmental soil physics. Academic Press, San-Diego, 2nd edition. Jackson R.B., Canadell J., Ehleringer J.R., Mooney H.A., Sala O.E. & Shulze E.D. (1996). A global analysis of root distributions for terrestrial biomes. Oecologia 108: 389-411. Jones H.G. (1992). Plants and microclimate. A quantitative approach to environmental plant physiology. Cambridge University Press, Cambridge. 2nd ed. Keeley J.E. (1986). Resilience of Mediterranean shrub communities to fires. In Dell A., Hopkins J.M. & Lamont B.B. (eds.), Resilience in mediterranean-type ecosystems. Dr. Junk Publishers, Dordrecht, 95-112 pp. Keeley J.E. (1991). Seed germination and life history syndromes in the California Chaparral.The Botanical Review 57: 81-116. Keeley J.E. (1994). Seed-germination patterns in fire-prone Mediterranean-climate regions. In Arroyo M.T.K., Zedler P.H., & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecossystems in Chile, California and Australia. Springer-Verlag, New York, 239-273 pp.

22 23

Klock G.O. & Helvey J.D. (1976). Soil water trends following wildfire on the Entiac Experimental Forests. Proceedings of the 15th Tall Timbers Fire Ecology Conference Tall Timber Research Station, Talahassee, U.S.A., 193-200 pp. Kozlowski T.T. (1982). Water supply and tree growth. Forestry abstracts 43: 57-95. Kozlowski T.T., Kramer P.J. & Pallardy S.G. (1991). The physiological ecology of woody plants. Academic Press Inc., San Diego. Krysanova V., Müller-Wohlfeil D.I. & Becker A (1998). Development and test of a spatially distributed hydrological / water quality model for mesoscale watersheds. Ecological Modelling 106: 261- 289. Kummerow J. (1981). Structure of roots and root systems. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 269- 288 pp. Kummerow J., Kummerow M. & Trabaud L. (1990). Root biomass, root distribution and the fine-root growth dynamics of Quercus coccifera L. in the garrigue of Southern France. Vegetatio 87: 37- 44. Larcher W. (1975). Physiological plant ecology. Springer-Verlag, New-York. Le Houérou H.N. (1973). Fire and vegetation in the Mediterranean Basin. In Proceedings of the 13th Annual Tall Timbers Fire Ecology Conference, Tall Timber Research Station, Talahassee, U.S.A., 237-277 pp. Le Houérou H.N. (1981). Impact of man and his animals on mediterranean vegetation. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 1-52 pp. Le Houérou H.N. (1992). Vegetation and land-use in the Mediterranean Basin by the year 2050: A prospective study. In Jeftic L., Milliman J.D. & Sestini G. (eds.), Climatic Change and the Mediterranean. Edward Arnold, London, 175-232 pp. Linder C.R., L.A., Moore & Jackson R.B. (2000). A universal molecular method for identifying underground plant parts to species. Molecular Ecology 9: 1549-1559. Makkink G.F. (1957). Testing the Penman formula by means of lysimeters. International Journal of Water Engineering 11: 277-288. Margaris N.S. (1981). Adaptive strategies in plants dominating Mediterranean-type ecosystems. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 309-315 pp. Martinez F. & Rodriguez J.M. (1988). Distribución vertical de las raices del matorral de Doñana. Lagascalia 15: 549-557. Martínez F., Merino O., Martín A., García Martín D., & Merino J. (1998). Belowground structure and production in a mediterranean sand dune shrub community. Plant and Soil 201: 209-216. Mazzoleni S. (1989). Fire and Mediterranean plants: germination responses to heat exposure. Annali di Botanica 47: 227 - 233. McArthur R. & Wilson E.O. (1967) The theory of island biogeography. Princetown University Press, Princetown. Midoun M., Picard C., Prosper-Laget V. & Rebattu L. (1998). Modification of hydrous-physical soil behaviour after the passage of a fire. Proceedings of the 3rd International Conference on Forest Fire Research. Luso-Coimbra, Portugal, 1687-1706 pp. Milchunas D.G., Lee C.A., Lauenroth W.K. & Coffin D.P. (1992). A comparison of 14C, 86Rb and total excavation for determination of root distributions of individual plants. Plant and Soil 144: 125- 132. Monteith J.L. (1965). Evaporation and Environment. In State and Movement of Water in Living Organisms. Proceedings of the 19th Symposium of the Society of Experimental Biology. Cambridge University Press,Cambridge, 205-234 pp.

23 24

Monteith J.L., Huda A.K.S. & Midya D. (1989). RESCAP: a resource capture model for sorghum and pearl millet. In Virmani S.M., Tandon H.L.S., Alagarswamy (eds.), Modelling the growth and development of sorghum and pearl millet. ICRISAT Research Bulletin 12: 30-34. Mualem Y. (1976). A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12: 513-522. Muetzelfeldt R.I. & Taylor J. (2001). Developing forest models in the Simile visual modelling environment. In Rennolls K. (ed.), Proceedings of IUFRO 4.11 Conference: Forest Biometry, Modelling and Information Science. University of Greenwich, Greenwich, UK. Naveh Z. (1975). The evolutionary significance of fire in the Mediterranean region. Vegetatio 29: 199- 208. Naveh Z. (1994). The role of fire and its management in the conservation of mediterranean ecosystems and landscapes. In Moreno, J.M. & Oechel W.C. (eds.), The role of fire in mediterranean-type ecosystems. Springer-Verlag, New York, 163-185 pp. Nepstad D.C., Carvalho C.R., Davidson E.A., Jipp P.H., Lefebvre P.A., Negreiros G.H., Silva E.D. da, Stone T.A., Trumbore S.E. & Vieira S. (1994). The role of deep roots in the hydrological and carbon cycles of amazonian forests and pastures. Nature 372: 666-669. Noble I.R. & Gitay H. (1996). A functional classification for predicting the dynamics of landscapes. Journal of Vegetation Science 7: 329-336. Noble I.R. & Slatyer R.O. (1980). The use of vital attribute to predict successional changes in plant communities subject to recurrent disturbance. Vegetatio 43: 5-21. Oliveira M.R. & Portas C.A. (1987). Estudo da distribuição radical numa consociação de festuca, azevém, e trevo branco e num luzernal. Pastagens e Forragens 8: 215-230. Pacheco C.A. (1989). Influência de técnicas de não mobilização e de mobilização sobre aspectos estruturais e hídricos de solos com vinha, bem como sobre o respectivo sistema radical. Consequências das relações hídricas solo-vinha na produção. Tese de Doutoramento, Universidade Técnica de Lisboa. Papanastasis V.P. & Romanas C.R. (1977). Effect of high temperatures on seed germination of certain Mediterranean half-shrubs. Ministry of Agriculture Forest Research Institute Bulletin No. 86. Paruelo J.M. & Sala O.E. (1995). Water losses in the Patagonian steppe: A modelling approach. Ecology 76: 510-520. Pausas J.G. (1999). Mediterranean vegetation dynamics: modelling problems and functional types. Plant Ecology 140: 27-39. Porta J., López-Acevedo M. & Roquero C. (1999). Edafologia para la agricultura y el medio ambiente. Ediciones Mundi-Prensa, Madrid, 2ª edicion. Portas C.A. & Taylor H.M. (1976). Growth and survival of young plant roots in dry soil. Soil Science 121: 170-175. Portas C.A. (1973). Development of root systems during the growth of some vegetable crops. Plant and Soil 39: 507-518. Priestley C.H.B. & Taylor R.J. (1972). On the assessment of surface heat flux and evaporation using large scale parameters. Monthly Weather Review 100: 82-92. Pyne S.J., Andrews P.A. & Laven R.D. (1996). Introduction to Wildland Fire. John Wiley & Sons. New York. 2nd edition. Quézel P. (1985). Definition of the Mediterranean region and the origin of its flora. In Gómez-Campo C. (ed.), Plant conservation in the Mediterranean Area. Dr. W. Junk Publishers, Dordrecht, 9-23 pp. Rego F.C. & Botelho H.S. (1992). Soil water regimes as affected by prescribed fire in young Pinus pinaster forests in Northern Portugal. In Trabaud L., Prodon R. (eds.), Fire in mediterranean ecosystems. Commission of European Communities, Brussels, 423-432 pp. Ribeiro O. (1991). Portugal, o Mediterrâneo e o Atlântico. Livraria Sá da Costa, Lisboa.

24 25

Richner W., Liedgnens M., Bürgi H., Soldati A. & Stamp P. (2000). Root image analysis and interpretation. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 305-341 pp. Ritchie J.T. (1972). Model for predicting evaporation from a row crop with incomplete cover. Water Resources Research 8: 1204-1213. Rodrigues M.L., Pacheco C.A. & Chaves M.M. (1995). Soil-plant water relations, root distribution and biomass partitioning in Lupinus albus L. under drought conditions. Journal of Experimental Botany 46: 947-956. Running S.W. & Cougham J.C. (1988). A general model of forest ecosystem processes for regional applications I. Hydrologic balance, canopy gas exchange and primary production processes. Ecological modelling 42: 125-154. Running S.W. & Gower S.T. (1991). FOREST-BGC, a general model of forest ecosystem processes for regional applications II. Dynamic carbon allocation and nitrogen budgets. Tree physiology 9: 147-160. Saxton K.E., Rawls W.J., Romberger J.S. & Papendick R.I. (1986). Estimating generalized soil-water characteristics from texture. Soil Science Society of America Journal 50: 1031-1036. Schenk H.J. & Jackson R.B. (2002a). The global biogeography of roots. Ecological Monographs 72: 311- 328. Schenk H.J. & Jackson R.B. (2002b). Rooting depths, lateral root spreads, and belowground/aboveground allometries of plants in water-limited ecosystems. Journal of Ecology 90: 480-494. Schuurman J.J. & Goedewaagen M.A. (1971). Methods for the examination of root systems and roots. Centre for Agricultural Publishing and Documentation, Wageningen. Silva J.S. & Rego F.C. (1998a). Factors affecting the establishment of woody species after fire in Central Portugal. In Trabaud L. (ed.), Fire management and landscape ecology. International Association of Wildland Fire, Fairfield, 103-114 pp. Silva J.S. & Rego F.C. (1998b). Estimation of seedling densities of Mediterranean woody species after fire. In Viegas D.X. (ed.), Proceedings of the 3rd International Conference on Forest Fire Research, Luso, Portugal, 1753-1764 pp. Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S. & van de Geijn S.C. (eds.) (2000). Root methods, a handbook. Springer-Verlag, Berlin. Smit A.L., George E. & Groenwold J. (2000). Root observations and measurements at transparent interfaces with soil. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 235-272 pp. Soto B. & Diaz-Fierros F. (1997) Soil water balance as affected by throughfall in gorse (Ulex euroapeus, L.) shrubland after burning. Journal of Hydrology 195: 218-231. Tavares M.P. (1989). O pinhal bravo das dunas do litoral entre Douro e Mondego. Produção lenhosa e crescimento do sistema radical. Estação Florestal Nacional, Lisboa. Tiktak A. & Bouten W. (1992). Modelling soil water dynamics in a forested ecosystem III: Model description and evaluation of discretization. Hydrological Processes 6: 455-465. Tiktak A. & Bouten W. (1994). Soil water dynamics and long term water balances of a Doulgas fir stand in the Netherlands. Journal of Hydrology 156: 265-283. Tiktak A. & Grinsven H. (1995). Review of sixteen forest soil-atmosphere models. Ecological Modelling 83: 35-53. Trabaud L. (1992). Influence du regime des feux sur les modifications à court terme et la stabilité à long terme de la flore d’une garrigue de Quercus coccifera. Revue d’Ecologie (Terre et Vie) 47: 209- 230. Valbuena L., Tarrega R. & Luiz E. (1990). Influence of temperature on germination of seeds of Cistus laurifolius and Cistus ladanifer. In Proceedings of the 1st International Conference on Forest Fire Research, Coimbra, Portugal.

25 26

Van Genuchten, M.T. (1980). A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44: 892-898. Vereecken H., Maes J., Darius P.& Feyen J. (1989). Estimating the soil moisture retention characteristic from texture, bulk density and carbon content. Soil Science 148: 389-403. Walter H. (1979). Vegetation of the Earth and ecological systems of the geo-biosfere. Springer-Verlag, New York, 2nd edition. Wells C.G., Campbell R.E., DeBano L.F., Lewis C.E., Fredriksen R.L., Franklin E.C, Froelich R.C. & Dunn P.H. (1979). Effects of Fire on Soil. U.S.D.A. Forest Service General Technical Report WO-7. Williams M., Law B.E., Anthoni P.M. & Unsworth, M.H. (2001). Use of a simulation model and ecosystem flux data to examine carbon-water interactions in ponderosa pine. Tree Physiology 21: 287-298. Zwolinski M.J. (2000). The role of Fire in Management of Watershed Responses. In Proceedings of Conference Land stewardship in the 21st Century. The contributions of watershed management. Tucson, U.S.A., 367-370 pp.

26 27

2 BELOWGROUND TRAITS OF MEDITERRANEAN WOODY PLANTS IN A PORTUGUESE SHRUBLAND1

Abstract: Belowground traits vary widely. Apart from the influence of the environment both genetic and ontogenic factors are responsible for this variation. For mediterranean woody plants there is also evidence of a relationship between regenerative strategies and root system characteristics. With the general aim of studying these different aspects, the root systems of 17 obligate seeders and 16 resprouters from 10 different species and different developmental stages were excavated at Tapada Nacional de Mafra in Central West Portugal. Root systems were photographed, weighted and measured. Root length and the average root diameter were determined using digital image software. Root-to- shoot ratio (R/S) and the specific root length (SRL) were computed for all plants. Basal section was used as an indicator of plant development. A principal component analysis (PCA) was performed in order to study the relationships between variables and between plants. The analysis showed a clear distinction of plants according to the respective developmental stage but also according to the regenerative characteristics of the different species. Allometric relationships were found between root biomass, shoot biomass and basal section. Statistical tests showed that resprouters had higher maximum rooting depth, average root diameter and R/S and lower SRL, than obligate seeders. A decrease of R/S and SRL with basal section was verified for a sub-sample of four species. Key words: Root development, mediterranean shrublands, regenerative strategies, root systems, allometric relationships.

1 Based on paper: Silva J.S., Rego F.C. & Martins-Loução M.A. (2002). Belowground traits of mediterranean woody plants in a portuguese shrubland. Ecologia Mediterranea 28: 5-13.

27 28

Introduction The origins of the diversity of root systems can be seen as an optimisation of two primary functions: acquisition of soil-based resources (water and nutrients) and anchorage (Fitter, 1996). In the specific case of the Mediterranean regions of the world, root systems have probably evolved to deal with the strong spatial and temporal limitations on availability of water, which are typical of these regions (Canadell & Zedler, 1995). This evolution has resulted in various root type morphologies, characteristic of mediterranean woody species. Several attempts were made to classify root systems in terms of basic structural characteristics (Cannon, 1949; Hellmers et al., 1955; Specht & Rayson, 1957). However the well known plasticity of root systems as a result of environmental conditions, has limited the use of these classifications (Fitter, 1996). Bengough et al. (2000) suggested the study of variability as an alternative to seeking for homogeneous features. A simple classification of plants into deep rooted and shallow rooted is generally accepted, although no strict limits are normally established to define the two types. These two basic types of root systems correspond to different adaptations to the highly seasonal water availability typical of Mediterranean- type climates and have been associated with the strategy to regenerate after natural disturbances (fire and grazing). Typical resprouters normally present deep root systems whereas seeders are shallow rooted (Keeley, 1986; Correia & Catarino, 1994; Bell, 2001). In fact, resprouters need to have deep root systems to supply the growth of new shoots since they can not rely on a seed bank to regenerate after fire or other kind of disturbance (Keeley & Zedler, 1978; Clemente et al., 1996). Although sharing the same climatic conditions as resprouters, obligate seeders present specific water saving adaptations such as a higher stomatal control or leaf hairs, allowing these species to survive in drier conditions (Keeley, 1986; Correia, 1988; Correia & Catarino; 1994; Silva & Rego, 1998). These water saving mechanisms partly explain why obligate seeders withstand summer drought without the use of deep root systems. The study of root systems have also showed the existence of different patterns of root development. In the case of many mediterranean plants it is known the preferential allocation of resources to roots at early stages (Canadell & Zedler, 1995) and there is evidence that woody plants in general present a decreasing trend of root-to-shoot ratio with age (Kozlowski et al., 1991). The fact that many species, especially those under dry conditions, may develop deep tap roots at early stages preceding the development of

28 29 lateral roots (Spurr & Barnes, 1980) is an evidence of the specificity of rooting patterns at different developmental stages. Ontogeny seems then to be at least as important as phylogeny to the overall variation of root systems and the respective traits. It is an obvious fact that root systems are difficult to study. These difficulties of root studies increase in the case of plants from mediterranean ecosystems because soils are frequently shallow and heterogeneous and also because many species are growing deep root systems (Kummerow, 1981). Additional problems are found with plants from dense shrub communities because different individuals form an intricate root net which makes it extremely difficult to trace individual root systems. Consequently there is a considerable lack of knowledge of basic root system characteristics of mediterranean woody plants, and no information at all could be found concerning the belowground traits of the species studied in the present work. The relative scarcity of root studies in mediterranean ecosystems is well reflected in the planetary compilation of root distribution data by Jackson et al. (1996) where the Mediterranean Region is represented by only 4 studies out of 250. However it is generally recognised the importance of obtaining information on root systems for modelling the functions of ecosystems both at the plant as at the community level (Caldwell & Richards, 1986; Pagès, 2000) or even in global scale simulation models (Zeng, 2001). The general objective of this paper is to assess the existence of relationships concerning the belowground traits of mediterranean woody plants. In particular the present paper is focused on: a) the relationships between different belowground traits, b) the relationships between plant development and belowground traits, c) the relationships between regenerative strategies and belowground traits.

Methods Plants were collected at Tapada Nacional de Mafra, a public estate located in the Central West region of Portugal. Tapada Nacional de Mafra, is a protected area with 827 ha, about 30 km Northwest of and 12 km East from the coast (38º 56’ 37’’ to 38º 58’ 30’’ N and 9º 15’ 52’’ to 9º 18’ 43’’ W). The lowest altitude is 90 m and the highest is 358 m. Soil is a sandy loam classified as a humic cambisol (FAO classification) derived from sandstone. Bedrock is normally located below 2 meters. Mean annual precipitation is 798 mm and mean annual temperature is 14.6 ºC. Summer precipitation (June, July, August) accounts for only 3.1 % of the total annual rainfall.

29 30

Most of the area is constituted by shrublands dominated by Erica scoparia L. and Erica lusitanica Rudolphi. Between September 2000 and April 2001 the complete root systems of 33 plants from 10 woody species were hydraulically excavated (Böhm, 1979). Although we have tried to include a wide variety of woody species, our choice was limited by the floristic composition of the study area. Within these limits plants were chosen according to their developmental stage and regenerative characteristics. From the ensemble of plants, 17 were obligate seeders (Lavandula luisieri (Rozeira) Rivas Martinez (4 plants), Cistus crispus L. (9 plants) and Cistus salvifolius L. (4 plants)) and 16 were resprouters (Erica scoparia L. (2 plants), Erica lusitanica Rudolphi (1 plant), Crataegus monogyna Jacq. (4 plants), Ulex jussiaei Webb (3 plants), Daphne gnidium L. (4 plants), Pistacia lentiscus L. (1 plant) and Myrtus communis L. (1 plant)). In order to obtain 4 individuals representative of 4 classes of basal section, 2 obligate seeders (L. luisieri and C. crispus) and 2 resprouters (D. gnidium and C. monogyna) were sampled more intensively, constituting a pooled sub-sample of 16 plants. These 4 species apparently represented different root system types, within the obligate seeder and resprouter strategies. Basal section was used as an indicator of the plants developmental stage. The 4 classes of basal section were defined as: class 1 for plants up to 5 mm2, class 2 from 5 to 25 mm2, class 3 from 25 to 125 mm2 and class 4 for plants showing a basal section higher than 125 mm2. Despite having sampled nine C. crispus plants, no plants were collected at class 4, given the small size of the individuals present at the excavation site. No replications could be obtained for each combination species/basal section class due to the enormous amount of work required for the excavation and processing of complete root systems. All plants were excavated in spots where soil depth did not seem to limit the vertical development of roots, thus allowing the full expression of potential growth of deep roots. Besides having observed the plants used in this study, we were able to observe in the field other partially excavated individuals from the same species. This allowed to recognise some basic morphological characteristics, enabling a more complete description of root systems. After excavation, the maximum rooting depth, the maximum average root width and the basal section of each plant were measured. Maximum rooting depth represents the depth achieved by the deepest root, maximum average root width was computed as the average between the maximum horizontal width and the respective orthogonal width

30 31 and basal section was obtained as the cross sectional area at the stem base. In the case of plants with several stems, basal section was obtained by summing all the stem sections. In order to obtain the total root length and the average root diameter, root systems were photographed using a Fujifilm MX 2900 Zoom (Fuji Photo Film Co., Ltd., Tokyo) digital camera, featuring a maximum resolution of 2.3 millions of pixels. Digital images were analysed using software WinRhizo 4.1b (Regent Instruments, Quebéc). Some remarks have to be made on the determination of root length. Root length is difficult to measure when dealing with extensive root systems. Since most of the root length is associated with fine roots, it is extremely difficult to have an accurate estimate of total root length from shrubs or trees collected in the field. The main source of errors is the excavation process (Böhm, 1979, Caldwell & Virginia, 1989), where considerable amounts of fine roots are inevitably lost. Therefore root length was determined for comparison purposes only and it should be strictly interpreted on a relative basis and not as absolute values. For biomass determination plants were oven-dried at 85 ºC for 48 hours and weighted, separately for root and shoot fractions. In addition to these variables the specific root length (SRL; cm of total root length/g of root dry weight) and the root-to- shoot ratio (R/S; root dry weight/shoot dry weight) were computed for each plant. All root variables (including R/S ratio) were log transformed and used to perform a principal components analysis (PCA) for all plants. The principal components (PC’s) extracted by the analysis were related to the different individual plants collected, in order to evaluate which were the main variables responsible for the variation found and their relationship with species strategies and plant development. Allometric relationships between variables were established considering the results obtained from the PCA. Given that most variables did not follow a normal distribution (Shapiro Wilk W test), the Mann-Whitney U test was used to determine the significance of differences between obligate seeders and resprouters, for all variables. Tests were performed both with pooled samples of obligate seeders and resprouters and within developmental stages 2 and 3 (not enough plants within stages 1 and 4). The sub-sample of 16 plants referred above was used to assess the effect of the developmental stage on R/S and SRL for the different species. Species nomenclature followed Castroviejo (1999).

31 32

Results All C. monogyna plants presented thick, structural roots, with a smooth pale brownish bark. L. luisieri was characterised by a fibrous root system composed of an intricate horizontal network of very thin pale roots. Despite having only collected young U. jussiaei plants, our observation of partially excavated roots from adult individuals revealed that this species presented deep, thick, poorly lignified tap roots, showing a wrinkled white bark. C. salvifolius and C. crispus plants showed very similar characteristics. Roots of both species were relatively simple in terms of branching, with few dark brown structural roots exploring surface soil layers. The two Erica species were also very similar in terms of root system characteristics. Both presented a lignotuber to which a few deep, black, strongly lignified, tap roots were connected. Both species were also showing an intricate network of horizontal, not very widelly spreading fine roots exploring surface soil layers. D. gnidium plants showed a rooting pattern similar to the Erica plants with reddish, deep, tap roots and with only a few horizontal roots. Root tissues were poorly lignified which made coarse roots soft and rubber-like, similar to some succulent shrubs. All D. gnidium plants showed a swelling region, similar to typical lignotubers, just below the stem base. M. communis was characterised by the presence of coarse, pale brown laterally spreading roots, connected to a main deep reaching root. Given the impossibility of observing more than one root system in the field we did not retain the morphological characteristics of P. lentiscus. A quantitative description of all plants grouped by species is presented in Table 2.1. Considering all excavated plants, root biomass ranged from 111.8 g (E. scoparia) to 0.2 g (C. crispus) and shoot biomass ranged from 141.8 g (E. scoparia) to 0.3 g (C. crispus). Maximum rooting depth ranged from 185 cm (D. gnidium) to 12 cm (C. crispus). Other species, E. lusitanica and E. scoparia, also showed deep roots (160 cm and 140 cm, respectively). Root system width ranged from 95 cm (L. luisieri) to 5 cm (C. monogyna). Total measured root length ranged from 2791.0 cm (L. luisieri) to 65.7 cm (C. monogyna). Another C. monogyna plant had the second highest length of roots (1956.8 cm). The average root diameter ranged from 3.6 mm (C. monogyna) to 0.7 mm (C. crispus). R/S ratio ranged from 3.2 (D. gnidium) to 0.2 (L. luisieri). The three highest values of R/S ratio were from D. gnidium and the three lowest were from L. luisieri. Only seven plants presented values of R/S higher than 1. SRL ranged from 848.4 cm.g-1 (C. crispus) to 15.6 cm.g-1 (E. scoparia). The six highest values of SRL were from three C. crispus plants and three L. luisieri plants. 32

Table 2.1 Descriptive parameters (mean ± SE) of the root systems of 33 plants excavated at Tapada Nacional de Mafra, distributed by species. Legend for abbreviations: n – number of plants; Reg. strat. – regenerative strategy; Maxim. root. depth - maximum rooting depth; Aver. root diam. – average root diameter; R/S – root-to-shoot ratio; SRL – specific root length; s – obligate seeder; r – resprouter.

Species nReg. Basal section Shoot biomass Maxim. root. Root system Root length Aver. root Root biomass R/S SRL strat. (mm2) (g) depth (cm) width (cm) (cm) diam. (mm) (g) (g.g-1) (cm.g-1) C. crispus 9s 16.7 ± 3.6 2.6 ± 1.0 30.2 ± 5.0 19.1 ± 3.6 302.5 ± 66.1 1.0 ± 0.1 1.4 ± 0.4 0.7 ± 0.1 322.8 ± 88.7 C. salvifolius 4s 17.0 ± 10.7 3.1 ± 0.9 26.5.0 ± 5.0 15.6 ± 2.2 165.2 ± 29.8 1.4 ± 0.2 1.9 ± 0.9 0.5 ± 0.2 162.6 ± 56.2 C. monogyna 4r 86.4 ± 51.8 25.3 ± 16.9 45.3 ± 11.9 34.3 ± 15.2 610.7 ± 452.8 2.3 ± 0.3 19.7 ± 12.8 0.9 ± 0.1 51.3 ± 24.4 D. gnidium 4r 101.8 ± 82.3 31.2 ± 28.5 120.0 ± 36.6 31.8 ± 6.4 652.9 ± 225.7 2.1 ± 0.3 18.3 ± 13.8 1.8 ± 0.6 109.6 ± 41.0 E. lusitanica 1r 152.1 ± nd.3 43.1 ± nd.5 160.0 ± nd.6 48.5 ± nd. 1738.6 ± 4nd.4 1.5 ± nd. 36.7 ± nd.8 0.9 ± nd. 47.4 ± nd.4 E. scoparia 2r 94.7 ± 78.4 72.9 ± 68.9 90.0 ± 50.0 25.8 ± 11.8 1253.6 ± 494.4 1.5 ± 0.6 57.0 ± 54.8 0.7 ± 0.1 179.6 ± 163.9 L. luisieri 4s 66.4 ± 51.0 27.6 ± 23.9 28.3 ± 2.3 50.1 ± 16.9 1077.7 ± 585.9 1.2 ± 0.2 5.1 ± 4.1 0.3 ± 0.1 475.1 ± 120.2 M. communis 1r 41.4 ± nd.0 8.3 ± nd . 120.0 ± nd.6 73.5 ± nd. 676.0 ± 2nd.4 2.0 ± nd. 11.0 ± nd.8 1.3 ± nd. 61.4 ± nd.4 P. lentiscus 1r 19.4 ± nd.0 2.7 ± nd . 50.0 ± nd.7 20.0 ± nd. 393.5 ± 2nd.4 1.5 ± nd. 2.2 ± nd . 0.8 ± nd. 176.4 ± nd.4 U. jussiaei 3r 11.5 ± 2.9 3.3 ± 1.7 67.7 ± 28.7 14.0 ± 4.5 562.9 ± 212.4 1.1 ± 0.2 2.6 ± 1.5 0.8 ± 0.1 284.3 ± 57.2

33

34 1 PC2 DIAMETER R/S

BIOMASS DEPTH

WIDTH PC1 LENGTH

SRL

1.1 B PC2 2,5

Cm2 Cm3 1,5 Dg3 Es4 Dg1 Dg4 Mc3 Cm1 Cs3 Dg2 Cc3 Cm4 El4 0,5 Cs2 Uj2 Cc2 Cc2 Pl2

Uj2 PC1 Cc2 Cc2 -0,5 Cs2 Uj2 Cc3 Cs1 Es2 Cc2 Cc2 Cc1 Ll2 Ll4 Ll1 -1,5 Ll3

-2,5 -2,5 -1,5 -0,5 0,5 1,5 2,5 Fig. 2.1 PCA diagrams. A represents the components loadings for each variable and B represents the components scores for each plant individual. Legend for variables: DIAMETER – Average root diameter; BIOMASS – Root biomass; DEPTH – Maximum rooting depth; WIDTH – Root system width; LENGTH – Root length; R/S – Root-to-shoot ratio; SRL – Specific root length. Legend for species: Cc – Cistus crispus; Cs – Cistus salvifolius; Cm – Crataegus monogyna; Dg – Daphne gnidium; El – Erica lusitanica; Es – Erica scoparia; Ll – Lavandula luisieri; Mc – Myrtus communis; Pl – Pistacia lentiscus; Ru – Rubus ulmifolius; Uj – Ulex jussiaei. Symbols in bold correspond to obligate seeders. The developmental stage, as obtained by the respective basal section, is indicated by the number following the species symbol. Stage 1: 0 to 5 mm2; stage 2: 5 to 25 mm2; stage 3: 25 to 125 mm2; stage 4: > 125 mm2.

35

The first two principal components extracted by the PCA (Fig. 2.1A) explained 80 % of the total variance. Root length and root width were best correlated with PC1 (loadings 0.94 and 0.89, respectively). Root diameter and SRL were best correlated with PC2 (loadings 0.83 and –0.84, respectively). Root biomass was best correlated with PC1 (loading 0.82) and R/S was best correlated with PC2 (loading 0.74). Maximum rooting depth was poorly correlated with both principal components (loadings 0.69 and 0.44 for PC1 and PC2, respectively). The plot of component scores (Fig. 2.1B) presented a clear arrangement of plants according to their developmental stage, as defined by the basal section classes, with the only exceptions of a C. crispus and a C. salvifolius plant. Plants were also arranged according to the respective species and the two regenerative strategy groups. Most obligate seeders presented lower scores for PC2 (specially L. luisieri and C. crispus) while most resprouters presented higher scores for PC2 (specially D. gnidium and C. monogyna). This pattern was less evident at developmental stage 2. Given the results obtained with the PCA, allometric relationships were established through linear regression using the log-transformed values of root biomass, shoot biomass, root system length and root system width as independent variables and the log-transformed values of basal section as the dependent variable (Table 2.2). The linear regressions for root and shoot biomass presented coefficients of determination of 0.79 and 0.83 (p<0.001 for both regressions), respectively. The linear regression slopes were very similar (1.003 and 1.046, respectively). In the cases of root system length and root system width the relationships were still highly significant (p<0.001) showing coefficients of determination of 0.61 and 0.55, respectively.

Table 2.2 Allometric relationships obtained by linear regression between basal section (mm2) and four different root variables: root biomass, shoot biomass, root system length and root system width. All variables were log-transformed. Biomass data is indicated in decigrams in order to obtain only positive values. For each linear regression it is indicated the intercept (a), the slope (b), the coefficient of determination (r2) and the associated probability (p).

Variables a b r2 p Root biomass (0.1g)) 0.335 1.003 0.79 < 0.001 Shoot biomass (0.1g) 0.615 1.046 0.83 < 0.001 Root system length (cm) 4.383 0.528 0.61 < 0.001 Root system width (cm) 2.059 0.357 0.55 < 0.001

36

According to the Mann-Whitney U test, the ensemble of obligate seeders presented significantly lower values than the ensemble of resprouters for maximum rooting depth (p<0.001), average root diameter (p<0.01), root biomass (p<0.01) and R/S (p<0.01). Resprouters presented significantly lower values than obligate seeders for SRL (p< 0.01). Basal section, shoot biomass, root width and root length were not significantly different. Due to the smaller number of individuals, these differences could be only partially confirmed within developmental stages 2 (maximum rooting depth, p<0.05; root biomass, p<0.05) and 3 (maximum rooting depth, p<0.05). The graph of SRL plotted against the log-transformed values of basal section (Fig. 2.2) showed decreasing values for all four species and a clear separation between the two obligate seeders (C. crispus and L. luisieri) and the two resprouters. The same decreasing trend was observed for the R/S ratio values of the four species but in particular for D. gnidium. This trend can also be observed in the images of Fig. 2.3.

3,5 3,0 2,5 2,0 1,5 1,0

Root-Shoot (g.g-1) 0,5 0,0

1000 C. crispus

800 L. luisieri C. monogyna 600 D. gnidium 400

SR L (cm.g-1) 200

0 01234567 Log [Basal Section (mm2)]

Fig. 2.2 Relationships between Basal Section and two root system indices: Specific Root Length (SRL) and Root-to-Shoot ratio (Root-Shoot), for two obligate seeders ( Cistus crispus and Lavandula luisieri; represented in bold) and two resprouters (Daphne gnidium and Crataegus monogyna; normal lettering).

37

Fig. 2.3 Images of 16 plants representing different developmental stages (as defined in Figure 1) of two resprouters: A – Daphne gnidium (stages 1,2,3 and 4), B – Crataegus monogyna (stages 1,2,3 and 4); and two obligate seeders: C – Lavandula luisieri (stages 1,2,3 and 4), D – Cistus crispus (stages 1,2,2 and 3). The vertical bars represent 0.5 m. Arrows indicate the ground surface

38

Discussion The study of a very heterogeneous group of individuals normally has to deal with important constraints in terms of statistical analysis. However it may allow to study the factors responsible for the existing variability, as shown by the PCA using different belowground traits. Although apparently obvious in some aspects, the results obtained allow a better understanding of the different role of phylogeny and ontogeny in the differentiation of roots. The PCA showed that the developmental stage/basal section of the individuals was the basic cause for root system differentiation within our set of plants and studied variables. This trend was essentially related to increasing values of root length, root system width and root biomass. Root diameter and root length were associated with different principal components, suggesting that these two variables represent two separate factors in the distinction of the root systems of our set of plants. According to this, plants should be distinguished either by root length or by root diameter. However, root length was essentially able to distinguish different developmental stages, whereas root diameter was more related with species differentiation. Obligate seeders C. crispus and L. luisieri had in general lower scores of PC2 than resprouters suggesting that the average root diameter may be a distinctive trait between the two groups, as confirmed by the statistical test. The symmetric position of SRL relative to root diameter is in accordance with the use of this coefficient as an indicator of root diameter (Fitter, 1985). Another distinctive trait between seeders and resprouters seems to be the maximum rooting depth which was poorly correlated with either axis of the PCA. Thus, depth seems to work quite independently from the remaining root system variables including basal section. Two possible reasons may explain the weak relationship between maximum rooting depth and other traits, namely basal section. One is related to the fact that some species (e.g. D. gnidium and L. luisieri) seem to achieve maximum rooting depth early in the development, and then decrease or stop the vertical growth of roots (Fig. 2.3). The other, results from the existence of an inherent maximum rooting depth attained by each species. The early achievement of maximum rooting depth is undoubtedly an ecological advantage and seems to be a common feature of many mediterranean woody species (Canadell and Zedler, 1995). Although representing just a small fraction of the total root system, deep roots play a fundamental role to deal with water stress during the dry season especially at early developmental stages (Canadell et al., 1996). These roots reach deeper layers where water depletion is not so frequent as at the surface soil layers. Deep roots may be

39 responsible for more than 75% of the total water extracted during dry periods (Nepstad et al., 1994). Within the same developmental stage, differences concerning maximum rooting depth seem to be associated with the regenerative strategy of each species (Keeley, 1986; Correia & Catarino, 1994). In general and for similar developmental stages, obligate seeders presented relatively shallow root systems whereas resprouters presented deeper root systems. Within our database, L. luisieri and D. gnidium represented two extremes of this opposed trends. Root growth was essentially directed horizontally for L. luisieri and vertically for D. gnidium. In ecological terms, the significantly higher values found for root biomass, root diameter, and R/S ratio of resprouters may be interpreted just as a consequence of the existence of deep roots, despite the weak relationships found within the PCA diagram. Studies from other Mediterranean-type regions of the World refer relationships between belowground traits and regenerative strategies similar to those reported here, namely in Australia (Bell, 2001), California (Hellmers et al., 1955) and South Africa (Higgins et al., 1987). The decreasing trend of the R/S ratio and SRL for increasing values of basal section is to be expected. An increasing allocation of resources to the shoot as plants grow has been referred by different authors (Nobel, 1996; Bazzaz, 1997; Grace, 1997) and it is specially important in mediterranean plants for which the early development of roots may be critical for surviving the first dry season after germination. The expected decreasing trend observed for SRL (Fitter, 1985) was much more marked for the two obligate seeder species, which may hint at different strategies of root growth. Apparently, differences are due to the preferential investment on structural roots and storage tissues of the two resprouters at early developmental stages, contrasting with the preferential investment on fine roots of the two obligate seeders at similar developmental stages. Despite the existence of a broad range of R/S ratios corresponding to different species and different developmental stages, root biomass and shoot biomass were similarly and consistently correlated with basal section. This was confirmed by both the linear regressions and the PCA. The existence of close relationships between stem variables (diameter and/or basal section) and root biomass, has been found in different studies on tree species ( e.g. Santantonio et al., 1977; Drexhage & Colin, 2001; Hoffmann & Usoltsev, 2001; Ranger & Gelhaye, 2001). The establishment of a single consistent allometric relationship for different species between root biomass and stem variables (Santantonio et al., 1977), is in accordance with the widespread pipe model

40 theory (Shinozaki et al., 1964). Stem section seems to limit root and shoot development in a similar way for different species and different developmental stages. Given the different sample sizes of each developmental stage and each species, it must be pointed out that the allometric relationships found are only valid within the scope of the present work. Sampling limitations have also to be kept in mind when interpreting the differences obtained between regenerative strategies. The similarity of belowground traits among obligate seeders certainly reflected the phylogenetic relationships between plants since only two genera could be sampled. Thus, the existence of common traits can be associated to a common regenerative strategy, or just simply to the fact that part of the plants were closely related in terms of phylogeny. Special sampling designs have been established to overcome this difficulty (Nicotra et al., 2002) but they are not always applicable within a single plant community as in our case. Therefore, the significant differences found between the two regenerative groups have to be interpreted also in view of the role of phylogeny. However our results are supported by similar findings (Hellmers et al., 1955; Higgins et al., 1987; Bell, 2001) in other Mediterranean-type regions, which lead us to assume that the same adaptive mechanisms existing elsewhere are also valid in the plant community where this study was undertaken. One of the aspects lacking in our study was the assessment of the influence of site conditions on belowground traits. Most authors agree on the fundamental influence of environmental factors on root systems characteristics (e.g. Spurr & Barnes, 1980; Kummerow, 1981; Fitter, 1996; Atkinson, 2000). In our case the absence of evident soil constraints for the development of roots and the proximity of the excavation sites, lead us to assume the existence of relatively homogeneous environmental conditions allowing the comparison of root systems on a genetic and an ontogenic basis. Nevertheless we have to admit that the existence of differences in plant density or micro-relief may have definitely contributed to the overall variance of belowground traits.

References

Atkinson D. (2000). Root characteristics: Why and what to measure. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer Verlag, Berlin, 1-32 pp. Bazzaz F.A. (1997). Allocation of resources in plants: state of the science and critical questions. In Bazzaz F.A. & Grace J. (eds.), Plant Resource Allocation. Academic Press, San Diego, 1-37 pp. Bell D.T. (2001). Ecological response syndromes in the flora of southwestern Western Australia: Fire resprouters versus reseeders. The Botanical Review 67: 417-440.

41

Bengough A.G., Castrignano A., Pagès L. & Van Noordwijk M. (2000). Sampling strategies, scaling and statistics. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S. & van de Geijn S.C. (eds.), Root methods, a handbook. Springer Verlag, Berlin, 147-173 pp. Böhm W. (1979). Methods of studying root systems. Springer Verlag, Berlin. Caldwell M.M. & Richards J.H. (1986). Competing root systems: morphology and models of absorption. In Givnish T.J. (ed.), On the Economy of Plant Life and Function. Cambridge University Press, New York, 251-273 pp. Caldwell M.M. & Virginia R.A. (1989). Root systems. In Pearcy R.W., Ehleringer J.R., Mooney H.A. & Rundel P. (eds.), Physiological plant ecology. Field methods and instrumentation. Chapman and Hall, London, 367-398 pp. Canadell J, Jackson R.B., Ehleringer J.R., Mooney H.A., Sala O.E. & Schulze E.-D. (1996). Maximum root depth of vegetation types at the global scale. Oecologia 108: 583-595. Canadell J. & Zedler P.H. (1995). Underground structures of woody plants in mediterranean ecosystems of Australia, California and Chile. In Arroyo M.T., Zedler P.H. & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecosystems in Chile, California and Australia. Springer Verlag, Berlin, 177-210 pp. Cannon W.A. (1949). A tentative classification of root systems. Ecology 30: 452-458. Castroviejo S. (ed.), (1999). Flora Iberica. Real Jardin Botánico, CSIC, Madrid. Clemente A., Rego F. & Correia O. (1996). Demographic patterns and productivity of post-fire regeneration in Portuguese mediterranean maquis. International Journal of Wildland Fire 6: 5- 12. Correia O. & Catarino F. (1994). Seasonal changes in soil-to-leaf resistance in Cistus sp. and Pistacia lentiscus. Acta Oecologica 15: 289-300. Correia O. (1988). Contribuição da fenologia e ecofisiologia em estudos da sucessão e dinâmica da vegetação mediterrânica. PhD. thesis, University of Lisbon. Drexhage M. & Colin F. (2001). Estimating root system biomass from breast-height-diameters. Forestry 74: 491-497. Fitter A.H. (1985). Functional significance of root morphology and root system architecture. In Fitter A.H., Atkinson D., Read D.J., Usher M.B. (eds.), Ecological interactions in soil. Special publication of the British Ecological Society, No. 4). Blackwell Scientific, Oxford, 87-106 pp. Fitter A.H. (1996). Characteristics and functions of root systems. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 1-20 pp. Grace J. (1997). Toward models of resource allocation by plants. In Bazzaz F.A. & Grace J. (eds.), Plant resource allocation. Academic Press, San Diego, 279-291. Hellmers H. Horton J.S., Juhren G. & O’Keefe J. (1955). Root systems of some chaparral plants in Southern California. Ecology 36: 667-678. Higgins K.B., Lamb A.J. & Van Wilgen B.W. (1987). Root systems of selected plant species in mesic mountain fynbos in the Jonkershoek Valley, South-Western Cape Province. South African Journal of Botany 53: 249-257. Hoffmann C. & Usoltsev V.A. (2001). Modelling root biomass distribution in Pinus sylvestris forests of the Turgai depression of Kazakhstan. Forest Ecology and Management 149: 103-114. Jackson R.B., Canadell J., Ehleringer J.R., Mooney H.A., Sala O.E. & Shulze E.D. (1996). A global analysis of root distributions for terrestrial biomes. Oecologia 108: 389-411. Keeley J.E. & Zedler P.H. (1978). Reproduction of chaparral shrubs after fire: a comparison of sprouting and seeding strategies. American Midland Naturalist 99: 142-161. Keeley J.E. (1986). Resilience of Mediterranean shrub communities to fires. In Dell A., Hopkins J.M. & Lamont B.B. (eds.), Resilience in mediterranean-type ecosystems. Dr. Junk Publishers, Dordrecht, 95-112 pp.

42

Kozlowski T.T., Kramer P.J. & Pallardy S.G. (1991). The physiological ecology of woody plants. Academic Press Inc., San Diego. Kummerow J. (1981). Structure of roots and root systems. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 269- 288 pp. Nepstad D.C., Carvalho C.R., Davidson E.A., Jipp P.H., Lefebvre P.A., Negreiros G.H., Silva E.D. da, Stone T.A., Trumbore S.E. & Vieira S. (1994). The role of deep roots in the hydrological and carbon cycles of amazonian forests and pastures. Nature 372: 666-669. Nicotra A., Babicka N. & Westoby M. (2001). Seedling root anatomy and morphology: an examination of ecological differentiation with rainfall using phylogenetically independent contrasts. Oecologia 130: 136-145. Nobel P.S. (1996). Ecophysiology of roots of desert plants, with special emphasis on Agaves and Cacti. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 823-844 pp. Pagès L., Asseng S., Pellerin S. & Diggle A. (2000). Modelling root system growth and architecture. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S. & van de Geijn S.C. (eds.), Root methods, a handbook. Springer Verlag, Berlin, 113-146 pp. Ranger J. & Gelhaye D. (2001). Belowground biomass and nutrient content in a 47-year-old Douglas-fir plantation. Annals of Forest Sciences 58: 423-430. Santantonio D., Hermann R.K. & Overton W.S. (1977). Root biomass studies in forest ecosystems. Pedobiologia 17: 1-31. Shinozaki K., Yoda K., Hozumi K. & Kira T. (1964). A quantitative analysis of plant form the pipe model theory II. Further evidence of the theory and its application in forest ecology. Japanese Journal of Ecology 14: 133-139. Silva J.S. & Rego F.C. (1998). Factors affecting the establishment of woody species after fire in Central Portugal. In Trabaud L. (ed.), Fire management and landscape ecology. International Association of Wildland Fire, Fairfield, 103-114 pp. Specht R.L. & Rayson P. (1957). Dark Island heath (ninety mile plain, South Australia). Australian Journal of Botany 5: 103-114. Spurr S.H. & Barnes B.V. (1980). Forest Ecology. Krieger Publishing Company, Malabar. Zeng X. (2001). Global vegetation root distribution for land modelling. Journal of Hydrometeorology 2: 525-530.

43

3 ROOT DISTRIBUTION OF MEDITERRANEAN WOODY PLANTS; INTRODUCING A NEW EMPIRICAL MODEL2 Abstract: The root distributions of 42 plants from 18 mediterranean woody species were characterised by using an adjustable new empirical model. Plants were classified in terms of developmental stage and regenerative strategy. Vertical distributions of root length and root biomass were determined, standardised, and computed as cumulative data. The root distribution data were used to test the new proposed model against three other models taken from the literature. Tests have shown that the new model presented the best performance among the four. Significant differences were obtained between developmental stages and between regenerative strategies, using model-derived parameters. Despite the heterogeneous characteristics within the group of studied plants, different rooting patterns could be detected using the new modelling approach. The paper discusses these patterns in terms of the ecological characteristics of the different species. Key words: Mediterranean woody plants, root biomass, root distribution models, root length, rooting patterns.

2Based on paper: Silva J.S., Rego F.C. & Martins-Loução M.A. (2003). Root distribution of mediterranean woody plants. Introducing a new empirical model. Plant Biosystems 137(1) in press.

44

3.1 Introduction The different rooting patterns that exist in nature are the result of both genetically and environmentally determined characteristics. As a consequence, root systems may show a wide variation both between and within species, depending on site conditions, the presence of neighbouring plants, developmental stage and genotype (Canadell & Zedler, 1995; Fitter, 1996). However it is recognised that different species are associated with different rooting patterns (Spurr & Barnes, 1980) and that these have a direct influence on the distribution of roots in the soil (Fitter, 1996). Studies on root distribution are normally concerned with root biomass and/or with root length, as a function of depth (Lynch, 1995). In physiological terms, root biomass is a measure of the role of roots as sinks, whereas root length is a direct indicator of the potential for the absorption of nutrients and water (Atkinson, 2000). Given the importance of these two root variables, most plant community models that quantify the uptake of water and solutes or the carbon partitioning within the plant, use root biomass and/or root length distributions as an input (e. g. Caldwell & Richards, 1986; Klepper & Rickman, 1990; Shani & Dudley, 1996). In this way, most root distribution models were developed for plant communities but only very few were designed for plant individuals (Hoffmann & Usoltsev, 2001). However, the high diversity of root systems that exists in natural conditions should also favour a single plant approach (Bengough et al., 2000). Models at the individual plant level may allow the study of the factors that determine root distribution patterns, just because each individual is a direct unique result of those factors. Some of the simplest root distribution models were developed to calculate the distribution of roots with depth under non-limiting growing conditions (Pagès et al., 2000). The model by Monteith (1989) defines root length density (the length of roots per volume of soil) as a simple inverse square root function of depth. Other models can be adjusted to a particular species or plant community by changing one single parameter, such as those developed by Gerwitz & Page (1974) and Gale & Grigal (1987). A similar approach was also used to describe the horizontal biomass decay with distance from the stem base of individual trees (Drexhage & Gruber, 1998). More flexible models have been adjusted to root distributions at the community level such as the logistic dose-response curve (LDR) which includes two parameters (Schenk & Jackson, 2002). The parameterisation obtained by fitting a model to a certain root distribution allows a simple characterisation of the respective rooting pattern.

45

In the specific case of mediterranean woody plants there is evidence of a relationship between rooting patterns and regenerative strategies. Regenerative strategies are very important for mediterranean shrublands since they assure their resilience to frequent disturbances such as fire or grazing (Naveh, 1975). Studies have shown that resprouting species have extensive root systems and are often deep rooted whereas seeder species have less developed root systems and are normally shallow rooted (Keeley, 1986; Bell, 2001). The adaptive strategies of plants in terms of root system characteristics have also been related to different patterns of root development. The fact that many species, especially those from dry conditions, may develop deep tap roots at early stages preceding the development of lateral roots (Spurr & Barnes, 1980) is an evidence of the specificity of root distributions at different developmental stages. Few efforts have been made to characterise in a quantitative manner the complete root systems of woody plants. Most works dealing with root systems have been developed on agricultural plants and/or plants at the seedling stage, given the considerable efforts involving the excavation of roots in the field. This paper aims to characterise the root distributions of a set of mediterranean woody plants using a new modelling approach. The dual objective of the present paper is to: a) compare the performance of a new empirical model with other known models, at fitting root distribution data; b) use the proposed modelling approach to describe and interpret the root distributions of different mediterranean woody plants according to the respective species, regenerative strategy and developmental stage.

3.2 Methods

Study regions The study site was at Tapada Nacional de Mafra in the Central West Region of Portugal, where 34 plants were excavated. Tapada Nacional de Mafra is a protected area with 827 ha, about 30 km Northwest of Lisbon and 12 km East from the coast (38º 58’ 30’’ N and 9º 15’ 52’’ W). The lowest altitude is 90 m and the highest is 358 m. Soils are humic cambissols derived from sandstone. Mean annual precipitation is 798 mm and mean annual temperature is 14.6 ºC. Summer precipitation (June, July, August) accounts for only 3.1 % of total annual rainfall. The area is mostly covered by shrublands and dominant species are Erica scoparia L. and Erica lusitanica Rudolphi.

46

Other important species are Crataegus monogyna Jacq., Ulex Ulex jussiaei Webb, Daphne gnidium L., Pistacia lentiscus L., Myrtus communis L. and Rubus ulmifolius Schott. On more xeric sites and open areas we also find Lavandula luisieri (Rozeira) Rivas Martinez and different species of Cistus. In order to increase the data set used for testing the root distribution models, we have also included unpublished data concerning 8 plants collected at Serra da Malcata in a previous work. Serra da Malcata is a natural reserve located in the Central East Region of Portugal close to the Spanish border (40o 19’ N and 7o 09’ W). Soils are mostly schist lithossols. The annual average rainfall is 812 mm and mean annual temperature is 11.8 ºC. Within the Reserve boundaries the lowest altitude is 425 m and the highest is 1078 m. Vegetation is essentially constituted by shrublands dominated by Chamaespartium tridentatum (L.) P. Gibbs, Erica australis L., Erica umbellata L., Cytisus multiflorus (L’Hér.) Sweet, Cytisus striatus (Hill) Rothm. and Cistus ladanifer L.

Sampling procedures The general sampling objective was to obtain a broad set of woody species typical from the studied regions including different developmental stages. We have preferred to collect few individuals from many species instead of replications of a few species because of the exploratory nature of the research task supporting this study. A total of 42 plants from 18 species were studied. Plants from Mafra were hydraulically excavated (Böhm, 1979), using a pump from a fire truck and an adjustable nozzle, whereas in Malcata only hand tools were used. In both cases the root systems of all plants were integrally excavated. Much care was taken in order to minimise the loss of fine roots. Plant stems were wired in order to hold root systems in their original position and photos were taken before the complete soil removal. The vertical distance from the soil surface to the tip of the furthest root was measured in order to determine the maximum rooting depth. Basal section was used as an indicator of the developmental stage of the plants. Plants were assigned to three different classes of basal section (the cross sectional area at the stem base). The class limits were defined according to BS=π.r2 which is the formula for calculating the basal section (BS), using an increasing stem radius r: BS1 ≤ 12.6 mm2 (13 plants), 12.6 mm2< BS2 ≤ 50.3 mm2 (15 plants), BS3 > 50.3 mm2 (14

47 plants). Plants were also classified according to the respective regenerative strategy into obligate seeders (5 species) and resprouters (13 species). After excavation, the root systems of Mafra were photographed and dry weighted. Photos were taken using a Fugifilm MX 2900 Zoom digital camera (Fuji Photo Film Co., Ltd., Tokyo), featuring a maximum resolution of 2.3 million pixels. Digital images were analysed using software WinRhizo 4.1b (Regent Instruments, Quebéc). Module “Root distribution”, was used to measure the root length of each plant within 4 cm deep horizontal sample layers separated by 1 cm (the minimum value accepted by the program), corresponding to approximately 80% of the total root length. In the case of plants from Malcata the procedures were similar for root biomass but root length was not measured. Biomass distribution was determined for individuals from both regions, by cutting plants in 5 cm sections. Each section was limited by two lines perpendicular to the vertical axe of each plant. Biomass was oven-dried at 85º C for 48 hours and weighted.

Data analysis For each layer, root length and root biomass were standardised as a fraction of total root length and total root biomass respectively, assuming values between 0 and 1. Using the standardised data, the cumulative root biomass and cumulative root length were computed from the surface layer to the maximum rooting depth. Cumulative root length and cumulative root biomass are thus non-dimensional variables and are referred to as cumulative root fraction (Yr), assuming values of 0 at the soil surface and 1 at the maximum rooting depth. Yr=0.5 corresponds to half of the cumulative root fraction and the corresponding depth was used as a root distribution parameter identified as d50.

When referring to root biomass, this parameter was identified as db50 and when referring to root length it was identified as dl50. Two models used for representing the cumulative distribution of roots as a function of depth (d) at the community level, were adjusted to the data: the Gale & Grigal model (Gale & Grigal, 1987; Jackson et al., 1996; Schenk & Jackson, 2002),

d Yr = 1− β (1) and the logistic dose-response curve (LDR), (Schenk & Jackson, 2002).

48

1 Yr = c (2) ⎛ d ⎞ 1+ ⎜ ⎟ ⎝ d 50 ⎠ Symbol β is an adjustable parameter in the Gale & Grigal model. It assumes lower values for higher concentrations of roots at surface layers and values closer to 1 for more homogeneous root distributions. The parameter d50 can be determined as d50=logβ0.5. In the LDR model symbol c is a dimensionless curve shape parameter and d50 is obtained directly from the equation. We have also tested a standard logistic model. This model was used by Rego et al. (1994) for representing the cumulative biomass distribution of the canopy of single plants: 1 Y = (3) r 1+ e a+bd Symbols a and b are both adjustable parameters depending on the curve shape.

The value of d50 is given by d50=–a/b (Rego et al., 1994). Due to the limitations shown by these functions, namely in what concerns the adjustment to data at the beginning and at the end of the biomass and root length distributions, we have developed a modification of the LDR model, hereafter identified as MLDR: 1 Yr = c (4) ⎛ Maxd − d ⎞ 1+ ⎜ ⎟ ⎝ d.D ⎠ where c and D are adjustable parameters and Maxd is the actual maximum rooting depth. In this model d50 can be obtained as d50=Maxd/(D+1). According to this definition, D >1 corresponds to distributions where Yr=0.5 is attained above Maxd/2, whereas D <1 corresponds to distributions where Yr=0.5 is attained below Maxd/2. The performance of the four models was evaluated by non-linear regression by computing the percentage of variation explained by each model (r2) when fitted to root distribution data. Since data from cumulative distributions are not independent, the high values obtained for the r2 have to be interpreted having this fact in mind. Although the significance of the adjustments can not be estimated as for independent data, the value of r2 can still be used for comparison purposes, in order to determine which model fits best a given data set. The distribution of the r2 did not follow a normal distribution as shown by the Shapiro-Wilk W test. According to this, the Friedman Anova was used to

49 compare the values of r2 obtained with the four models, complemented for paired comparisons by the Mann-Whitney U test (Sokal & Rohlf, 1995). The results were used to choose the best model among the four under test.

Values of db50 and dl50 were computed for all plants from Mafra and Malcata using the chosen model. The Mann-Whitney U test was used to compare the db50 and dl50 values of each regenerative strategy and each basal section class. Since species were sampled unequally, paired comparisons were performed using the mean values of each species. Different rooting patterns were analysed by plotting the model fitted to the root length and root biomass data of nine plants from Mafra, each one corresponding to the highest basal section of the respective species. In this case the depth scale was standardised for comparing the different plants, thus allowing the analysis of relative root distributions. 3.3 Results

Testing the models The Friedman Anova (Table 3.1) showed that the values of r2 of the four models were significantly different for both root length (p<0.001; n=34) and root biomass (p<0.001; n=42). The logistic model presented the highest average rank for root biomass (3.40; mean r2=0.98) whereas the MLDR presented the highest rank for root length (3.76; mean r2=0.99). The comparisons between the logistic and the MLDR models showed that only the results obtained for root length data set were significantly different (p<0.001) as shown by the Mann-Whitney U test. According to the significance of the tests we concluded that the MLDR model should be preferred to the other three models given that a significantly higher percentage of variation could be explained by the MLDR model. This decision has affected all further steps of the study since only the MLDR was used to characterise root biomass and root length distributions.

Root distributions

Within the whole set of plants the values of Maxd, db50 and dl50 reflected the diversity of root distribution patterns (Table 3.2 and Table 3.3).

Table 3.1 Comparison of average ranks (1 to 4) and mean r2 of four models fitted to root biomass and root length data, as obtained by the Friedman Anova (p<0.001 for both root biomass and root length

50 distributions). Values of mean r2 followed by the same letter did not present significant differences (p>0.05) as obtained by paired comparisons using the Mann-Whitney U test.

Model Root biomass Root length Average rank Mean r2 Average rank Mean r2 Gale and Grigal 1.65 0.86 a 1.38 0.84 a LDR 1.70 0.93 a 2.08 0.95 b Logistic 3.40 0.98 b 2.77 0.97 c MLDR 3.26 0.98 b 3.76 0.99 d

Table 3.2 Maximum rooting depth averaged by species and basal section class. Maximum rooting depth represents the depth achieved by the deepest root. Values are averages and the number of plants is shown in brackets. The range was obtained as the difference between the maximum and the minimum values observed. Basal section classes are defined as: BS1 ≤ 12.6 mm2, 12.6 mm2< BS2 ≤ 50.3 mm2, BS3 > 50.3 mm2.

Species Regenerat. Region Maximum rooting depth (cm) strategy BS1 BS2 BS3 Range Arbutus unedo respr. Malcata − − 301(2) 30 Chamaespartium tridentatum respr. Malcata − 401(1) − 0 Cistus crispus seeder Mafra 251(4) 401(4) − 39 Cistus salvifolius seeder Mafra 251(2) 401(1) − 15 Crataegus monogyna respr. Mafra 261(1) − 651(2) 47 Cytisus multiflorus seeder Malcata − − 601(1) 0 Cytisus striatus seeder Malcata − − 351(1) 0 Daphne gnidium respr. Mafra 751(2) − 1651(2) 170 Erica australis respr. Malcata − − 351(1) 0 Erica lusitanica respr. Mafra − − 1601(1) 0 Erica scoparia respr. Mafra − 401(1) 1401(1) 100 Lavandula luisieri seeder Mafra 261(2) 261(1) 351(1) 10 Myrtus communis respr. Mafra − 1201(1) − 0 Pistacia lentiscus respr. Mafra − 501(1) − 0 Quercus faginea respr. Mafra 551(1) − − 0 Quercus pyrenaica respr. Malcata − − 601(2) 10 Rubus ulmifolius respr. Mafra − 661(3) − 51 Ulex jussiaei respr. Mafra 401(1) 821(2) − 87 All species 36(13) 56(15) 79(14) 170

51

Table 3.3 Values of db50 and dl50 averaged by species and by basal section (BS) class. Basal section classes are defined as: BS1 ≤ 12.6 mm2, 12.6 mm2< BS2 ≤ 50.3 mm2, BS3 > 50.3 mm2. The range was obtained as the difference between the maximum and the minimum values observed within all plants from each species.

Species db50 (cm) dl50 (cm) BS1 BS2 BS3 Range BS1 BS2 BS3 Range Arbutus unedo − − 1.7 2.7 − − − − Chamaespartium tridentatum − 11.0 − 0.0 − − − − Cistus crispus 3.3 4.9 − 5.6 12.1 19.0 − 13.7 Cistus salvifolius 2.8 3.7 − 1.5 10.0 15.9 − 6.6 Crataegus monogyna 1.9 − 9.7 9.5 17.2 − 37.1 33.2 Cytisus multiflorus − − 9.0 0.0 − − − − Cytisus striatus − − 3.4 0.0 − − − − Daphne gnidium 4.0 − 6.9 4.7 28.0 − 58.2 76.5 Erica australis − − 5.8 0.0 − − − − Erica lusitanica − − 3.3 0.0 − − 31.9 0.0 Erica scoparia − 4.6 0.5 4.1 − 21.4 43.3 21.9 Lavandula luisieri 3.3 1.9 3.9 2.2 7.9 19.1 26.0 18.4 Myrtus communis − 5.0 − 0.0 − 26.3 − 0.0 Pistacia lentiscus − 2.9 − 0.0 − 22.7 − 0.0 Quercus faginea 7.1 − − 0.0 24.6 − − 0.0 Quercus pyrenaica − − 10.8 0.2 − − − − Rubus ulmifolius − 7.5 − 12.2 − 23.4 − 32.0 Ulex jussiaei 7.6 15.8 − 20.8 17.9 40.9 − 46.2 All species 4.3 6.4 5.5 25.8 16.8 23.6 39.3 76.5

The highest values of Maxd were observed for D. gnidium (185 cm), E lusitanica (160 cm) and E. scoparia (140 cm), at BS3 and the lowest values of Maxd were observed for C. crispus (16 cm), C. salvifolius (16 cm) and D. gnidium (15 cm) at

BS1. The lowest value of db50 was observed for A. unedo (0.4 cm) at BS3 and the highest for U. jussiaei (26.2 cm) at BS2. The lowest value of dl50 was observed for C. salvifolius (3.2 cm) at BS1 and the highest for D. gnidium (79.9 cm) at BS3. Tests between regenerative strategies including all plants, have shown that resprouters had significantly higher values of dl50 than obligate seeders (p<0.01) but no significant results were obtained for db50. When restricting the tests to plants from the first two BS classes, tests have shown that resprouters had significantly higher values of db50

(p<0.05) and dl50 (p<0.01).

52

Tests between BS classes have shown that BS3 had significantly higher values of dl50 than BS2 (p<0.05) and BS1 (p<0.01). The comparison between BS1 and BS2 for dl50 was not significant. None of the comparisons between BS classes for db50 was statistically significant. The cumulative root distributions of nine plants plotted using a standardised depth scale (0-100 % of Maxd) are shown in Fig. 3.1 together with the MLDR model fitted to the data. The corresponding root system images are shown in Fig. 3.2.

Fig. 3.1 Root distribution of nine plants excavated at Tapada Nacional de Mafra as represented by the fitted MLDR model (see text and equation 4). D and c are the model parameters. The solid line represents the cumulative root biomass distribution and the dotted line represents the cumulative root length distribution.

53

C. crispus and C. salvifolius plants revealed a remarkable resemblance of root biomass (D = 9.06 and D = 9.93, respectively) and root length (D = 1.06 and D = 1.89, respectively) distributions. C. monogyna presented a considerable separation between the two curves showing distinct distribution patterns for biomass (D = 6.03) and length (D = 0.47). D. gnidium presented a considerable accumulation of root biomass close to the surface (D = 18.94) but a much deeper concentration of root length (D = 0.88).

Fig. 3.2 Root systems of nine plants excavated at Tapada Nacional de Mafra. The corresponding cumulative root distributions are shown on Fig. 3.1. The vertical bar represents 0.5 m.

54

E. lusitanica and E. scoparia plants showed very similar rooting patters with a very high accumulation of root biomass close to the surface (D = 48.06 and D = 266.87, respectively) and not very different root length distribution curves (D = 4.47 and D = 2.72, respectively). L. luisieri showed very different patterns for root biomass (D = 8.09) and root length (D = 0.39). M. communis showed a high accumulation of root biomass (D = 23.19) and root length (D = 4.19) at the surface layers. A similar result was obtained for R. ulmifolius, with the two curves almost overlapping (D = 13.13 for root biomass and D = 7.54 for root length). 3.4 Discussion

Methodology Among the four models tested, the MLDR showed a considerable flexibility and capacity to fit root distribution data from a broad range of woody plants. The use of this simple mathematical function allowed a very close description of root distribution patterns as obtained with the described sampling procedures. Also the joint use of root biomass and root length coefficients reflected the diversity of root systems present within the data set. Although this approach could potentially lead to an empirical classification of root systems by establishing class limits for the two parameters, such an attempt was not followed here given the need of a much broader root distribution database. A comprehensive classification would only be possible with replications for each species, each developmental stage and different environmental conditions. The enormous amount of work required for the excavation of woody plants in the field was a serious obstacle to obtain such a complete set of data. If the MLDR model is to be used to fit root distributions of plant communities where the maximum rooting depth is often unknown, it should be taken into account the fact that the deepest root data always correspond to Yr=1. If this condition is considered and accepted, the use of the MLDR model for fitting root distribution data at the community level is perfectly possible. Silva & Rego (unpublished) have fitted the MLDR model to the vertical distribution of root counts from trench profiles, obtaining similar results in terms of the explained variance, as those obtained in the present study. Thus there is a potential use of the MLDR model to work as an input in functional models both at the community as at the individual plant level. Some remarks have to be made on the methods used to obtain root distribution data specially in what concerns root length. Root length distribution was affected to an

55 unknown degree by the methodology used to excavate the individual root systems, given the inevitable loss of fine roots (Wallace et al., 1974; Böhm, 1979; Caldwell & Virginia, 1989; Bengough et al., 2000) which are responsible for most of the total root length. Although we may assume an equal effect along the whole root system, the loss of fine roots during the excavation works has eventually contributed to hide some of the normal accumulation of root length at surface layers. Another aspect to take into account is the possibility of changes in the plants root system structure after excavation, compared to the original root arrangement in undisturbed soil, with consequences on the overall root distribution. However, our perception of the field work says that very little changes are introduced when dealing with coarse structural roots of woody plants. Possible deformations on the arrangement of thinner roots were prevented by a careful observation and image registration of the plants in the field and with a careful handling and disposal for biomass and root length determination. Finally we should also mention the existence of possible errors associated with the process of image taking, processing and analysis. However, deviations on root length measurements may be either in the sense of diminishing or increasing the actual values (Richner et al., 2000). Despite the obvious limitations of the methodology used for measuring the length of roots of woody plants, we have assumed that the referred errors were similarly distributed among all plants and within each plant. Moreover, the use of standardised data and the comparative nature of the present study are reasons for considering the possible errors as acceptable in view of the goals pursued.

Root distribution patterns Root distribution data revealed in particular for adult plants, a separation between seeders and resprouters. Within the resprouters group, plants with deep roots showed a distinct root distribution pattern (Daphne, Erica, Myrtus, Ulex). In particular the lignotuberous species E. lusitanica and E. scoparia presented a high biomass concentration close to the soil surface which accounts for the importance of the lignotuber on biomass partitioning in these plants. Some authors question the decision of considering the lignotuber as part of the root system since it works as a storage organ for starch, normally located very close to the soil surface or even at the stem base (Kummerow, 1981; Canadell & Zedler, 1995), and not exactly as a structural root. This decision is obviously critical for the resulting root distribution pattern. Two other resprouting species with a relatively high concentration of biomass close to the surface,

56

D. gnidium and M. communis, did not present typical woody lignotubers but similar swollen-less-lignified root crowns, which apparently play the same role of storing resources and dormant buds for plant resprouting. With the exception of M. communis, all deep rooted species presented nearly vertical tap roots. Tap roots although having a strategic role on plant survival during the dry season, they often have little importance both in terms of length and biomass (Kummerow et al., 1990; Nepstad et al., 1994; Canadell & Zedler, 1995; Canadell et al., 1996). The implication of this rooting pattern is a higher relative concentration of roots close to the surface (as can be observed in Fig. 3.1) for increasing maximum rooting depths. Most deep rooted plants presented also a lateral expansion of roots close to the surface in what can be considered as a dual root system (Canadell & Zedler, 1995) composed of tap roots and surface roots. Different studies have shown that in drought-prone regions this rooting pattern may be advantageous and is present in several species (Hellmers et al., 1955; Specht & Rayson, 1957; Kummerow,1981; Krämer et al., 1996). In the particular case of the two phylogenetically close Erica species (Castroviejo (1999) despite the similar root biomass distribution pattern, a different root length distribution pattern was observed due to the less pronounced lateral expansion of the root system of E. scoparia (at BS3). This may be explained by the fact that this individual was part of a dense maquis whereas the E. lusitanica plant was an isolated individual. Although we could not test this hypothesis, it is known that plant interaction may have a determinant effect on root distribution (Atkinson, 2000). The resprouters C. monogyna and R. ulmifolius followed distinct strategies when compared to the remaining species of this group. The first presented only thick structural roots at the surface and an almost complete absence of fine roots at this level. This rooting pattern explains the difference between the root biomass and the root length curves observed in Fig. 3.1. In the case of R. ulmifolius this species only presented fine roots both at the surface as at deeper layers. This explains the similarity of the root biomass and root length curves. The fact of being a climber species explains the absence of structural roots and consequently the rooting patterns exhibited by this species. In the case of seeders there was a remarkable resemblance of rooting patterns of C. crispus and C. salvifolius plants, corresponding to shallow root systems as it is common of most seeder species (Keeley, 1986; Bell, 2001). In particular the case of L luisieri is paradigmatic since this highly specialised seeder manages to survive in very dry conditions with a shallow root system. The strategy of this species, seems to be the

57 development of a fibrous root system at the maximum rooting depth, with the consequent increase in water uptake efficiency. In terms of developmental stages, only root length distribution seemed to be particularly affected as a consequence of increasing maximum rooting depth. The absence of consistent differences of root biomass distribution among BS classes was partially due to the concentration of biomass in storage surface organs by resprouters.

On the contrary, the values of dl50 showed a consistent increase for all species along developmental stages, as a consequence of increasing values of maximum rooting depth. The present work showed that the parameterisation of root distributions using an adjustable model such as the MLDR may be an interesting way of characterising species and functional groups according to their adaptation to environmental conditions. Further research is need in what concerns the study of the intra-specific variability of individual root distributions according to different environmental conditions, for which the approach presented here may also be used.

References

Atkinson D. (2000). Root characteristics: Why and what to measure. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 1-32 pp. Bell D.T. (2001). Ecological response syndromes in the flora of southwestern Western Australia: Fire resprouters versus reseeders. The Botanical Review 67: 417-440. Bengough A.G., Castrignano A., Pagès L. & Van Noordwijk M. (2000). Sampling strategies, scaling and statistics. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S. & van de Geijn S.C. (eds.), Root methods, a handbook. Springer Verlag, Berlin, 147-173 pp. Böhm W. (1979). Methods of studying root systems. Springer Verlag, Berlin. Caldwell M.M. & Richards J.H. (1986). Competing root systems: morphology and models of absorption. In Givnish T.J. (ed.), On the economy of plant life and function Cambridge University Press, Cambridge, 251-273 pp. Caldwell M.M. & Virginia R.A. (1989). Root systems. In Pearcy R.W., Ehleringer J.R., Mooney H.A. & Rundel P. (eds.), Physiological plant ecology. Field methods and instrumentation. Chapman and Hall, London, 367-398 pp. Canadell J. & Zedler P.H. (1995). Underground structures of woody plants in mediterranean ecosystems of Australia, California and Chile. In Arroyo M.T., Zedler P.H. & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecosystems in Chile, California and Australia. Springer Verlag, Berlin, 177-210 pp. Canadell J., Jackson R.B., Ehleringer J.R., Mooney H.A., Sala O.E. & Schulze E.-D. (1996). Maximum root depth of vegetation types at the global scale. Oecologia 108: 583-595. Castroviejo S. (ed.), (1999). Flora Iberica. Real Jardin Botánico, CSIC, Madrid. Drexhage M. & Gruber F. (1998). Architecture of the skeletal root system of 40-year-old Picea abies on strongly acidified soils in the Harz Mountains (Germany). Canadian Journal of Forest Research 28: 13-22.

58

Fitter A.H. (1996). Characteristics and functions of root systems. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 1-20 pp. Gale M.R. & Grigal D.F. (1987). Vertical root distribution of northern tree species in relation to successional status. Canadian Journal of Forest Research 17: 829-834. Gerwitz A. & Page E.R. (1974). An empirical mathematical model to describe plant root systems Journal of Applied Ecology 11: 773-781. Hellmers H. Horton J.S., Juhren G. & O’Keefe J. (1955). Root systems of some chaparral plants in Southern California. Ecology 36: 667-678. Hoffmann C. & Usoltsev V.A. (2001). Modelling root biomass distribution in Pinus sylvestris forests of the Turgai depression of Kazakhstan. Forest Ecology and Management 149: 103-114. Jackson R.B., Canadell J., Ehleringer J.R., Mooney H.A., Sala O.E. & Shulze E.D. (1996). A global analysis of root distributions for terrestrial biomes. Oecologia 108: 389-411. Keeley J.E. (1986). Resilience of mediterranean shrub communities to fires. In Dell A., Hopkins J.M. & Lamont B.B. (eds.), Resilience in mediterranean-type ecosystems. Dr. Junk Publishers, Dordrecht, 95-112 pp. Klepper B. & Rickman R.W. (1990). Modelling crop root growth and function. Advances in Agronomy 44: 113-132. Krämer S., Miller P.M. & Eddleman L.E. (1996). Root system morphology and development of seedling and juvenile Juniperus occidentalis. Forest Ecology and Management 86: 229-240. Kummerow J. (1981). Structure of roots and root systems. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 269- 288 pp. Kummerow J., Kummerow M. & Trabaud L. (1990). Root biomass, root distribution and the fine-root growth dynamics of Quercus coccifera L. in the garrigue of Southern France. Vegetatio 87: 37- 44. Lynch J. (1995). Root architecture and plant productivity. Plant Physiology 109: 7-13. Monteith J.L., Huda A.K.S. & Midya D. (1989). RESCAP: a resource capture model for sorghum and pearl millet. In Virmani S.M., Tandon H.L.S., Alagarswamy (eds.), Modelling the growth and development of sorghum and pearl millet. ICRISAT Research Bulletin 12: 30-34. Naveh Z. (1975). The evolutionary significance of fire in the Mediterranean region. Vegetatio 29: 199- 208. Nepstad D.C., Carvalho C.R., Davidson E.A., Jipp P.H., Lefebvre P.A., Negreiros G.H., Silva E.D. da, Stone T.A., Trumbore S.E. & Vieira S. (1994). The role of deep roots in the hydrological and carbon cycles of amazonian forests and pastures. Nature 372: 666-669. Pagès L., Asseng S., Pellerin S. & Diggle A. (2000). Modelling root system growth and architecture. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 113-146 pp. Rego F.C., Pereira J.R., Fernandes P. & Almeida A.F. (1994). Biomass and aerial strucuture characteristics of some mediterranean shrub species. Proceedings of the 2nd International Conference on Forest Fire Research. Coimbra, Portugal, 377-384 pp. Richner W., Liedgnens M., Bürgi H., Soldati A. & Stamp P. (2000). Root image analysis and interpretation. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 305-341 pp. Schenk H.J. & Jackson R.B. (2002). The global biogeography of roots. Ecological Monographs 72: 311- 328. Shani U. & Dudley L.M. (1996). Modelling water uptake by roots under water and salt stress: Soil-based and crop response root sink terms. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 1-20 pp. Sokal R.R. & Rohlf F.J. (1995). Biometry. W.D. Freeman and Company, New York.

59

Specht R.L. & Rayson P. (1957). Dark Island heath (ninety mile plain, South Australia). Australian Journal of Botany 5: 103-114. Spurr S.H. & Barnes B.V. (1980). Forest Ecolgy. Krieger Publishing Company, Malabar. Wallace A., Bamberg S.A. & Cha J.W. (1974). Quantitative studies of roots of perennial plants in the Mojave desert. Ecology 55: 1160-1162.

60

4 ROOT DISTRIBUTION OF A MEDITERRANEAN SHRUBLAND IN PORTUGAL3 Abstract: The distribution of roots of an Erica (Erica scoparia and Erica lusitanica) dominated mediterranean maquis was studied using three different approaches: root counts on trench walls (down to 120 cm), estimation of the maximum rooting depth using an allometric relationship and estimation of fine root biomass and fine root length using soil cores (down to 100 cm). Roots were classified according to diameter (fine, < 1.0 mm; small, 1.0-5.0 mm; medium, 5.1-10.0 mm; coarse, >10.0 mm) and species (Erica sp., Pteridium aquilinum, Rubus ulmifolius and Ulex jussiaei). The depth corresponding to 50% of all roots (D50) was determined by fitting a new model to the cumulative root distribution. Fine roots represented 96% of root counts. Root counts of Erica represented 59%, Ulex 34%, Rubus 6% and Pteridium 1%. Overall root counts showed a D50 of 26 cm. D50 was higher for Ulex (40 cm) and Erica (22 cm), and lower for Pteridium (9 cm) and Rubus (3 cm). D50 for fine roots was 27 cm, for small roots 11 cm, for medium roots 6 cm and for coarse roots 4 cm. The estimated average maximum rooting depth of the 28 deepest Erica roots was 222 cm. The deepest Erica root was estimated to reach 329 cm. 82% of roots growing deeper than 125 cm were not reaching more than 175 cm. Root length density ranged from 4.6 cm/cm3 at 10 cm to 0.8 cm/cm3 3 3 at 80 cm. Root biomass ranged from 7.7 mg/cm at 10 cm to 0.6 mg/cm at 40 cm. D50 for root biomass was 12 cm and D50 for root length was 14 cm. Fine root biomass was estimated as 1.6 kg/m2 and the respective root length as 18.7 km/m2. Key words: Root distribution, maximum rooting depth, root biomass, root length, mediterranean shrubland, Erica.

3 Submitted to Plant and Soil by J.S. Silva & F.C. Rego 2002.

61

4.1 Introduction The distribution of roots in the soil has a direct influence on the capability of plants to extract water and solutes (Fitter, 1996). Because of its vital role, information on root distribution is essential for a comprehensive understanding of the ecophysiology of plants. Root distribution data are frequently used as an input in water and nutrient uptake models (e.g. Caldwell & Richards, 1986; Monteith et al., 1989; Klepper & Rickman, 1990; Shani & Dudley, 1996; Pagès et al., 2000) or even in global scale simulation models (Zeng et al., 1998; Zeng, 2001). Models of root distribution have been developed normally under the form of a function which, in most cases, has adjustable parameters in order to suit the specificity of the crop or the plant community. Many root distribution models have been proposed in the literature in order to represent root distribution patterns (Gerwitz & Page, 1974; Gale & Grigal, 1987; Monteith et al., 1989; Drexhage & Gruber, 1998; Jobbágy & Jackson, 2000; Shenk & Jackson, 2002a). Similarly many different methodologies have been used to obtain root distribution data: root maps, soil cores, soil monoliths, complete root excavation. These different techniques allow to express the distribution of roots under different forms: root counts, root length, root biomass, root surface, root volume (Schuurman & Goedewaaggen, 1971; Böhm, 1979; Caldwell & Virginia, 1989; Atkinson, 2000). In the specific case of mediterranean ecosystems not many studies have been concerned with belowground processes and especially with the distribution of roots in the soil. Besides the normal difficulties encountered when working with roots, this lack of studies is also motivated by the existence of extensive root systems, the frequent existence of shallow bedrock or the intricate network of roots present in dense maquis- type shrublands (Kummerow, 1981; Canadell & Zedler, 1995). The relative scarcity of studies for this type of ecosystems is well reflected in the planetary compilation of root distribution data by Jackson et al. (1996) where mediterranean ecosystems are represented by only 11 (less than 5%) out of 250 studies and all the Mediterranean Region is represented by only 4 studies. Among the root distribution studies in the Mediterranean Region, the species Quercus ilex (Canadell & Rodà, 1991; Djema, 1995) and Quercus coccifera (Kummerow et al., 1990; Cañellas & Ayanz, 2000) have deserved special attention. It is also worth to mention the works of Martinez & Rodriguez (1988) and Martinez et al. (1998) in shrublands of Southern Spain. Few works have been focused on the root distribution of different species from the same plant community, thus sharing the same soil and climate conditions. In a

62 common environment, different rooting patterns should reflect different ecological adaptive strategies but also the competition between plants with different characteristics (Casper & Jackson, 1997). Another aspect lacking in most previous works is the comprehensive study of maximum rooting depth. This is critical for mediterranean conditions since deep roots are fundamental for many plant species to overcome the water stress typical of the dry season and also to be able to survive after a natural disturbance such as fire or grazing (Kummerow, 1981; Nepstad et al., 1994; Canadell & Zedler, 1995; Shenk & Jackson, 2002b). Moreover, in order to understand the uptake mechanisms of a community of plants, it is important to know the distribution of deep roots along the soil profile. Although many works refer values of the maximum rooting depth observed for a certain species, it is rare that studies approach the distribution of deep roots at the community level (Canadell et al., 1996). This is most of all explained by the tremendous difficulties to access this information, especially when dealing with deep rooted species. In order to contribute to a better knowledge of mediterranean ecosystems, the present paper approaches different aspects of the root distribution of a typical mediterranean maquis. In particular this study aims to: i) determine and compare the vertical root distribution of the different species and the different diameter classes, ii) estimate the maximum rooting depth and the distribution of deep roots, iii) determine the distribution of fine roots in terms of root length and root biomass. 4.2 Methods The study was conducted at Tapada Nacional de Mafra, a protected area with 1187 ha, about 30 km Northwest of Lisbon (38º 56’ 37’’ to 38º 58’ 30’’ N and 9º 15’ 52’’ to 9º 18’ 43’’ W). The studied plant community was largely dominated by Erica scoparia L. forming a dense maquis with an average plant height around 180 cm. Other species present were: Erica lusitanica Rudolphi, Rubus ulmifolius Schott, and Ulex jussiaei Webb. The only important herbaceous species was Pteridium aquilinum (L.) Kuhn. According to growth ring counts, the age of the oldest shrubs was estimated to be around 46 years, although we suppose that important disturbances may have occurred during this period. Mean annual precipitation is 798 mm and mean annual temperature is 14.6 ºC. The rare occurrence of rocks and the existence of a deep soil facilitate the natural development of roots. The methodology developed to study the roots of this plant community was based on the trench profile method complemented by the extraction of soil cores

63

(Schuurman & Goedewaagen, 1971; Böhm, 1979; Caldwell & Virginia, 1989; Van Noordwijk et al., 2000). The trench profile method allowed the extensive sampling of the shrub community in terms of root counts and maximum rooting depth, whereas the soil cores were specifically used to estimate the biomass and length of fine roots. Six trenches (numbered 1 to 6) were excavated with a backhoe equipment in June 1999. The site was chosen close to a forest road following a contour line on a 15% slope. The excavation was performed on the upper roadside and perpendicularly to the road axis. Trenches were separated by 8 m except trenches 3 and 4 which were separated by 27 m. Trenches were 3 m long x 1 m wide and were excavated down to the road level. The bottom of each trench was kept flat but depth in relation to soil surface varied among the trenches and within each trench according to the local relief characteristics. The minimum depth achieved was 120 cm and the maximum depth was 230 cm. Before the trench excavation, all plants inside the area to be excavated and 0.5 m on both sides were identified and the respective diameters at the stem base were measured. The fronds of Pteridium aquilium were measured exactly as the other species, despite its particular growing characteristics. For each trench we have disposed of the two side walls except for trench 5 where only one wall presented suitable conditions for root studies. Trench walls were also used to collect soil samples at different depths. Before root counting each wall surface was flattened and smoothed using hand tools and then referenced by means of a system of co-ordinates. On each prepared wall the basic procedure consisted on mapping the section of cut roots down to 120 cm deep on a transparent plastic sheet. Since parts of the trench walls have collapsed during and after excavation, the length of each root map was different according to the length of prepared wall. A total area of 16.4 m2 of soil profile was sampled using vertical maps. Cut roots were drawn on a 1:1 scale using different colours to distinguish the different species. Given the impossibility to distinguish in the field the roots of the two Erica species it was decided to register only the respective genus. This resulted in the differentiation of four categories: Erica (Erica scoparia and Erica lusitanica), Pteridium (Pteridium aquilinum), Rubus (Rubus ulmifolius) and Ulex (Ulex jussiaei), hereafter referred to as “species” for simplification purposes. Roots from Erica presented a black coloured bark and a light brown strongly lignified cortex, even in the smallest diameters. Roots from Pteridium were soft and brown coloured. Rhizomes presented hairs at the surface and a white cortex. The bark of Rubus roots was light brown becoming whitish coloured for finer roots. Ulex roots were white and poorly lignified, in all diameter classes, coarser roots presenting a rough

64 surface with a typical wrinkle-like texture. Specific root organs such as lignotubers from Erica and rhizomes from Pteridium were also included in the root mapping. No attempts were made to separate dead from live roots given the difficulties to distinguish these two categories in the field. However, obviously decaying roots and rhizomes were discarded from the root distribution counts. These criteria were consistent for all sampling procedures. After mapping, roots were counted over a 10x10 cm grid and registered according to species and diameter. Four diameter classes were considered: < 1.0 mm (fine roots), 1.0-5.0 mm (small roots), 5.1-10.0 mm (medium roots), >10.0 mm (coarse roots). The number of root counts of each grid square was used as a measure of root density (number of roots/dm2). The average root density by species and by diameter class was computed separately for each trench. The mean values for the plant community were computed as the weighted averages of the six trenches, using the sampled surface of each trench as the respective weight. Statistical comparisons were preceded by a Kolmogorov-Smirnov test in order to check the normal distribution of the variables under study. As most of them did not verify the normality assumptions, the existence of statistical differences in root density among species and among diameter classes was tested as paired comparisons by the Mann-Whitney U test. The distribution of roots along the soil profile was separately computed for each species and each diameter class following the procedure just described for the average root densities. In order to compare the different root distribution patterns, a two parameter logistic-type function specifically developed with this purpose, was fitted to the cumulative root fraction (Yr) as a function of depth (d). The cumulative root fraction was obtained by dividing the root density at each depth by the sum of root densities of the profile and then computing the cumulative series of these values for all depths. This cumulative standardised root distribution is equal to 0 at the soil surface and is equal to 1 at the maximum depth of the profile. The fitted model (hereafter referred to as MLDR) takes the form: 1 Yr = c (1) ⎛ Maxd − d ⎞ 1+ ⎜ ⎟ ⎝ d.D ⎠ where c and D are the model parameters and Maxd is the maximum depth of the studied profile (120 cm). The depth corresponding to 50% of the cumulative root fraction (Yr = 0.5) is given by:

65

Maxd D = (2) 50 (D +1)

High values of D50 are associated to deep root distributions whereas low values of D50 are associated to a higher concentration of roots close to the soil surface. Maximum rooting depth of deep rooted plants was estimated by mapping the cut roots at the bottom of the trenches according to a procedure similar to the one used on the trench walls. A total area of 9.9 m2 was sampled using this technique. In this case the diameters of cut roots were measured instead of being assigned to classes. The possible existence of a relationship correlating the diameter of cut roots and the respective maximum rooting depth was on the basis of this procedure. This relationship was determined by measuring the diameter of tap roots from excavated plants at regular intervals and the respective vertical distance to the root tip. This procedure could only be used for Erica since no adult Ulex plants (also deep rooted) could be excavated at the study site. Tap roots from three individuals (two E. scoparia and one E. lusitanica) were measured with a digital calliper every 5 cm, starting from the root tip and continuing upwards. An asymptotic equation of the form: a L = c (3) ⎛ R ⎞ 1+ ⎜ d ⎟ ⎝ b ⎠ where a, b and c are constants, Rd is the root diameter and L is the correspondent vertical distance to the root tip, was fitted to the observed data (n = 63). The non-linear regression revealed a consistent relationship between Rd and L, being able to explain 89% of the variance (Fig. 4.1). Estimations of maximum rooting depth using this relationship were made under the assumptions that: Erica tap roots were in general following an orientation similar to the sampled roots (4.1±0.4 degrees from the vertical), and that no solid sandstone (found on 8% of the mapped surface) was preventing roots to grow following their natural orientation. Both assumptions were reasonably confirmed by field observations of partially excavated individuals and road cuts in the study area. The estimation of maximum rooting depth was computed separately for each individual root by summing L to the depth at which the root was cut (i.e. the trench depth at that point).

66

160

140

120

100

80

60

40 Vertical distance to root tip (cm) −2.5 20 y = 144.3 []1+ ()x 1.8 r 2 = 0.89 0 24681012 Root diameter (mm)

Fig. 4.1 Relationship between root diameter and the vertical distance to the root tip for Erica

Roots were assigned to 25 cm classes of maximum rooting depth in order to determine the distribution of deep roots at the plant community level. Another feature which has deserved some attention was the spatial distribution of deep roots at the bottom of the trenches. With this purpose a chi-square goodness-of-fit test for randomness was performed separately on each of the six trenches, complemented by the computation of dispersion coefficients (Sokal & Rohlf, 1995). In order to complement the information given by the vertical root maps, core samples were extracted from two trenches (trenches 2 and 5). On each trench, 5 vertical transects were established, each consisting of 7 depths (10, 20, 30, 40, 60, 80, and 100 cm) for core extraction. This technique was used specifically to sample fine roots, for which a small soil core of 68,7 cm3 was extracted using a metal ring (inner diameter of 5.4 cm) at every sampling depth. In the laboratory soil was removed using water and roots were retained by a fine mesh sieve. Root dry weight was obtained after oven- drying the roots for 48 hours at 85 º C. Root length was obtained by counting the intersections of the roots disposed on a grid of 1 cm x 1 cm using the line intercept method (Marsh, 1971; Tennant, 1975). In this case no distinction was made between species. In order to obtain an estimation of biomass and root length per unit area of soil surface, a simple exponential model of type: y = a.d b (4)

67

was adjusted to the root distribution data. In this model y is the root variable (biomass or length) at a certain depth, a and b are the model parameters and d is depth. The model was adjusted to the means of 10 transects (r2 = 0.92 for root biomass and r2 = 0.97 for root length). Each model was integrated according to 10 cm increments from 0 to the estimated maximum rooting depth. This allowed to extrapolate the values of root density beyond 100 cm, which was the depth limit for actual data. Similarly to the procedure described for the distribution of root counts, an estimation of D50 was also computed for root biomass and root length. In order to detect differences between the two methodology approaches, values of D50 from soil cores were statistically compared with values of D50 of root counts (only fine roots) from the same trenches. 4.3 Results

Soil characterisation Organic matter decreased exponentially (Table 4.1) with depth, showing very low values at 120 cm. The average thickness of the A horizon (including A1 and A2) ranged from 39 cm (trench 6) to 102 cm (trench 4) with an overall value of 67.2±9.4 cm. Trenches 4 and 5 were presenting a thick organic layer due to relatively poor drainage conditions at these spots. Values of pH increased with depth revealing a decrease in exchangeable cations. According to an analysis performed on non-replicated samples (from trench 6), the sum of exchangeable cations (Ca2+, Mg2+, Na+ and K+) ranged between 7.87 cmol/ kg at 10 cm to 5.9 cmol/ kg at 120 cm. The same decreasing trend could be observed in particular for potassium (K2O) but not for phosphorus (P2O5) which was present in very low concentrations at all soil layers. Texture was relatively homogenous along the soil profile and within trenches (sandy loam), although relatively higher percentages of silt were observed at trench 2 and relatively higher percentages of clay were observed at trench 4.

Table 4.1 General soil characteristics (mean±SE, n=6).

Depth (cm) Org. mat. (%) pH K2O (mg/kg) P2O5 (mg/kg) Texture 10 5.4±0.5 6.1±0.3 354.9±78.1 < 8 Sandy loam 30 2.1±0.2 6.5±0.2 214.1±41.9 < 8 Sandy loam 60 0.6±0.2 7.1±0.2 73.3±09.4 < 8 Sandy loam 120 0.3±0.1 7.3±3.0 37.2±15.2 < 8 Sandy loam

68

Aboveground characterisation of the plant community The two Erica species were largely dominant (Table 4.2) both in terms of stem density (64% of the total) as in terms of basal area (84% of the total). Among the two Erica species, E. scoparia presented the highest stem density and the highest basal area (representing 82% and 63% of both species, respectively). The remaining three species were far less important, Pteridium being the second most dense species (16%) and Ulex representing the second highest basal area (9%).

Vertical mapping Root counts on vertical maps revealed that the average root density at the community level within the studied profile was 19.24±4.31 roots/dm2 (Table 4.3). According to the adjustment of the MLDR model to the overall root counts, 50% of all roots were found within the top 26 cm of soil (i.e.. D50 = 26 cm). Considering all root counts distributions, the MLDR model was able to explain between 96.9% and 99.9% of the variance. When considering the different species separately, the results for the two Erica species represented the highest average density of roots (59%), followed by Ulex (34%), Rubus (6%) and Pteridium (1%). Root densities of the four species were significantly different (p<0.05) for fine and for small but not for medium and coarse roots. Fine roots represented 96% of the overall (all species) root density. This proportion of fine roots was similar in all species except in the case of Pteridium (76%). With the exception of Pteridium, all species presented significantly higher root densities for fine and small roots and non-significant differences for medium and coarse roots. Most part of rhizomes from Pteridium were classified as decaying roots (65%). Decaying roots represented only 3 to 4% for the remaining species.

Table 4.2 Aboveground plant cover before trench excavation. Basal area is the sum of individual cross sections measured at the stem base per m2 (mean±SE, n=6).

Species Stem density (n/m2) Basal area (cm2/m2) Erica 17.4±4.6 29.5±5.6 Pteridium 4.5±1.8 1.7±0.6 Rubus 3.2±0.8 0.7±0.2 Ulex 2.1±0.7 3.0±0.9 All species 27.2±3.2 34.9±4.3

69

Table 4.3 Average root density (number of root counts/dm2) including all depths (mean±SE, n=6). Means followed by the same letter are not significantly different (p<0.05) according to the Mann-Whitney U test. The first letter refers to differences among diameter classes (columns) and the second letter refers to differences among species (rows). Symbol “_” represents no data.

Species < 1 mm 1-5 mm 5-10 mm > 10 mm All diameters Erica 11.16±4.06 a,ax 0.22±0.06 bb,ax 0.01±0.00 c,ax 0.01±0.00 c,ax 11.41±4.11 ax Pteridium 0.10±0.04 a,cx 0.02±0.01 ab,cx 0.00±0.00 b,ax 0.00±0.00 b,ax 0.13±0.05 cx Rubus 1.12±0.38 a,bx 0.04±0.01 bb,cx 0.00±0.00 c,ax _ 1.16±0.39 bx Ulex 6.36±1.87 a,ab 0.10±0.02 bb,bx 0.01±0.00 c,ax 0.01±0.00 c,ax 6.48±1.89 ab All species 18.75±4.23 a,xx 0.39±0.06 bb,xx 0.03±0.01 c,xx 0.02±0.01 c,xx 19.24±4.31xx

The vertical distribution of root counts (Fig. 4.2) was considerably different for each species. Erica presented the highest root densities at all depths with a single exception at the 90-100 cm layer where Ulex presented a higher value. The values of

D50 obtained from the mean root distribution were higher for Ulex (40 cm) and Erica (22 cm), and lower for Pteridium (9 cm) and Rubus (3 cm). All species presented roots at all studied depths except Pteridium which did not present roots below 80 cm of soil depth, as obtained from vertical map counts. All species have shown a consistently decreasing trend with depth except Ulex, which presented a peak of root counts at the 20-30 cm layer. Roots were more concentrated at upper layers for increasing root diameter (Fig.

4.3). The value of D50 for fine roots was considerably higher (25 cm) when compared to the other classes (10 cm for small roots, 6 cm for medium roots and 4 cm for coarse roots). Roots from all diameter classes were found down to 120 cm except coarse roots which were not found below 110 cm in any of the studied trenches. When considering trenches separately a significant positive correlation was observed between the overall D50 of each trench and the average thickness of the A horizon (r2 = 0.68; n = 6).

Estimation of maximum rooting depth The horizontal maps revealed that 77% of all roots detected at the bottom of the trenches were roots from Erica and 23% were roots from Ulex. The mean estimated maximum rooting depth of all Erica plants was 222.1±10.1 cm (n=6). This value was obtained using the averages of the 28 deepest roots of each trench.

70

Root density (n/dm2) Root density (n/dm2)

0 10203040 012345

0-10

20-30

40-50 Erica Pteridium

9 cm

Depth (cm) Depth 60-70 22 cm

80-90

100-110

0,0 0,5 1,0 0,0 0,5 1,0

02,557,510 0 5 10 15 20

0-10

20-30

40-50 Rubus Ulex

60-70 3 cm Depth (cm) Depth

40 cm 80-90

100-110

0,0 0,5 1,0 0,0 0,5 1,0

Fig. 4.2 Average density of root counts for each species including all diameters. (mean±SE, n=6). Insets represent the MLDR model fitted to the cumulative root fraction. Horizontal lines and corresponding depths indicate the value of D50.

The value 28 corresponded to the minimum number of deep roots mapped in a single trench (trench 2) within the six trenches. The estimated maximum rooting depth for the deepest root was 329 cm. When considering the 6 trenches separately, the estimated maximum rooting depth ranged from 192.8±5.5 cm at trench 4 to 256.1±5.9 cm (n=28) at trench 1 (Fig. 4.4).Although we could not find an equation to estimate the maximum rooting depth of Ulex, most observed plants had very deep reaching tap roots and the deepest root mapped was at 214 cm deep.

71

Root density (n/dm2) Root density (n/dm2) 0 1020304050 012345

0-10

20-30

40-50 < 1 mm 1-5 mm

60-70 10 cm Depth (cm) Depth 25 cm

80-90

100-110

0,0 0,5 1,0 0,0 0,5 1,0

0 0,2 0,4 0,6 0,8 1 0 0,1 0,2 0,3 0,4

0-10

20-30

40-50 5-10 mm > 10 mm

6 cm 4 cm 60-70 Depth (cm) Depth

80-90

100-110

0,0 0,5 1,0 0,0 0,5 1,0

Fig. 4.3 Average density of root counts for each diameter, including all species (mean±SE, n=6). Insets represent the MLDR model fitted to the cumulative root fraction. Horizontal lines and corresponding depths indicate the value of D50.

Moreover the average diameter of Ulex roots was higher (11.3±2.4 mm) than the average root diameter of Erica (7.3±0.5 mm) although the difference was not statistically significant. The estimated distribution of maximum rooting depths (Fig. 4.5) indicated that 82% of roots growing deeper than 125 cm were not reaching more than 175 cm in depth. A chi-square test for a Poisson distribution goodness-of-fit, showed for each trench that the distribution of roots was not random (distribution rejected at p<0.001). The coefficient of dispersion (estimated variance/estimated mean) ranged from 3.0 to 7.1, thus indicating a markedly clumped distribution of deep roots.

72

Fig. 4.4 Schematic representation of the estimated average maximum rooting depth (mean of the deepest 28 roots) of Erica plants at each trench. The solid line represents the soil surface. The broken straight line represents the maximum depth of excavation (bottom of the trenches). Both the above and the belowground parts of each plant have been drawn to scale in order to represent the average height and the average maximum rooting depth respectively. Each trench is represented by an image obtained from an Erica scoparia individual.

Root density (n/dm2) 00,511,5

125-149

150-174

175-199

200-224

225-249

Depth (cm) Depth 250-274

275-299

300-325

> 325

Fig. 4.5 Distribution of the estimated maximum rooting depths of Erica (mean±SE; n=2, n=3 and n=4 for the first, second and third depth classes, respectively; n=6 for the remaining classes).

73

Core samples

Values of root length density ranged from 4.6±0.7 cm/cm3 at 10 cm of soil depth to 0.8±0.2 cm/cm3 at 80 cm of soil depth. Values of root biomass ranged from 7.7±2.8 mg/cm3 at 10 cm of soil depth to 0.6±0.2 mg/cm3 at 40 cm of soil depth. According to the results shown on Fig. 4.6, the distribution of fine root biomass was similar (D50 = 12 cm) to the distribution of fine root length (D50 = 14 cm), although displaying a much higher variability. The MLDR model was able to explain 99.6% (biomass) and 99.7%

(length) of the variance. Values of D50 were significantly lower (p<0.001 ) than the value obtained with root counts for fine roots. The integration of the exponential model fitted to the average root distribution from core samples, allowed to estimate fine root biomass per unit area as 1.6 kg/m2 and the respective root length as 18.7 km/m2.

Root biomass (mg/cm3) Root length (cm/cm3)

024681012 02468

10

20

30

40

60 12 cm 14 cm Depth (cm) Depth

80

100 0,0 0,5 1,0 0,0 0,5 1,0

Fig. 4.6 Biomass and length of fine roots per unit of soil volume (mean±SE, n=10) as obtained from core samples. Gaps at 50 cm, 70 cm and 90 cm on the y axis correspond to non-sampled depths. Insets represent the MLDR model fitted to the cumulative root fraction. Horizontal lines and corresponding depths indicate the value of D50.

74

4.4 Discussion Our results have showed the existence of quite distinct patterns of root distribution for the species present at the plant community. The different root distribution patterns may be associated to the existence of distinct ecological strategies for water and solute uptake. In fact the results suggest the existence of a niche separation (Casper & Jackson, 1997) among the four species associated to different adaptive strategies. Root distribution patterns from vertical maps essentially reflected the root distribution of individual plants from each species as could be observed at the study site. Erica plants were showing what was described by Canadell & Zedler (1995) as a dual root system consisting of deep tap roots and laterally spreading roots at the upper layers. Plants of Ulex, were also deep rooted but not showing an important lateral root development at upper layers. The consequent lower development of fine roots at upper layers accounted for the peak of root counts below the first 20 cm and the higher value of D50. The clumping distribution of roots at the bottom of the trenches is in accordance with the root system architecture of individual Erica and Ulex plants. In fact tap roots were growing in a nearly vertical direction, thus originating a patchy distribution of roots on the horizontal maps, each patch apparently belonging to one individual plant. Deep roots are fundamental for plants to overcome the water stress verified in the upper soil layers typical of mediterranean conditions during the dry season (Canadell et al., 1996). This rooting pattern has been found to be typical of climates with a dry summer and important winter rainfall (Shenk & Jackson, 2002b) which is in accordance with the conditions of the study site. However in this specific case it is possible that more roots have been preferentially growing vertically due to neighbour competition (Atkinson, 1978) because of the extreme density of the stand. Besides the competition between plants of the same species, it is very likely the existence of inter-specific competition. Although we have not approached this specific aspect, different works refer the existence of asymmetric competition in mixed stands (e.g. McKay, 1988; Casper & Jackson, 1997; Leuschner et al., 2000) as a result of distinct levels of soil exploitation efficiency by different species. In the case of Rubus, individual plants were showing a high density of fine roots close to the soil surface but also having a few deeper roots. This explains the low value of D50 and the decreasing pattern obtained with vertical root counts. Pteridium plants were showing a dense network of horizontally oriented rhizomes very close to the soil surface, to which relatively sparsely distributed fine roots were directly connected. About this species we

75 should take into account its particular characteristics in terms of life cycle. In fact the life cycle of Pteridium is highly seasonal (Pakeman & Marrs, 1994) leading to a high variability in aboveground structures (fronds decay at the beginning of the cold season), but apparently also belowground as could be observed in this study by the high percentage of decaying rhizomes. In this way our results should be interpreted taking into account the fact that they report observations made during the growth period (June - August), thus reflecting the characteristics of this species in this particular season. Root distribution differences between each diameter class are partially a consequence of the preferential location of structural roots from the different species at upper soil layers. In the particular case of medium and coarse roots there was a considerable contribution of Pteridium rhizomes for the obtained results. These roots located at upper soil layers play a fundamental role in plant anchorage (Coutts, 1983; Fitter & Ennos, 1989) and in soil stability (Ziemer, 1981). Thus the deeper and more uniform distribution of finer roots and the shallower and more heterogeneous distribution of coarser roots is not surprising even if different species were contributing to this pattern. Since this is, to our knowledge, the only study about the root distribution of an Erica-dominated shrub community, it is difficult to have other references for comparison. Moreover among the panoply of methods which can be found in the literature, most of them are based on biomass estimation rather than root counts. This is evident in the compilation of eleven sclerophyllous shrubland root distribution studies of different regions presented by Jackson et al. (1996), where all of them are biomass studies. In this work and also in the seventeen studies compilation by Shenk & Jackson (2002a) the overall root distribution for sclerophyllous shrublands provided an average

D50 estimate of 19 cm which is lower than our estimate based on root counts (26 cm). Another compilation of fine root (in this case, roots ≤ 2 mm diameter) data from six sclerophyllous vegetation studies by Jackson et al. (1997) indicated a value of 14 cm for

D50 which is much lower than our estimate for fine roots based on root counts (27 cm), but exactly the same as our estimate for fine roots based on root length. In fact we have observed a remarkable difference between the results obtained with core samples and those obtained with root counts. There is not always a direct relationship between root counts and root length or root biomass and the differences between the two variables observed in our study seem to confirm this aspect reported in other studies (Van Noordwijk et al., 2000). Besides any methodological constraints, the final root count distribution definitely resulted from the influence of trenches 4 and 5 which were

76 presenting a thick organic A horizon. In these trenches roots were in general more uniformly distributed following the organic matter distribution along the profile. In what concerns the estimation of total (dead and live) fine root biomass, our estimates are higher than the six study average of 0.52 kg/m2 referred by Jackson et al. (1997) and higher than the value of 0.78 kg/m2 estimated by Martinez et al. (1998) for a mediterranean sand dune shrub community in Spain. Apparently the intricate root mat composed of lignified Erica roots represents a higher biomass than what is reported by the referred studies. We also admit that a small percentage of thicker roots (> 1mm) have been included in the soil cores, hence contributing to this comparatively high value. On the other hand, our estimates of root length density (18.7 km/m2) match well with the estimation of 17.5 km/m2 (Jackson et al., 1997) for sclerophyllous shrubs and trees. It is remarkable the fact that both estimates are in general higher than values reported for forest ecosystems (e.g. Jackson et al., 1997; Vande Walle et al., 1998; Wiesenmüller, 1998). In our case this is not surprising if we consider that the basal area of the shrubland under study was higher than the normal values of adult forest stands conducted for timber production (20-30 m2/ha for most species of the temperate region). Many works express the root occupation in the soil by computing the average distance between roots instead of the root length density. Distance between roots can be computed as the inverse of root length density (Miller & Ng, 1977). In our case the distance between roots estimated for the first meter of soil depth was 0.9 cm. This is a much lower value than the 2.0 cm estimated by Kummerow et al. (1977) for a Californian chaparral, the 2.8 cm estimated by Hoffmann & Kummerow et al. (1978) for a Chilean matorral and the values of 1.9 cm and 1.7 cm estimated by Martinez & Rodriguez (1988) and Martinez et al. (1998) for shrub communities in Southern Spain. The higher soil occupation by roots verified in this study reveals an important difference in terms of the belowground characteristics of a thick maquis when compared with the sparser shrub communities studied in the referred studies. Besides the different floristic composition of our shrub community, we may speculate that the existence of such an important soil occupation is only possible because of the relatively high nutrient content and the high (for Mediterranean standards) values of precipitation verified. Thus there seems to exist a different pattern of belowground occupation at more mesic mediterranean sites such as in our case. Our estimates of maximum rooting depth for the two Erica species are reasonably close to the values reported by Canadell et al. (1996) for other Ericaceae

77

(Arbutus unedo, 350 cm; Erica arborea, 200 cm). The estimated distribution of maximum rooting depths suggests that the bulk of deep roots was roughly located within a 50 cm soil layer between 125 cm and 175 cm, and that only few roots reached deeper soil layers. The reasons for this distribution were not assessed within the scope of the present study but we may speculate that deeper roots can be associated to higher aboveground biomass and leaf surface (i.e. bigger plants) as suggested by Schenk & Jackson (2002b). Since this are the first data to our knowledge on the root distribution of Ulex, no additional information was available from the literature reporting the maximum rooting depth of Ulex plants or even of other phylogenetically close leguminous shrubs. However our results strongly suggest the existence of root systems at least as deep as those studied for Erica. When applying the same allometric relationship found for Erica to Ulex, we have obtained an absolute maximum rooting depth of 349 cm and an average for the 28 deepest roots of each trench of 221 cm. The importance of water uptake by deep roots was confirmed by Silva et al. (2002) for this same shrub community. Moisture measurements have revealed a cumulative soil water depletion at the end of the dry season of 18 mm at the 160-180 cm depth layer and 12 mm at the 120-140 cm depth layer. Although many questions will remain to be answered concerning the belowground structure of mediterranean shrublands and its vital role for the functioning of these plant communities, the present work may help understanding some of these aspects, in particular for Erica dominated maquis, common in the western Mediterranean Region. Further studies on this type of plant communities, namely those directed to belowground physiological processes, may benefit from the results presented here.

References

Atkinson D. (1978). The use of soil resources in high density planting systems. Acta Horticulturae 65: 75-90. Atkinson D. (2000). Root characteristics: Why and what to measure. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 1-32 pp. Böhm W. (1979). Methods of studying root systems. Springer Verlag, Berlin. Caldwell M.M. & Richards J.H. (1986). Competing root systems: morphology and models of absorption. In Givnish T.J. (ed.), On the Economy of Plant Life and Function. Cambridge University Press, New York, 251-273 pp.

78

Caldwell M.M. & Virginia R.A. (1989). Root systems. In Pearcy R.W., Ehleringer J.R., Mooney H.A. & Rundel P. (eds.), Physiological plant ecology. Field methods and instrumentation. Chapman and Hall, London, 367-398 pp. Canadell J, Jackson R.B., Ehleringer J.R., Mooney H.A., Sala O.E. & Schulze E.-D. (1996). Maximum root depth of vegetation types at the global scale. Oecologia 108: 583-595. Canadell J. & Rodá F. (1989). Root biomass of Quercus ilex in a montane mediterranean forest. Canadian Journal of Forest Research 21: 1771-1778. Canadell J. & Zedler P.H. (1995). Underground structures of woody plants in mediterranean ecosystems of Australia, California and Chile. In Arroyo M.T., Zedler P.H. & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecosystems in Chile, California and Australia. Springer Verlag, Berlin, 177-210 pp. Cañellas I. & Ayanz A. (2000). Biomasss of root and shoot systems of Quercus coccifera shrublands in Eastern Spain. Annals of Forest Sciences 57: 803-810. Casper B.B. & Jackson R.B. (1997). Plant competition underground. Annual Review of Ecology and Systematics 28: 545-570. Coutts M.P. (1983). Root architecture and tree stability. Plant and Soil 71: 171-188. Djema A. (1995). Cuantification de la biomassa e mineralomassa subterranea de un bosque de Quercus ilex. MSc. thesis. Instituto Agronomico Mediterraneo de Zaragoza. Drexhage M. & Gruber F. (1998). Architecture of the skeletal root system of 40-year-old Picea abies on strongly acidified soils in the Harz Mountains (Germany).Canadian Journal of Forest Research 28: 13-22. Fitter A.H. & Ennos R.A. (1989). Architectural constrains to root system function. In Robinson D. (ed.), Roots and the soil environment. Aspects of Applied Biology 2:15-22. Fitter A.H. (1996). Characteristics and functions of root systems. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 1-20 pp. Gale M.R. & Grigal D.F. (1987). Vertical root distribution of northern tree species in relation to successional status. Canadian Journal of Forest Research 17: 829-834. Gerwitz A. & Page E.R. (1974). An empirical mathematical model to describe plant root systems Journal of Applied Ecology 11: 773-781. Hoffmann A., & Kummerow J. (1978). Root studies in the Chilean Matorral. Oecologia 32: 57-69. Jackson R.B., Canadell J., Ehleringer J.R., Mooney H.A., Sala O.E. & Shulze E.D. (1996). A global analysis of root distributions for terrestrial biomes. Oecologia 108: 389-411. Jackson R.B., Mooney H.A. & Schulze E.-D. (1997). A global budget for fine root biomass, surface area, and nutrient contents. Proceedings of the National Academy of Science of the United States of America 94: 7363-7366. Jobbágy E. & Jackson R. (2000). The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecological Applications 10: 423-436. Klepper B. & Rickman R.W. (1990). Modelling crop root growth and function. Advances in Agronomy 44: 113-132. Kummerow J. (1981). Structure of roots and root systems. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 269- 288 pp. Kummerow J., Krause D., & Jow W. (1977). Root systems of chaparral shrubs. Oecologia 29: 163-177. Kummerow J., Kummerow M. & Trabaud L. (1990). Root biomass, root distribution and the fine-root growth dynamics of Quercus coccifera L. in the garrigue of Southern France. Vegetatio 87: 37- 44. Leuschner C., Hertel D., Coners H. & Büttner V. (2000). Root competition between beech and oak: a hypothesis. Oecologia 126: 276-284.

79

Marsh B. (1971). Measurements of length in random arrangements of lines. Journal of Applied Ecology 8: 265-267. Martinez F. & Rodriguez J.M. (1988). Distribución vertical de las raices del matorral de Doñana. Lagascalia 15: 549-557. Martínez F., Merino O., Martín A., García Martín D., & Merino J. (1998). Belowground structure and production in a mediterranean sand dune shrub community. Plant and Soil 201: 209-216. McKay H.M. (1988) The influence of pine on the form of Sitka spruce fine roots. Journal of Experimental Botany 39: 1263-1266. Miller P.C. & Ng E. (1977). Root:shoot biomass relations in shrubs in Southern California and Central Chile. Madroño 24: 215-223. Monteith J.L., Huda A.K.S. & Midya D. (1989). RESCAP: a resource capture model for sorghum and pearl millet. In Virmani S.M., Tandon H.L.S., Alagarswamy (eds.), Modelling the growth and development of sorghum and pearl millet. ICRISAT Research Bulletin 12: 30-34. Nepstad D.C., Carvalho C.R., Davidson E.A., Jipp P.H., Lefebvre P.A., Negreiros G.H., Silva E.D. da, Stone T.A., Trumbore S.E. & Vieira S. (1994). The role of deep roots in the hydrological and carbon cycles of amazonian forests and pastures. Nature 372: 666-669. Pagès L., Asseng S., Pellerin S. & Diggle A. (2000). Modelling root system growth and architecture. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 113-146 pp. Pakeman R.J. & Marrs R.H. (1994). The effects of control on the biomass, carbohydrate, content and bud reserves of bracken (Pteridium aquilinum L. Kuhn), and an evaluation of a bracken growth model. Annals of Applied Biology 124: 479-493. Schenk H.J. & Jackson R.B. (2002a). The global biogeography of roots. Ecological Monographs 72: 311- 328. Schenk H.J. & Jackson R.B. (2002b). Rooting depths, lateral root spreads, and belowground/aboveground allometries of plants in water-limited ecosystems. Journal of Ecology 90: 480-494. Schuurman J.J. & Goedewaagen M.A. (1971). Methods for the examination of root systems and roots. Centre for Agricultural Publishing and Documentation, Wageningen. Shani U. & Dudley L.M. (1996). Modelling water uptake by roots under water and salt stress: Soil-based and crop response root sink terms. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 1-20 pp. Silva J.S., Rego F.C. & Mazzoleni S. (2002) Fire effects on soil water dynamics in a mediterranean shrubland. In Proceedings of the 4th International Conference on Forest Fire Research. Luso, Portugal. Sokal R.R. & Rohlf F.J. (1995). Biometry. W.D. Freeman and Company, New York. Tennant D. (1975). A test of a modified line intersect method of estimating root length. Journal of Ecology 99: 995-1001. Van Noordwijk M., Brouwer F., Meijboom M., Oliveira M. do R. & Bengough A.G. (2000). Trench profile techniques and core break methods. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 211-234 pp. Vande Walle I., Willems S., Lemeur R. (1998). Root length and distribution in the mineral soil of a mixed deciduous forest (experimental forest aelmoeseneie). Silva Gandeensis 63: 1-15. Wiesenmüller J., Santos W., Denich M. & Vlek P.L. (1998). Modelling of fine root distribution under secondary vegetation in NE Amazonia – a qualitative and quantitative assessment. In Proceedings of the Third SHIFT-Workshop. Manaus, Brazil, 185-189 pp. Zeng X. (2001). Global vegetation root distribution for land modelling. Journal of Hydrometeorology 2: 525-530. Zeng X., Dai Y., Dickinson R.E. & Shaikh M. (1998) The role of root distribution for climate simulation over land. Geophysical Research Letters 25: 4533-4536.

80

Ziemer R.R. (1981). Roots and stability of forested slopes. In Proceedings of the Symposium on Erosion and Sediment Transport in Pacific Rim Wetlands. Christchurch, New Zealand, 343-361 pp.

81

5 FIRE EFFECTS ON SOIL WATER DYNAMICS IN A MEDITERRANEAN SHRUBLAND4

Abstract: With the purpose of studying the effects of fire on the natural dynamics of soil water, an experiment was set in a mediterranean shrubland in Central West Portugal. Soil moisture was measured in two closely located plots at six depths down to 170 cm, during 2000 (reference period) and after an experimental fire set on one of the plots in June 2001 (treatment period). Moisture differences between homologous days

(i) of the treatment and the reference periods were computed both for the control (Dci) and for the burned (Dbi) plots. Results showed significantly higher values of Dbi both at the dry and at the rainy seasons. Values of Dci were subtracted from Dbi in order to estimate the net effect of fire on soil water storage (Si). Values of Si showed a maximum of 108.2 mm in the dry season and a maximum of 145.2 mm in the rainy season. At the upper soil layers values of Si decreased as vegetation recovery proceeded along the dry season. Key words: Fire effects, soil water, mediterranean ecosystems, Root distribution

4 Based on paper: Silva J.S., Rego F.C. & Mazzoleni S. (2002). Fire effects on soil water dynamics in a mediterranean shrubland. In Proceedings of the 4th International Conference on Forest Fire Research. Luso, Portugal.

82

5.1 Introduction The effects of fire on mediterranean ecosystems have been thoroughly studied all over the five Mediterranean Regions of the World. However the knowledge concerning the effects of fire belowground is still very limited. One of the major aspects to be understood is the effect of fire on the natural dynamics of soil water, since it is the basic limiting factor for plant growth in mediterranean ecosystems. Immediate consequences of fire are the elimination of the transpiring surface of plants and the exposure of soil to the direct action of meteorological agents. These two basic consequences have opposed effects in terms of soil water (Pyne et al., 1996; Zwolinski, 2000). When compared to undisturbed situations, the elimination of plant transpiration leads to an increase of soil water storage (Tiedmann et al., 1979), whereas the exposure of soil to rain, wind and solar radiation contributes to a lower water content. At the surface soil layers water is lost by direct evaporation. Also the effect of rainfall on soil water recharge is diminished due to a decrease in infiltration accompanied by an increase in runoff. Different works report the formation of an hydrophobic soil layer after intense fires (e.g. DeBano, 1966; Ferreira, 1990; Midoun et al., 1998). This process works together with the absence of vegetation cover and both are responsible for the lower infiltration rate, thus contributing to reduce water content and to increase runoff (DeBano, 2000). The relative importance of each of these processes depends on fire intensity, vegetation type, soil/relief characteristics and meteorological conditions. Few works report on the consequences of fire on soil water storage. Rego & Botelho (1992) found slightly higher moisture contents during summer at a high intensity prescribed burning plot, when compared with a low intensity and a control plot, in a young Pinus pinaster stand. Klock & Helvey (1976) have detected considerably lower soil water deficits in a mature mixed conifer forest during the years following a wildfire. Similar results were obtained by Soto & Diaz-Fierros (1997) in a mediterranean shrubland dominated by Ulex europaeus. However Campbell et al. (1977) reported a lower soil water content one year after a wildfire in a Pinus ponderosa burned stand, when compared to an undisturbed nearby plot. In this case only surface (down to 30 cm) moisture was measured and soil water differences in the burned area were attributed to increased evaporation and runoff. Contradictory conclusions about the effects of fire on soil water where reported by Wells et al. (1979) referring to studies from various sources. The existence of different conclusions from different studies

83 reflects two opposite processes: the increase of evaporation from the immediate sub- surface and the decrease of plant transpiration affecting the whole profile. Given that these two mechanisms have different expression at different soil layers it is important to study the whole soil profile, if possible down to the maximum rooting depth and taking into account the root distribution of plants. The study of fire effects on soil water may help understanding the frequently studied dynamics of vegetation in the post-fire period. This knowledge is particularly important for mediterranean ecosystems where soil water is the basic limiting factor for plant growth. The present study aims to contribute to a better knowledge of this specific aspect of fire effects in mediterranean ecosystems. In particular, we aim to determine the effect of a simulated wildfire on soil moisture dynamics at different soil layers in a typical mediterranean shrubland in Central West Portugal.

5.2 Methods The study was carried out at Tapada Nacional de Mafra in the Central West Region of Portugal. Tapada Nacional de Mafra is a protected area with 827 ha, about 30 km Northwest of Lisbon and 12 km East from the coast (38º 58’ 30’’ N and 9º 15’ 52’’ W). The lowest altitude is 90 m and the highest is 358 m. Soils are humic cambissols derived from sandstone. Mean annual precipitation is 798 mm and mean annual temperature is 14.6 ºC. June, July and August are the three driest months accounting for only 3.1% of total annual rainfall. The study area was a shrubland dominated by Erica scoparia L. and Erica lusitanica Rudolphi, both species being hereafter referred to as Erica. Other important species were Crataegus monogyna Jacq., Ulex jussiaei Webb, Daphne gnidium L., Rubus ulmifolius Schott. and Pteridium aquilinum (L) Kuhn. The experiment was designed as two neighbouring plots with four internal replications, each replication consisting of one tube for moisture measurement. Plots and replications were disposed linearly along a contour line on a 15% slope. One plot was kept intact as a control (control plot) and the other was burned (burned plot) by an experimental fire on the 4th of June 2001. Fire was planned to simulate a typical summer fire for which an area of approximately 0,2 ha was burned using fire torches. During the fire treatment air temperature ranged from 23.3 ºC to 28.7 ºC, relative humidity ranged from 81% to 82% and wind ranged from 1.5 m/s to 2.8 m/s. Within each plot replications were separated by 8 meters and the 2 plots were separated by 27 meters including a fire-break, 10 m

84 wide. On each plot soil moisture (% volume) was measured using a TDR (Time Domain Reflectometry) equipment consisting on a measuring device (TRIME-FM3; IMKO GmbH, Ettlingen) connected to a tube-access probe (TRIME-T3). The probe was driven inside Tecanat tubes inserted in the soil, 2 m deep, in order to measure moisture at different soil layers. During fire, tubes (the remaining aboveground top) were covered with sacs filled with soil in order to insulate them from high temperatures. For each tube two crossed measurements were made at 15 cm, 30 cm, 50 cm, 90 cm, 130 cm and 170 cm, from April 2000 to December 2001. Moisture measurements were part of a broader study, but for the purpose of the present study two measuring periods were considered: a reference period from the 4th of June 2000 to the 20th of December 2000 and a treatment (post-fire) period from the 4th of June 2001 to the 20th of December 2001. Within each period we have distinguished a dry season going from the 4th of June to the 19th of September and a rainy season going from the 21st of September to the 20th of December. The 19th and the 21st of September were coincident with the first significant rain (> 10 mm precipitation) marking the end of the dry season, for 2000 and 2001, respectively. In average measurements were made once every two days although during periods with no precipitation only one measurement per week was made. Due to equipment problems the largest interval between consecutive measurements was 17 days in July 2000. Meteorological data was obtained on a daily basis along the whole period of measurements using a Weather Monitor II station (Davis Instruments Co., San Francisco) located 2 km away from the experiment. According to laboratory analysis of the 48 samples taken at all measurement depths of all replications, soil texture was classified as a sandy loam although a few samples have been classified as loams (3 samples from the burned plot and 2 samples from the control plot) and silt loams (2 samples from the control plot). The vertical root distribution for each of the two plots was determined by counting the number of root tips on 3 m long trench walls located at the lower part of the slope, 4 m away from each tube. Roots were drawn and counted on plastic sheets down to 120 cm. Besides the exclusion of obviously decaying roots, no distinction between dead and live roots was attempted. Maximum rooting depth was indirectly accessed for the two dominant Erica species by using an allometric relationship between root diameter and vertical distance down to the root apex. Estimation of maximum rooting depth was obtained by measuring the diameters of the cut root at the bottom of the trenches and then using the allometric relationship to determine the correspondent vertical distance (Silva & Rego,

85 unpublished). Vegetation was characterised in terms of plant height, number of stems and basal area (sum of cross sectional areas per square meter) on each of the two plots. The colonisation of the burned area was followed until October 2001 simply by counting the number of P. aquilinum fronds, the number of resprouting Erica plants and by estimating the respective average height on a 1,6 m2 plot close to each tube. All statistical comparisons between plots, between periods and between depths were performed using one mean value for each replication (i.e. sample size=4). Gaps between measurements were filled using linear interpolations. The effect of fire on soil water was assessed by comparing the moisture differences between each day (i) of the treatment period (Mti) and the homologous Julian day of the reference period (Mri), obtained at the two plots. The use of moisture differences from two years to compare plots was preferred to the simple use of raw data from one year. In fact given the reduced number of replications, even small variability levels within plots would have been sufficient to hide the effects of fire. The distribution of moisture differences

(Di=Mti-Mri) was tested for normality using a Kolmogorov-Smirnov test. The mean values of Di of the control plot (Dci) were compared with the mean values of Di of the burned plot (Dbi) for each depth using t tests. Multiple comparisons between depths were performed using Tuckey tests. Values of Dbi and Dci were converted into mm of soil water and estimated down to 180 cm deep, by integrating measurements using 20 cm increments. Since the distinct meteorological conditions of the two periods have also accounted for part of Dbi, these values were compared with Dci. Given that meteorological conditions were similar in both plots, the difference (Si) between Dbi and Dci was assumed to correspond to the net effect of fire on soil water storage.

5.3 Results Before fire the two plots showed a similar plant cover. The average shrub height was 196±20.6 cm for the control plot and 187±34.8 cm for the burned plot. The average stem densities were 26±3.2 stems/m2 and 29±5.2 stems/m2, respectively. The average basal area (the sum of cross sectional areas at the stem base, per unit area) were 41±5.4 cm2/m2 and 29±5.0 cm2/m2, respectively. The two Erica species were largely dominant both in terms of stem density (58±12.7% and 56±24.7% for the control and the burned plots respectively) as in terms of basal area (87±5.8% and 71±16.2%, respectively). The remaining species were far less important, P. aquilinum being the second most dense

86 species (18±8.0% and 20±16.6%, respectively) and U. jussiaei representing the second highest basal area (10±3.8% and 16±7.6% respectively). Fire killed nearly all aerial parts of plants and the litter/duff layer was reduced by 1.5 to 4 cm. Resprouting of plants started two weeks after fire. Fronds of P. aquilinum and resprouts from the two Erica species were the first signs of vegetation recovery (Fig. 5.1). Seven weeks after fire the number of resprouting plants of Erica had stabilized and the number of P. aquilinum fronds attained 83% of the maximum registered. Erica attained a maximum of 3.5 resprouting plants per square meter and a maximum average height of 14.2 cm. P. aquilinum attained a maximum average density of 8.6 fronds per square meter and a maximum average height of 63.7 cm. This maximum size for P. aquilinum was followed by a decrease in October, when fronds started decaying. In both plots roots were distributed following a typical exponentially decreasing curve. The first 20 cm of soil contained 36% and 28% of all roots counted respectively in the control and the burned plots (Fig. 5.2). However roots extended much below the 120 cm deep studied profile. The relationship used to estimate the maximum rooting depth (r2 = 0.89, n = 64) was a typical logistic function of form y=a/[1+(x/b)c] where a, b and c are constants, x is the root diameter and y is the correspondent vertical distance to the root apex. The rooting depth of the 28 deepest Erica roots of each trench was estimated to be 242±12.8 cm for the control plot and 206±7.4 cm for the burned plot. According to the observations made on completely excavated individual root systems of Erica spp. and U. jussiaei plants, both species present nearly vertical deep tap roots.

250 2 2

200 1,5 150 1 100 0,5 50 Average height (cm) 0 0 Nr. plants or fronds/m

pr 4- 28 17 25 3- 10 11 19 e- 6- -6 -7 -7 8- -8 -9 -1 fir 01 -0 -0 -0 01 -0 -0 0- e 1 1 1 1 1 01

Fig. 5.1 Evolution of plant cover during the treatment period. Triangles and diamonds represent Erica and P. aquilinum respectively. Open symbols/dashed lines refer to plant height whereas closed symbols/solid lines are the density of P. aquilinum fronds or resprouting of Erica plants.

87

Root density (n/dm2) Root density (n/dm2) 0 1020304050 0 1020304050

0-10

20-30

40-50

60-70 Depth (cm)Depth 80-90

100-110 Control plot Burned plot

Fig. 5.2 Root density (number of root counts/dm2) at the control and the burned plots (mean±SE). Data collected before fire.

REFERENCE PERIOD (2000) TREATMENT PERIOD (2001) 60

50

40

30

20

10

Soil moisture (%vol.) 0

15 cm 30 cm

50 cm 90 cm 60 130 cm 170 cm 50

40

30

20

10 Soil moisture (%vol.) moisture Soil

BURNED PLOT 0 CONTROL PLOT

80 30 70 25 60 20 50 40 15 30 10

Rainfall(mm) 20 5 10 0 0 C) (º temp. daily Aver. 04-06 01-07 28-07 24-08 20-09 17-10 13-11 10-12 04-06 01-07 28-07 24-08 20-09 17-10 13-11 10-12 Time Time

Fig. 5.3 Values of soil moisture for each plot in the two study periods with the corresponding meteorological data (bars and solid line represent rainfall and average daily temperature, respectively).

88

Rhizomes from P. aquilinum did not go deeper than 80 cm with 50% of all roots located within the first 7 cm of soil. The values registered for precipitation were quite different when comparing the rainy season of the reference (2000) and the treatment (2001) periods, but very similar during the dry season. Total rainfall was 29.3 mm during the dry season of the reference period and 561.4 mm during the rainy season. In the treatment period total rainfall was 25.2 mm in the dry season and 242.6 mm in the rainy season (Fig. 5.3). In both plots and in both periods, soil moisture followed a general pattern which can be described as a decreasing trend along the dry season and a sudden increase as the wetting front reached the different soil layers along the rainy season. This pattern varied according to each of the studied depths. The 15 cm soil layer presented the highest variation of water content in both study periods (SD =9.6), whereas the 170 cm soil layer presented the lowest variation (SD=3.2). In general each layer presented a typical moisture pattern along each study period in particular during the dry season. In average the wetting front took between 44 (15 cm) and 74 days (170 cm) to reach the different soil layers after the beginning of the rainy season. In general terms fire caused a change in the typical decreasing patterns of soil moisture in the dry season, in particular at deeper soil layers, where roughly constant values were observed at the burned plot. At the beginning of each period (4th of July) the average moisture content was very similar (36.2% for the reference period and 35.9% for the treatment period; see Fig. 5.4) and not significantly different. Although no analysis concerning the hydraulic properties of the soil had been performed at this stage, the method developed by Saxton et al. (1986) using texture based empirical relationships, indicated a field capacity between 18% and 26%, a wilting point between 8% and 14% and a saturation water content between 39% and 47

%. The average values of Dbi were higher than the average values of Dci, both within the dry and the rainy seasons (Table 5.1). These values were significantly higher at all depths except at 130 cm and 170 cm in the dry season and at 15 cm in the rainy season. When considering both seasons together, all depths provided significantly higher values of Dbi. During the dry season the average Dbi showed a maximum at 50 cm (5.93±0.43%) and a minimum at 130 cm (3.08±0.95%). During the rainy season the average Dbi showed a maximum at 170 cm (8.29±2.23%) and a minimum at 15 cm

(2.83±0.58%). The values of Dci ranged from -1.93±0.68% (15 cm, dry season) to

89

3.08±0.29% (30 cm, rainy season). Comparisons between depths provided significant differences for both plots at both seasons except for the burned plot in the dry season.

Moisture differences between periods for the control (Dci) and the burned (Dbi) plots revealed distinct patterns at all soil layers (Fig. 5.5). In general Dbi showed consistently higher values than Dci, thus positive values of Si. The exceptions were the first two soil layers where negative values of Si were observed by middle November and also the initial period after fire at all layers. In general Dbi curves showed positive values while

Dci showed values close to zero. The exceptions were the last 28- 16 days (depending on the soil layer) of the study, when a strong decrease was exhibited by the two curves due to heavy rainfall in the reference period. At the first soil layer (0-20 cm) there was an initial strong increase of Si followed by a decrease after middle July. This decreasing trend started later for the 20-40 cm layer (early August) and the 40-60 cm layers (middle August) and it was practically absent at deeper layers during the whole study period. At the 80-100 cm soil layer and below this depth, Si has basically increased along the whole study period. In particular the deepest layer (160-180 cm) showed a stabilisation around 16 mm of soil water after the end of the dry period. The overall (0-

180 cm) values of Si have increased until middle August and then have stabilised around 100 mm until October. Eight days before the end of the dry season the overall values of

Si showed a maximum of 108.2 mm. However the highest values of the overall Si were attained by middle October showing a peak of 145.2 mm. By the end of the study period, the overall Si had stabilised around 107 mm of soil water.

90

45

40

35

30

Fire 25

20 Control plot Soil moisture (% vol.) Burned plot 15

10 03-05-01 18-06-01 03-08-01 18-09-01 03-11-01 19-12-01 Time

Fig. 5.4 Average soil moisture of the control and the burned plots during the treatment period and two weeks before fire.

Table 5.1 Mean ± SE values of Di (soil moisture differences in terms of % volume, between the treatment and the reference periods). Means sharing the same letter are not statistically different (p >0.05, n=4). Letters in the first column refer to comparisons between plots within each season (t tests). Letters in the second column refer to multiple comparisons between depths (Tuckey tests).

Depth Dry season Rainy season Control Burned Control Burned 15 cm -1.93±0.68 a, a 3.71±0.74 b, a 1.94±0.20 a, bc 2.83±0.58 a, a 30 cm -0.07±0.22 a, ab 4.38±0.84 b, a 3.08±0.29 a, c 4.72±0.56 b, ab 50 cm 0.09±0.33 a, ab 5.93±0.43 b, a 2.69±0.75 a, bc 7.12±0.44 b, ab 90 cm 1.78±0.99 a, b 4.96±0.45 b, a 0.53±0.50 a, abc 6.26±0.70 b, ab 130 cm 1.35±0.52 a, b 3.08±0.95 a, a -0.73±0.99 a, a 6.13±0.97 b, ab 170 cm 1.84±0.75 a, b 4.76±0.98 a, a 0.12±0.40 a, ab 8.29±2.23 b, b

91

0 - 20 cm 30

10

-10

-30 20 - 40 cm 30

10

-10

-30

40 - 60 cm 30

10

-10

-30

80 - 100 cm 30

10

-10

-30

120 - 140 cm 30

SOIL WATER (mm) SOIL WATER (mm) 10

-10

-30

160 - 180 cm 30

10

-10

-30

250 0 - 180 cm 150 Si

50 Dbi -50 0 0 2 2 2 1 1 1 4- 1- 8- 4- 0- 7- 3- 0- 06 07 07 08 09 10 11 12 -150 Dc i

Fig. 5.5 Differences in soil water storage between the treatment and the reference periods in the burned

(Dbi) and the control (Dci) plots. Si = Dbi - Dci represents the effect of fire on soil water. Si is represented by a moving average (n=7). Estimation for 0-180 cm is an integration using 20 cm increments.

92

5.4 Discussion According to our results fire has definitely modified the soil water dynamics of the studied shrub community. These differences could be detected not only during the dry season where it would be more obvious due to the strong decrease in transpiration, but in general along the whole study period. The establishment of a dry season and a rainy season based on a precipitation threshold, had the simple purpose of separating the period with no significant water inputs, from the period of typical soil water recharge due to rainfall. However this distinction had a different meaning for different layers since the deepest layers had continued the same dry season trend much beyond its end whereas the upper soil layers had a more immediate response to the first Autmn rains. In fact each soil layer has a specific drying period followed by a wetting period as the wetting front reaches the corresponding depth. Another aspect to take into account in what concerns the timings established within this study, has to do with the date of burning. Results could had been considerably different if fire was set later in the dry season. For example, a September fire would certainly have a different effect because of a lower vegetation recovery due to a shorter growth period and lower water availability for plant growth. In such a case, the effects verified in this study could be expected to occur in the following growing season, as observed by Klock & Helvey (1976) after an August wildfire. The distinct patterns observed for soil water dynamics in the dry season within the different soil layers are basically explained by the direct effect of evaporation, the soil water extraction by roots and the water flows in the soil. All three effects potentially contribute for a higher water depletion in the upper soil layers. As shown by the root distribution data, most of the roots were located at the upper layers. In particular the species which have presented the most important regenerative response to fire, P. aquilinum, had very shallow rhizomes. In contrast to this, deeper soil layers did not suffer the effects of direct evaporation at the soil surface and the only roots present were essentially from Erica. This apparently explains the fact that while at the upper layers there was a marked decrease of soil water, especially after the flush of P. aquilinum fronds in July, at deeper soil layers there was a roughly constant soil water content. At the upper layers this corresponded to a decrease in Dbi, presumably due to an increasing water extraction by roots from new P. aquilinum fronds but also from new

Erica shoots. In contrast, deeper layers have shown increasingly high values of Dbi, which is conclusive about the importance of water extraction by deep roots from Erica

93 during the dry season of the reference year. This in accordance to what has been reported for other deep rooted species in different Mediterranean Regions (Canadell & Zedler, 1995) and in general for the role of deep roots in most ecosystems (Canadell et al., 1996). During the rainy season, different patterns of soil water recharge could be associated to different soil layers. These patterns basically reflected the time that the wetting front took to reach the different depths. Again it was observed a much quick and sensitive response from the upper layers (0-40 cm) at both plots. However differences between plots were not so consistent during the rainy season at these layers, given the oscillation between negative and positive values of Si. Also given the diversity of factors acting at the upper layers it is not straightforward to determine which was the basic cause for the lower values of Si at this period and at this zone of the soil. Although we have not been able to prove this hypothesis within the scope of the present study, we can speculate that soil exposure at the burned plot was responsible for a higher evaporation in the treatment period during the unusually dry November and December months, than at the control plot. One important consequence of our findings is related to the influence of soil water on plant regrowth in the post-fire period. The high soil water content verified in the present study apparently explains the quick increase of plant cover at the burned plot. Similar post-fire conditions may be at the origin of the high water potentials and transpiration rates observed for regenerating seeder and resprouter species (Clemente, in prep.) in another mediterranean shrubland. However the results of this study also suggest that the unusual dry period in late Autmn 2001 was related to comparatively lower soil moisture levels in surface layers of the burned area, although the same area was characterised by a higher moisture level in the previous period. Unusually dry periods have been reported as possible critical factors which explain, in interaction with fire, the mortality of plants during regrowth periods (Mazzoleni & Pizzolongo, 1990). Experiments under controlled conditions showed that individuals of Erica arborea are sensitive to water conditions during resprouting after disturbance (Mazzoleni & Esposito, 1993). Our results confirm the interest of interactions between fire and climatic conditions in relation to vegetation dynamics processes. Some works report a detrimental effect of fire on soil water content (Campbell et al., 1977; Redmann, 1978; Wells et al., 1979) in the immediate post fire period, thus apparently showing opposite conclusions to those presented in this study. However

94 several reasons may be at the origin of these different results. In fact none of these works was performed under the same experimental conditions as our work, namely in what concerns the studied depths, the vegetation type, the climate and the burning season. One of the few works confirming an increase in soil water storage after fire was presented by Klock & Helvey (1976). The estimated increase of 108.2 mm of soil water storage due to fire effects in September is very similar to the 116 mm reported by Klock & Helvey (1976) for a mixed conifer forest one year after fire. However it is difficult to take conclusions from these similar results since the study conditions were distinct in what concerns the sampling design (no pre-fire measurements available) the burning period (August), climate and type of vegetation. We should note that the values of Si obtained in our study were probably overestimated at the first soil layer (0-20 cm) during the dry season and underestimated during the rainy season. These estimations were probably affected by the use of a 20 cm integration increment because of the highly variable and heterogeneous nature of this first soil layer. On the other hand the estimation for the whole profile (0-180 cm) should not be faced as the total impact of fire on soil water storage. In fact both the water uptake by the deepest roots during the reference period and the higher water percolation to soil layers below the studied depth during the treatment period, may have contributed to even higher increases in soil water storage as an effect of fire. The results of the present study unequivocally showed the existence of a much higher soil water content along the studied soil profile during the dry season, likely as a consequence of the reduction of transpiring vegetation by fire. Furthermore this effect was consistent along the whole studied period including the rainy season. Therefore, the more thoroughly studied detrimental effects of wildfires on infiltration rates do not seem to be the driving force controlling soil water dynamics, at least for our study conditions. Further research will be addressed to the evaluation of these effects under distinct field conditions and different fire characteristics and to the development of a model of water relations in relation to fire disturbance.

References

Campbell R.E., Baker P.F., Ffolliot P.F., Larson F.R. & Avery C.C. (1977). Wildfire effects on a ponderosa pine ecosystem. An Arizona case study. U.S.D.A. Forest Service Research Paper RM- 191. Rocky Mountains Forest and Range Experimental Station. Fort Colins. Canadell J., Jackson R.B., Ehleringer J.R., Mooney H.A., Sala O.E. & Schulze E.-D. (1996). Maximum root depth of vegetation types at the global scale. Oecologia 108: 583-595.

95

Canadell J. & Zedler P.H. (1995). Underground structures of woody plants in mediterranean ecosystems of Australia, California and Chile. In Arroyo M.T., Zedler P.H. & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecosystems in Chile, California and Australia. Springer Verlag, Berlin, 177-210 pp. DeBano L.F. (1966). Formation of a non wettable soils involves heat transfer mechanism. General Technical Report PSW-132. U.S.D.A. Forest Service Research Notes PSW-46. Pacific Southwest Forest and Range Experimental Station, Berkeley. DeBano L.F. (2000). Fire-Induced Water Repelency: An Erosional Factor in Wildland Environments. In Proceedings of Conference Land Stewardship in the 21st century. The Contributions of Watershed Management. Tucson, U.S.A., 307-310 pp. Ferreira A.D. (1990). Fire Effect on Soil Water Dynamics. Proceedings of the 1st International Conference on Forest Fire Research. Coimbra, Portugal, C07-1. Klock G.O. & Helvey J.D. (1976). Soil water trends following wildfire on the Entiac Experimental Forests. Proceedings of the 15th Tall Timbers Fire Ecology Conference Tall Timber Research Station, Talahassee, U.S.A., 193-200 pp. Mazzoleni S. & Esposito A. (1994). Vegetation regrowth after fire and cutting of mediterranean macchia species. In Trabaud L., Prodon R. (eds.), Fire in Mediterranean Ecosystems. Commission of European Communities, Brussels, 87-99 pp. Mazzoleni S. & Pizzolongo P. (1990). Post-fire regeneration patterns of mediterranean shrubs in the Campania Region, Southern Italy. In Goldammer J.G., Jenkins M.J. (eds.), Fire in Ecosystems Dynamics. SPB Academic Publ., The Hague, 43-51 pp. Midoun M., Picard C., Prosper-Laget V. & Rebattu L. (1998). Modification of hydrous-physical soil behaviour after the passage of a fire. Proceedings of the 3rd International Conference on Forest Fire Research. Luso, Portugal, 1687-1706 pp. Pyne S.J., Andrews P.A. & Laven R.D. (1996). Introduction to Wildland Fire. John Wiley & Sons. New York. 2nd edition. Redmann R.E. (1978). Plant and soil water potentials following fire in a northern mixed grassland. Journal of Range Management 31: 443-445. Rego F.C. & Botelho H.S. (1992). Soil water regimes as affected by prescribed fire in young Pinus pinaster forests in Northern Portugal. In Trabaud L., Prodon R. (eds.), Fire in mediterranean ecosystems. Commission of European Communities, Brussels, 423-432 pp. Saxton K.E., Rawls W.J., Romberger J.S. & Papendick R.I. (1986). Estimating generalized soil-water characteristics from texture. Soil Science Society of America Journal 50: 1031-1036. Soto B. & Diaz-Fierros F. (1997). Soil water balance as affected by throughfall in gorse (Ulex euroapeus, L.) shrubland after burning. Journal of Hydrology 195: 218-231. Tiedmann A.R., Conrad C.E., Dietrich J.H., Hornbeck J.W., Megahan W.F., Viereck L.A. & Wade, D.D. (1979). Effects of Fire on Water. U.S.D.A. Forest Service General Technical Report WO-10. Wells C.G., Campbell R.E., DeBano L.F., Lewis C.E., Fredriksen R.L., Franklin E.C, Froelich R.C. & Dunn P.H. (1979). Effects of Fire on Soil. U.S.D.A. Forest Service General Technical Report WO-7. Zwolinski M.J. (2000). The role of Fire in Management of Watershed Responses. In Proceedings of Conference Land stewardship in the 21st Century. The contributions of watershed management,. Tucson, U.S.A., 367-370 pp.

96

6 MODELLING SOIL WATER DYNAMICS IN A MEDITERRANEAN SHRUBLAND5 Abstract: A new soil water model was used to interpret the results of two extensive sets of measurements performed in 2000 and 2001 in a mediterranean shrubland in Portugal. One set corresponded to undisturbed vegetation (control plot) and the other to adjacent vegetation burned by an experimental fire in the 4th of June 2001 (burned plot). The variability in soil water was very high at the surface layers and decreased for deeper soil levels. The comparison of the soil hydraulic properties predicted by the model against real data provided reasonable agreements in both control and burned plots with a coefficient of determination of 0.82. The comparison of measured and modelled soil water measurements for 6 depths in both plots provided coefficients of determination ranging from 0.85 to 0.69. When assessing the effects of fire, the model and actual data showed the same basic trend (r2=0.69), revealing an increase in soil water storage. Key words: Soil water dynamics, mediterranean shrublands, modelling, soil hydraulic properties, fire.

5 Submitted to Annals of Forest Science by Giannino F., Silva J.S., Amato M., Mazzoleni S., Rego F.C. & Magalhães M.C. 2003.

97

6.1 Introduction In mediterranean ecosystems the basic limiting factor for plant growth is the availability of water (Daget, 1977). In these conditions the distribution of rain along the year and particularly during the dry season is critical for plant survival and plant growth. Therefore, a comprehensive understanding of the mechanisms involving mediterranean ecosystems has necessarily to include the study of soil water dynamics. Given the difficulties inherent to the acquisition of soil water data along the soil profile and during long periods of time, the possibility of using models for the simulation of soil water dynamics, is undoubtedly attractive. Soil water models have been developed as standalone models, but also many times as components of larger models of crop/ecosystem dynamics. This latter application partly explains the profusion of soil water models developed during the last two decades. Most of these models were developed for crops and forest ecosystems, many of them as modules of more comprehensive models, also including carbon and nutrient flows. Reviews on ecosystem models including soil water dynamics can be found in Ågren et al. (1991), Tiktak & Grinsven (1995), Ryan et al. (1996), and Waring & Running (1998). One of the problems which a soil water modeller has to solve is the establishment of soil hydraulic properties. This is basically a problem of finding how soil water potential varies with the soil water content. This relationship is fundamental for determining the different water flows in the soil. With this purpose many authors have developed empirical relationships (Mualem, 1976; Gupta & Larson, 1979; Van Genuchten, 1980; Saxton et al., 1986;Vereecken et al., 1989) which are supposed to apply to a wide range of soil types. Another fundamental component of soil water models is the simulation of evapotranspiration and its relationship with water uptake by plants and with evaporation. Also for evaprotranspiration there are multiple solutions which can be implemented in a soil water model (e.g. Makkink, 1957; Monteith, 1965; Priestley & Taylor, 1972). Other components may or not be included in a soil water model depending on its importance for the target situations to be simulated. This is the case of the role of the canopy and the litter layers on water interception before it reaches the soil, the simulation of runoff, the simulation of water leaving the system as seepage or even the contribution of the snowpack to soil water recharge. Also for modelling these processes different alternatives can be found in the literature in terms of relationships representing each one of them.

98

Few authors have been concerned with the simulation of the dynamic mechanisms of mediterranean ecosystems (e.g. Gracia et al., 1999; Sabaté et al., 2002) in particular in what concerns the soil water dynamics. To our knowledge none of the soil water models described in the literature was developed or validated using data from mediterranean shrublands. This lack of scientific information concerning the mediterranean ecosystems was at the origin of the development of the European project ModMED - Modelling Vegetation Dynamics in Mediterranean Ecosystems (DG XII). The present paper is a direct result of the work developed within the frame of this project and it approaches different aspects of soil water dynamics using data from a shrubland in Central Portugal. This paper aims to: i) present a new model of soil water dynamics designed for mediterranean ecosystems, ii) test the proposed model using two data sets from experimental plots in a mediterranean maquis in Portugal, iii) describe and interpret the soil water dynamics of the studied shrubland, including the disturbance caused by an experimental fire. 6.2 Methods

The model

Overview The model SWADY – Soil Water Dynamics, presented here, was developed using the modelling software Simile ver. 2.2 (Simulistics Ltd., Edinburgh). Details about this modelling environment can be found in Muetzelfeldt & Taylor (2001) and at http://www.ierm.ed.ac.uk/simile/. The model basically calculates the water content at different soil layers along a period of time using a daily time-step. For this purpose the soil is divided into layers each one defined by specific characteristics: thickness, organic matter, bulk density and texture. Two sub-models allow to compute the fluxes to and from each soil layer. The first sub-model computes the soil water retention curve according to the soil texture and organic matter content of each layer, whereas the other sub-model computes the evapotranspiration processes. The water inputs and outputs for the compartment Soil water content of each defined layer are: surface infiltration (only to the top layer), drainage-in (from each layer above), drainage-out (to each layer below), evaporation (from the top layer) and water uptake by plants (from all defined layers which include roots). The model diagram is shown in Fig. 6.1 and the input

99 variables necessary to run the model are listed in Table 6.1. The equations used in the model are grouped according to the two sub-models and the five different flows.

Soil water retention curve The soil water retention curves for each layer are estimated using the Mualem- Van Genuchten equation (Mualem, 1976; Van Genuchten, 1980) which allow to obtain the soil water content (θ) as a function of the pressure head (h). This equation is defined as:

θs − θr θ()h = θr m (1) ()1+ αh n where 3 -3 θr residual soil water content in m m (the soil water that is not bound by capillary forces when h and dθ dh become very small)

3 -3 θs saturated soil water content in m m α shape parameter in m-1 n dimensionless shape parameter m equal to 1−1 n The parameters for this equation are estimated using the empirical pedotransfer functions described in Wösten et al. (1999) which require information on texture, bulk density and organic matter content. As a result, each defined soil layer is characterised by a specific water retention curve. Threshold values for θ are computed to establish conditions for water uptake, drainage and surface infiltration.

Auxiliary variables

DrainageLayers Differencemm ThetaAll DifferenceTheta mmMax7Lay Climatic Input TemperatureRH Radiation Precipitation Wind ThetaMax

n Water Input SOIL PARAMETERS Uptake Clay Penman Monteith Vegetation Input EP ETP Org Matter LAI Drainage Out Silt Evaporation Bulk Dens SWC

Layers Input

TP Drainage In Potential ThetaSat Layer Thickness ThetaLim Traspiration m Theta

Root Density

Soil layers

Fig. 6.1 Schematic diagram of model SWADY drawn using the modelling environment Simile. The compartment SWC represents the Soil Water Content, circles with a cross represent variables, thick arrows represent flows to and from the compartment and the curved thin arrows represent influences between the different model entities. 100

101

Table 6.1 List of input variables.

Type of input Variable Units Climatic inputs Air temperature °C Air relative humidity % Radiation W m-2 Rainfall mm Wind m s-1 Vegetation input Leaf area index m2 m-2 Soil layer inputs Layer thickness m Root length density (% of total) %, (cm cm3) Bulk density g cm-3 % clay % % silt % Organic matter %

Evapotranspiration processes Potential evapotranspiration (Etp) is estimated using the climatic inputs of Table 6.1 and the FAO-Penman-Monteith equation (Allen et al., 1998). The partitioning between potential soil evaporation (Ep) and potential transpiration (Tp) is obtained according to the method proposed by Ritchie (1972) using the leaf area index (LAI). Tp was computed as follows: LAI ≤ 3 Tp = Etp(1− e −0.5LAI ) (2) LAI > 3 Tp = Etp Then Ep is computed as the difference between Tp and Etp.

Water input Before rainfall reaches the soil surface, the model considers the influence of canopy interception. This is computed as a proportion of LAI using an interception coefficient equal to 0.1 mm LAI-1day-1 (Hillel, 1998). The remaining precipitation is considered equal to throughfall plus stemflow. Then water enters the soil only until saturation. Water above saturation is lost from the system as runoff.

102

Drainage-in For the top layer it is the result of flow Water input. For all other layers it is a direct consequence of the Drainage-out flow from the above layer.

Evaporation Soil evaporation is only considered for the top layer of soil and only for a water content above θlim. This soil water threshold corresponds to h=1020 cm. Actual evaporation is computed from Ep using a logistic-type reduction function.

Uptake Water uptake by plants is calculated for each layer by partitioning the overall transpiration according to the relative root length present in that layer. Transpiration is zero when water content drops to θlim, as calculated for each layer using the pedotransfer functions.

Drainage-out Drainage-out is computed using a simplified approach. For all layers drainage- out occurs for all moisture in excess of saturated soil water content (θs). When soil moisture θi is below θs, (unsaturated soil conditions) the amount of water (di) leaving a layer i with a thickness wi to the layer i+1 is equal to: ()θ − θ w d = i i+1 i (3) i 2

where θi and θi+1 represent the soil water content of layers i and i+1, respectively. Drainage-out from the bottom layer occurs when the pressure head drops bellow 50 cm, and is lost from the system as deep percolation.

Study site The model was tested using data collected at Tapada Nacional de Mafra in the Central West Region of Portugal. Tapada Nacional de Mafra is a protected area with 827 ha, about 30 km Northwest of Lisbon and 12 km East from the coast (38º 58’ 30’’ N and 9º 15’ 52’’ W). The lowest altitude is 90 m and the highest is 358 m. Soils are humic cambissols (FAO classification) derived from sandstone. According to laboratory analysis using samples from the study site, texture was classified as a sandy loam, although 5 samples (out of 48) have been classified as loams and 2 samples have been classified as silt loams (FAO classification). Mean annual precipitation is 798 mm and

103 mean annual temperature is 14.6 ºC. June, July and August are the three driest months accounting for only 3.1 % of total annual rainfall. The study area is a maquis-type shrub community dominated by Erica scoparia L. and Erica lusitanica Rudolphi, both species being hereafter referred to as Erica. Other important species are Ulex jussiaei Webb, Rubus ulmifolius Schott. and Pteridium aquilinum (L) Kuhn. Two sets of data were obtained: one from undisturbed vegetation (control plot) and another from adjacent vegetation burned by an experimental fire in the 4th of June 2001 (burned plot).

Vegetation data Vertical root length distributions were determined using core samples. For each plot, 5 vertical transects were established each one consisting of 7 depths (10, 20, 30, 40, 60, 80, and 100 cm) for core extraction. Root length density (cm.cm-3) was determined by counting the number of root intersections from each sample on a 1 cm grid, using the line intercept method (Marsh, 1971; Tennant, 1975). Extrapolation to deeper soil layers was achieved by adjusting a simple exponential function of type: y = a.d b (4) where y is the estimated root length density, a and b are adjustable parameters and d is depth. Leaf area index (LAI) was estimated for Erica (the dominant species) as 2.8. This value was obtained using the weight of 5 plants and then estimating the weight and the surface of leaves using relationships from d’Armand et al. (1993) and Fernandes & Pereira (1993) for Erica arborea. This method was preferred to the use of direct measurements since for these species a considerable amount of radiation is intercepted by small twigs and not by leaves. At the burned plot plant growth was followed after fire by measuring the average plant height.

Validation data For each plot, soil moisture was measured at 4 access tubes. All 8 tubes were disposed linearly along a contour line on a 15 % slope. Within each plot tubes were separated by 8 meters, and the two plots were separated by 27 meters. On each tube, soil water content (% volume) was measured using a TDR (Time Domain Reflectometry) equipment consisting on a measuring device (TRIME-FM3; IMKO GmbH, Ettlingen) connected to a tube-access probe (TRIME-T3). The probe was driven inside the tubes inserted in the soil 2 m deep, in order to measure moisture at different soil layers. Tubes were inserted with and angle of 30º from the vertical in order to minimise the influence

104 on root development and water flows in the soil (Maertens & Clauzel, 1982; Merril, 1992). For each tube two crossed measurements were performed at 15 cm, 30 cm, 50 cm, 90 cm, 130 cm and 170 cm, from May 2000 to December 2001. For modelling purposes layer thickness were inputted as: 7.5 cm-22.5 cm; 22.5 cm-40 cm; 40 cm-70 cm; 70 cm-110 cm; 110 cm-150 cm and 150 cm-190 cm. The layer 0 cm-7.5 cm could not be measured given the limitations of the TDR technique used. The whole study period represented a total of 587 days between the 13th of May 2000 and the 20th of December 2001. Within this period, soil water content was measured for a total of 318 days. As a rule, measurements were more frequent (daily, whenever possible) during rainy periods and less frequent during periods with no precipitation (weekly, during summer). The largest interval between consecutive measurements was 17 days in July 2000, because of equipment problems.. Meteorological data was obtained on a daily basis along the whole study period using a Weather Monitor II station (Davis Instruments Co., San Francisco) located 2 km away from the experiment. In order to verify the suitability of the pedrotransfer functions and the Mualem- van-Genuchten equation to determine the soil water retention curve to our specific case, soil samples were submitted to different pressures using a Richards plate. Soil water content was gravimetrically determined at different levels of pressure head in order to obtain the soil water retention relationships for 3 x 6 samples (6 depths replicated 3 times) at each plot.

Model evaluation

Water retention curves In order to check the accuracy of the simulated soil water retention curves obtained with the pedotransfer functions and the Mualem-Van Genuchten equation, the resulting water content values were plotted against measurements obtained in the laboratory using the same values of pressure head.

Soil water content The SWADY model was run using actual data as starting values for the first day. Model simulations for both sets of data (control and burned plots) were plotted against actual data and linear regressions were established to evaluate the model. The time series of actual data was the average of 4 replications (4 tubes).

105

Predicting the effect of fire The performance of the model was also evaluated at predicting the effects of fire on soil water storage. In order to perform this simulation, LAI recovery after fire was roughly estimated. Based on ground cover estimations, it was assumed a maximum recovery up to 0.8 cm2.cm-2, due to the rapid colonisation of the burned plot by P. aquilinum fronds and to the resprouting of most Erica plants soon after fire. This recovery was assumed to follow a typical logistic growth curve similar to the one found for plant height after the fire experiment. The effect of fire on soil water was assessed using the procedure described in Silva et al. (2002) for the same study site. Two measuring periods were considered: a reference (pre-fire) period from the 4th of June 2000 to the 20th of December 2000 and a treatment (post-fire) period from the 4th of

June 2001 to the 20th of December 2001. Differences (Di) between soil moisture measurements at each day (i) of the treatment period (Mti) and the homologous Julian day of the reference period (Mri), were obtained at the two plots (Di=Mti-Mri). The use of moisture differences from two years to compare plots was preferred to the simple use of raw data from one year. In fact given the reduced number of replications, even small variability within plots would have been sufficient to hide the effects of fire. Values of

Dbi (moisture differences in the burned plot) and Dci (moisture differences in the control plot) were converted into mm of soil water and estimated down to 180 cm depth, by integrating measurements using 20 cm increments. Since the distinct meteorological conditions of the two periods have also accounted for part of Dbi, these values were compared with Dci. Given that meteorological conditions were similar in both plots, the difference between Dbi and Dci (Si=Dbi-Dci) was assumed to correspond to the net effect of fire on soil water storage. This procedure was used both for the measured and the modelled data and the resulting time series was compared. Dry season was defined as the period from the 4th of June (burning experiment) to the day of the first September rain event >10 mm (19th of September in 2000 and 21th of September in 2001). 6.3 Results

Soil water measurements The precipitation patterns along the whole study period were very different when comparing the rainy season of 2000 (561.4 mm) with the same period of 2001 (242.6 mm). This lead to different soil water recharge patterns in both periods. On the contrary,

106 the amounts of precipitation during the dry season were very similar in both years (29.3 mm and 25.2 mm, respectively). In both plots and in both years, soil moisture followed a general pattern typical of mediterranean climates, which can be described as a markedly decreasing trend along the dry season followed by a sudden increase as the wetting front reached the different soil layers along the rainy season. In average the wetting front took between 44 (15 cm) and 74 days (170 cm) to reach the different depths after the first significant rain event (>10 mm) marking the beginning of the rainy season in September. This pattern varied according to each of the studied depths. According to Table 6.2, at 15 cm soil water content presented the highest variation in both the control and the burned plots (SD=9.0% and SD=10.6%, respectively), whereas at 170 cm it presented the lowest variation (SD=3.7% and SD=3.5%, respectively). At 15 cm, soil water content ranged from 15.4% to 50.0% at the control plot and from 15.3 to 51.6% at the burned plot. At 170 cm, soil water content ranged from 20.3% to 36.5% at the control plot and from 22.0% to 35.9% at the burned plot.

Table 6.2 Descriptive statistics of soil water measurements (% vol.) representing 318 days between May 2000 to December 2001, for the control and the burned plots. SD – Standard Deviation; SE – Standard Error.

Plot Depth Mean SD SE Max. Min. Range Control 15 cm 31.5 9.0 0.5 50.0 15.4 34.6 30 cm 33.3 7.6 0.4 46.8 21.4 25.4 50 cm 30.8 6.8 0.4 40.6 20.1 20.5 90 cm 33.0 6.5 0.4 42.2 22.7 19.5 130 cm 29.3 4.7 0.3 35.6 21.5 14.1 170 cm 28.6 3.7 0.2 36.5 20.3 16.2 Burned 15 cm 32.4 10.6 0.6 51.6 15.3 36.2 30 cm 32.4 8.6 0.5 45.1 17.9 27.3 50 cm 33.5 6.5 0.4 43.0 22.6 20.4 90 cm 30.2 5.0 0.3 37.5 21.4 16.2 130 cm 27.1 3.9 0.2 33.0 19.4 13.6 170 cm 28.8 3.5 0.2 35.9 22.0 14.0

107

Model evaluation

Water retention curves Fig. 6.2 shows the graph of the measured vs. modelled data for the soil water retention curves. It includes both plots, all depths and three points of the curves (pressure head (h) equal to: 15700 cm, 100 cm and 1 cm). The linear regression slope was 0.98, the intercepts was -6.70 and the coefficient of determination was 0.82.

Soil water content Graphs of modelled and measured soil water contents are shown in Fig. 6.3 and Fig. 6.4, together with threshold lines, representing the wilting point (h=15700 cm), field capacity (h=100 cm) and saturation (h=1 cm), as obtained by the model. Each graph represents a total of 587 days between the 13th of May 2000 and the 20th of December 2001. Within this period, soil water content was measured in a total of 318 days. Measured data were compared with the corresponding modelled data using linear regressions, and the results of this analysis are shown in Table 6.3. The coefficients of determination ranged from 0.85 to 0.74 for the control plot and from 0.78 to 0.69 for the burned plot. Values of regression slopes ranged from 0.94 to 0.58 for the control plot and from 0.71 to 0.49 for the burned plot. Values of regression intercepts ranged from 6.73 to 0.11 for the control plot and from 12.74 to 8.82 for the burned plot. There was a reasonably good fit between both data sets although with a general underestimation of the soil water content, as expected from the relationships observed for the soil water retention curves (see Fig. 6.2). In general the modelled data sets presented lower levels of variability than the actual data. When comparing the modelled data series with the modelled soil water content thresholds (the horizontal lines in Fig. 6.3 and Fig. 6.4), we verify that, with few exceptions, values remained between the two extremes (saturation and wilting point. On the other hand soil water content has exceeded the soil saturation capacity at different depths during the 2000/2001 rainy period. According to the simulation exercise, this happened especially at deeper layers (50 cm and 130 cm both in the control plot and the burned plots). The deepest layers (130 cm and 170 cm) revealed considerably high minimum levels of soil water content. The autumnal minimum was specially high in the burned plot, after the experimental burning.

108

80

70 y = 0.98x - 6.70; r2 = 0,82

60

50

40

30

20

10 Modeled soil water content (% vol.) (% content Modeled soil water

0 0 1020304050607080 Measured soil w ater content (% vol.)

Fig. 6.2 Comparison between soil water content values obtained from laboratory-determined water retention relationships (measured) and the corresponding values obtained using the soil water retention sub-model (modelled) based in the method proposed by Wösten et al., (1999) for computing the parameters of the Mualem-Van-Genuchten equation. Soil water content values correspond to three pressure head levels applied to samples from six different depths, from both plots. The solid line represents the linear regression (n=36) and the broken line represents a reference y=x relationship.

Table 6.3 Results of linear regressions (n=318) for comparison between measured (abscissa) vs. modelled (ordinate) soil water data for the two plots and each depth.

Plot Depth Slope Intercept r2 Control 15 cm 0.94 0.11 0.74 30 cm 0.88 1.70 0.80 50 cm 0.76 2.57 0.85 90 cm 0.58 6.12 0.80 130 cm 0.63 6.73 0.81 170 cm 0.66 6.48 0.75 Burned 15 cm 0.71 9.17 0.74 30 cm 0.65 8.82 0.78 50 cm 0.57 9.00 0.70 90 cm 0.61 9.00 0.74 130 cm 0.62 10.44 0.76 170 cm 0.49 12.74 0.69

109

60 15 cm 50 40 30 20 10 0

60 30 cm 50 40 30 20 10 0

60 50 cm 50 40 30 20 10 0

60 90 cm 50 40 30 20 10 0 SOIL WATER CONTENT (% volume) volume) (% SOIL WATER CONTENT

60 130 cm 50 40 30 20 10 0

60 170 cm 50 40 30 20 10 0

80

60 2000 2001

40

20 RAINFALL (mm)

0 133 177 221 265 309 353 32 76 120 164 208 252 296 340

Julian days (years 2000 and 2001)

Fig. 6.3 Water content at six different depths at the control plot. Circles represent actual measurements and the continuous lines represent model simulations. The histogram shows the rainfall during the two years (separated by a broken line) of measurements. Horizontal broken lines represent saturation, field capacity and wilting point, respectively, as obtained by the model.

110

60 15 cm 50 40 30 20 10 0

60 30 cm 50 40 30 20 10 0

60 50 cm 50 40 30 20 10 0

60 90 cm 50 40 30 20 10 0

SOIL WATER CONTENT (% volume) volume) (% SOIL WATER CONTENT 60 130 cm 50 40 30 20 10 0

60 170 cm 50 40 30 20 10 0

80

2000 2001 60

40

20 RAINFALL (mm) 0 133 177 221 265 309 353 32 76 120 164 208 252 296 340 Julian days (years 2000 and 2001)

Fig. 6.4 Water content at six different depths at the burned plot. Circles represent actual measurements and the continuous lines represent model simulations. The histogram shows the rainfall during the two years (separated by a broken line) of measurements. Horizontal broken lines represent saturation, field capacity and wilting point, respectively, as obtained by the model. The arrow indicates the date of fire.

111

180 150 120 90 60 (mm) i 30 S 0 -30 -60 150 185 220 255 290 325 360 Julian day

Fig. 6.5 Modelled (line) and measured (circles) net effect of fire in terms of soil water storage (Si).

Si=Dbi- Dci, where Dbi represents the soil water storage difference for each day (i) between the treatment period and the reference period in the burned plot and (Dci) represents the same difference for the control plot. The estimation refers to the layer 0-180 cm and results from an integration using 20 cm increments. Fire occurred at the beginning of the time series (Julian day 155, 4th of June).

Predicting the effect of fire Fig. 6.5 presents the results of simulating the net effects of fire on total water storage (Si) in the 0-180 cm soil layer compared to actual data The actual soil water storage data was taken from Silva et al. (2002). Both curves presented an increase in soil water storage along the considered period (4th of June to 20th of December), due the reduction of transpiration activity from plants. This effect was not restricted to the dry season but it continued throughout the study period (i.e. before and after Julian day 264). The maximum moisture difference in the dry season was 79.4 mm for the model and 108.2 for real data. The maximum moisture difference in the rainy season was 82.3 mm for the model and 145.2 for real data. In average both curves showed the same basic trend (r2=0.69, for the linear regression between both data sets) although the simulation had underestimated soil water storage increment by 24.5 mm (average for the whole period), when compared to actual data. 6.4 Discussion Precipitation patterns along the study period are well an example of the irregularity of the mediterranean climate specially during the rainy season. The conditions during the dry season are more regular and very little rainfall contributes for the soil water recharge until the first September rains. Plant species have evolved with these conditions and have developed adaptations to overcome the risks of water shortage. One of the adaptations is the development of deep root systems. This characteristic was found to be associated with climates with important winter rainfall (Schenk & Jackson, 2002) and dry summers, as it is the case of our study site. Important

112 rainfall in winter allows water to recharge the deeper soil layers, which can be used by deep roots during the dry season. This agrees well with the fact that the values obtained with soil moisture measurements revealed rather constant and favourable moisture conditions at 170 cm depth. This is very important for plant survival since it allows the deep rooted species to keep relatively high transpiration and growth rates during the dry season. According to this, plants have the possibility of maintaining a constant evapotranspirative flux because of deep rooting (Williams et al., 2001). The shrub community under study, is basically composed of deep rooted Erica and Ulex plants. Silva & Rego (unpublished) have found that these plants had roots going much bellow the studied profile (deeper than 240 cm in the control plot and deeper than 200 cm in the burned plot). Apparently these plants make use of their deep root system to tap water during the dry summer months when moisture is not available at more superficial layers. As soon as the dry season ends, these plants use the high root density of the upper layers to take advantage of the higher water and mineral content of the soil. In our case, the also abundant Pteridium aquilinum did not present a deep root system (Silva & Rego, unpublished) having to rely on the superficial rhizomes as long as water was available. Simultaneously with the lowest water content and the decrease in air temperature, fronds normally start decaying in September and only reappear when soil moisture conditions and temperature became again favourable in the spring. Despite the mediterranean characteristics of the climate, we should mention the mesic characteristics of the study site. In fact the lowest soil moisture values encountered at the surface layer during summer, were always far from reaching the wilting point. Nevertheless at the lowest moisture levels verified at 15 cm, both in the control and the burned plots, plants certainly have much difficulty extracting water, thus having to rely on deeper roots. Thus it is apparent that the uppermost and the lowest soil layers are critical for mediterranean plants since they represent respectively the preferential uptake zone in moist and dry conditions (Canadell & Zedler, 1995). This was confirmed by the root distribution and the maximum rooting depth found for the shrub community. In fact plants tend to optimise their root distribution as a function of soil water content distribution along the year. This have allowed some authors to use soil water models for inferring about the potential distribution of roots in the soil (Musters & Bouten, 1999; Wijk & Bouten, 2001). The empirical pedotransfer functions and the Mualem-Van Genuchten equations used in Wösten et al. (1999) for the establishment of an European Database of Soil

113

Hydraulic Properties (HYPRES) revealed an underestimation of soil moisture when compared to laboratory values at similar pressure head levels. A deviation was to be expected since the pedotransfer equations resulted from empirical laws developed to suit a wide range of european soils. This underestimation of soil water content by the modelled soil water retention curves was the basic reason for the general underestimation trend of modelled soil moisture values along the year, when compared to field measurements. This means that a different model performance (better or worst) may be expected according to the accuracy of the water retention curve estimation, when applied to other soils. In this aspect the model is no exception within the universe of soil water models, where the dependence on empirical relationships derived from soil texture to estimate hydraulic properties, constitutes an important weakness (Feddes et al., 2001). Otherwise errors could also result from measurements, contributing to increase the misfit between modelled and measured data (Musters, 1998). Errors in soil water measurements have origin in many sources, from operating errors, to changes caused by the tubes in the soil structure (Rothe et al., 1997) and preferential growth of roots (Maertens & Clauzel, 1982; Merril, 1992), or even calibrating errors. From the model side there are simplifications which may also have contributed for some deviations observed. In particular we should mention the absence of input information on root distribution dynamics along the year and the fact that some soil hydraulic characteristics are not considered such as hydraulic conductivity or the hysteresis of the water retention curves (Feddes & Koopmans, 1998). Finally the model could also consider the particular effect of the litter and the duff layers in soil infiltration and evaporation. In fact the results (measured and modelled) obtained for the first depth (15 cm) are not representative of the processes occurring close to the surface. Together with the existence of specific mechanisms controlling water flows influenced by the existence of a litter and a duff layer, this region of the soil is highly and directly influenced by all meteorological agents. The existence of a wide range of different mechanisms influencing the inflows and the outflows of water, leads to a high variability of this first layer in terms of soil water content, which could not be simulated by our model. In modelling terms it acted essentially as a buffer layer, which is definitely a considerable simplification considering its specificity in terms of the water balance. This buffer layer was responsible for example for the delay between the first significant September rain event and the arrival of the wetting front at 15 cm depth. However in general terms the model was able to reproduce the soil water dynamics as

114 obtained with the measurements. In particular we should refer the results obtained for the simulation of the fire effect on soil water storage. These results have confirmed that the major short term fire effect in terms of soil water changes, is the reduction of plant transpiration. Thus both the model and the actual data revealed that the more thoroughly studied effects of fire such as runoff, lower infiltration, and higher evaporation do not seem to be the driving force controlling short term fire effects in terms of soil water dynamics. Although we may consider the simulation results quite encouraging, further developments of the SWADY model are to be expected. These should include a more soil-physics-based approach in order to provide a better simulation of the soil water flows according to different soil hydraulic characteristics. With this purpose a further development is planned to include the widely applied Richards equation (Feddes & Koopmans, 1998; Hillel, 1998). Another important improvement will consist in the adoption of modified pedotranfer empirical functions allowing to obtain a better estimate of soil hydraulic characteristics.

References

Ågren G.I., McMurtrie R.E., Parton W.J., Pastor J. & Shugart H.H. (1991). State-of-the-art of models of production-decomposition linkages in conífer and grassland ecosystems. Ecological Applications 1: 118-138. Allen R.G., Pereira L.S., Raes D. & Smith M. (1998). Crop evapotranspiration. Guidelines for computing crop water requirements. FAO Irrigation and Drainage Paper 56, FAO, Rome. Canadell J. & Zedler P.H. (1995). Underground structures of woody plants in mediterranean ecosystems of Australia, California and Chile. In Arroyo M.T., Zedler P.H. & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecosystems in Chile, California and Australia. Springer Verlag, Berlin, 177-210 pp. d’Armand D., Etienne, M., Legrand, C., Marechal, J. & Valette, J.C. (1993). Phytovolume, phytomasse et relations strucuturales chez quelques arbustes méditerranéens. Annales des Sciences Forestiéres 50: 79-89. Daget P. (1977). Le bioclimat mediterranean: Characteres generaux, modes de characterisation. Vegetatio 34: 1-20. Feddes R.A & Koopmans R.W. (1998). Agrohydrology. Wageningen Agricultural University, Wageningen. Feddes R.A., Hoff H., Bruen M., Dawson T.E., de Rosnay P. Dirmeyer P., Jackson R.B., Kabat P., Kleidon A., Lilly A. & Pitman, A.J., (2001). Modeling root water uptake in hydrological and climate models. Bulletin of the American Meteorological Society 82: 2797-2809. Fernandes P. & Pereira J.P. (1993). Caracterização de combustíveis na Serra da Arrábida. Silva Lusitana 1: 237-260. Gracia C.A., Tello, E., Sabaté, S., Bellot, J. (1999). Gotilwa: an integrated model of water dynamics and forest growth. In Rodà F., Retana J., Gracia C.A. & Bellot J. (eds.), Ecology of Mediterranean Evergreen Oak Forest. Springer-Verlag, Berlin, 163-178 pp.

115

Gupta S.C. & Larson W.E. (1979). Estimating soil water characteristics from particle-size distribution, organic matter percent and bulk density. Water Resources Research 15: 1633-1635. Hillel D. (1998). Environmental soil physics. Academic Press, San-Diego, 2nd edition. Maertens C. & Clauzel Y. (1982). Premières observations sur l’utilisation de l’endoscopie dans l’étude de l’enracinement in situ de plantes cultivées (Sorghum vulgare et Lolium multiflorum). Agronomie 2: 677-680. Makkink G.F. (1957). Testing the Penman formula by means of lysimeters. International Journal of Water Engineering 11: 277-288. Marsh B. (1971). Measurements of length in random arrangements of lines. Journal of Applied Ecology 8: 265-267. Merril S. (1992). Pressurized-wall minirhizotron for field observation of root growth dynamics. Agronomy Journal 84: 755-758. Monteith J.L. (1965). Evaporation and environment. In State and Movement of Water in Living Organisms. Proceedings of the 19th Symposium of the Society of Experimental Biology. Cambridge University Press, 205-234 pp. Mualem Y. (1976). A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12: 513-522. Muetzelfeldt R.I. & Taylor J. (2001). Developing forest models in the Simile visual modelling environment In Proceedings of the IUFRO 4.11 conference on Forest biometry, modelling and information science. Greenwich, UK. Musters P.A.D. & Bouten W. (1999). Assessing rooting depths of an Austrian pine stand by inverse modelling variability in maps of soil water contents. Water Resources Research 35: 3041-3048. Musters P.A.D. (1998). Temporal and spatial patterns of root water uptake in an Austrian pine stand on sandy soil. PhD thesis. University of Amsterdam. Priestley C.H.B. & Taylor R.J. (1972). On the assessment of surface heat flux and evaporation using large scale parameters. Monthly Weather Review 100: 82-92. Ritchie J.T. (1972). Model for predicting evaporation from a row crop with incomplete cover. Water Resources Research 8: 1204-1213. Rothe A., Weis W., Kreutzer K., Mathies D., Hess U. & Ansorge B. (1997). Changes in soil structure caused by the installation of time domain reflectometrty probes and their influence on the measurement of soil moisture. Water Resources Research 33: 1585-1593. Ryan M.G., Hunt E.R., McMurtrie R.E., Ågren G.I., Aber J.D., Friend A.D., Rastetter E.B., Pulliam W.M., Raison R.J. & Linder S. (1996). Comparing models of ecosystem function for temperate conifer forests. I. Model description and validation. In Breymeyer A.I., Hall D.O., Melillo J.M. & Ågren G.I. (eds.), Global change: effects on coniferous forests and grasslands. Wiley, New York, 313-362 pp. Sabaté S., Gracia C.A. & Sánchez A. (2002). Likely effects of climate change on growth of Quercus ilex, Pinus halepensis, Pinus pinaster, Pinus sylvestris and Fagus sylvatica forests in the Mediterranean region. Forest Ecology and Management 162: 23-37. Saxton K.E., Rawls W.J., Romberger J.S. & Papendick R.I. (1986). Estimating generalized soil-water characteristics from texture. Soil Science Society of America Journal 50: 1031-1036. Schenk H.J. & Jackson R.B. (2002). Rooting depths, lateral root spreads, and belowground/aboveground allometries of plants in water-limited ecosystems. Journal of Ecology 90: 480-494. Silva J.S., Rego F.C. & Mazzoleni S. (2002). Fire effects on soil water dynamics in a mediterranean shrubland. In Proceedings of the 4th International Conference on Forest Fire Research. Luso, Portugal. Tennant D. (1975). A test of a modified line intersect method of estimating root length. Journal of Ecology 99: 995-1001. Tiktak A. & van Grinsven H.J.M. (1995). Review of sixteen forest soil-atmosphere models. Ecological Modelling 83: 35-53.

116

Van Genuchten, M.T. (1980). A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44: 892-898. Vereecken H., Maes J., Darius P.& Feyen J. (1989). Estimating the soil moisture retention characteristic from texture, bulk density and carbon content. Soil Science 148: 389-403. Waring R.H. & Running S.W. (1998). Forest ecosystems. Analysis at multiple scales. Academic Press, San Diego. 2nd edition. Wijk M.T. van & Bouten W. (2001) Towards understanding tree root profiles: simulating hydrologically optimal strategies for root distribution. Hydrology & Earth System Sciences 5: 629-644. Williams M., Law B.E., Anthoni P.M. & Unsworth, M.H. (2001). Use of a simulation model and ecosystem flux data to examine carbon-water interactions in ponderosa pine. Tree Physiology 21: 287-298. Wösten J.H., Lilly A., Nemes A. & Le Bas C. (1999). Development and use of a database of hydraulic properties of European soils. Geoderma 90: 169-185.

117

7 DISCUSSÃO E CONCLUSÕES Neste capítulo final tentam agrupar-se, de uma forma tanto quanto possível consistente e harmonizada, a discussão dos resultados e as principais conclusões do trabalho desenvolvido no âmbito da presente tese. Apesar de algumas alterações introduzidas de forma a dar uma sequência lógica à leitura, uma boa parte deste capítulo final é fundamentalmente uma assumida transcrição para Português da secção Discussion dos cinco capítulos anteriores. Deste modo, o objectivo base do presente capítulo é tão somente o de se constituir como uma súmula em Língua Portuguesa do conhecimento adquirido no âmbito deste trabalho. 7.1 Estudo das raízes de plantas individuais

Características estruturais das raízes de algumas plantas lenhosas mediterrânicas A decisão de estudar um grupo heterogéneo de plantas quer ao nível da espécie quer ao nível do estado de desenvolvimento, embora acarretando importantes limitações ao nível do tratamento estatístico, permitiu por outro lado o estudo da variabilidade exibida pelas plantas amostradas. Pensamos que este objectivo foi conseguido tendo em conta os resultados obtidos pela Análise por Componentes Principais (PCA – Principal Component Analysis) utilizando diferentes parâmetros caracterizadores dos sistemas radicais. Esta análise demonstrou que o estado de desenvolvimento dos indivíduos amostrados foi o principal factor responsável pela variabilidade exibida pelos diferentes sistemas radicais. Tal como esperado, estados de desenvolvimento mais avançados corresponderam tendencialmente a um maior comprimento total das raízes, a uma maior extensão lateral do sistema radical e a uma maior biomassa radical. De acordo com o diagrama da PCA o diâmetro médio das raízes aparece em posição ortogonal relativamente ao comprimento total das raízes, sugerindo que estas duas variáveis representam formas diferentes e não correlacionadas de distinguir os sistemas radicais do conjunto de plantas amostrado. No entanto a segunda variável foi sobretudo capaz de distinguir estados de desenvolvimento ao passo que a primeira se apresentou mais relacionada com a diferenciação das várias espécies. Embora não exista uma separação clara entre as espécies de regeneração exclusiva por semente (géneros Cistus e Lavandula) e as espécies de regeneração vegetativa (restantes géneros), as primeiras obtiveram em geral maiores valores na segunda componente principal (PC2 – Principal

118

Component 2) sugerindo que o diâmetro médio das raízes pode ser uma característica distintiva entre os dois grupos, tal como foi confirmado pelo teste estatístico. A posição simétrica da variável comprimento radical específico(SRL – Specific Root Lenght) relativamente ao respectivo diâmetro está de acordo com a utilização deste último coeficiente como indicador do diâmetro das raízes (Fitter, 1985). Um outro parâmetro distintivo entre espécies de regeneração exclusiva por semente e espécies de regeneração vegetativa parece ser a profundidade máxima de enraizamento a qual, por outro lado, não teve grande correlação com qualquer dos eixos da PCA. Aparentemente a profundidade máxima de enraizamento parece ser uma variável bastante independente dos outros parâmetros estudados, incluindo a área basal. Podemos avançar duas razões possíveis para explicar o fraco relacionamento entre a profundidade máxima de enraizamento e a área basal. Por um lado o facto de os nossos resultados sugerirem que algumas espécies (como a D. gnidium e a L. luisieri) parecerem alcançar relativamente cedo a profundidade máxima de enraizamento. Por outro lado, a cada espécie parece estar associada uma profundidade máxima de enraizamento característica para cada espécie. Em relação ao primeiro aspecto o desenvolvimento precoce das raízes em profundidade é com certeza uma vantagem competitiva e parece ser uma característica comum a muitas espécies lenhosas mediterrânicas (Canadell & Zedler, 1995). Muito embora representando apenas uma pequena fracção de todo o sistema radical, as raízes profundas têm um papel fundamental na capacidade das plantas para ultrapassar os déficits hídricos verificados durante a estação seca, em particular durante os estádios iniciais de desenvolvimento. (Canadell et al., 1996). Estas raízes conseguem alcançar camadas de solo onde o teor de água é mais elevado que nos horizontes superficiais. As raízes profundas podem ser responsáveis por mais de 75% de toda a água extraída durante a época seca (Nepstad et al., 1994). As diferenças de profundidade máxima de enraizamento parecem estar também associadas às estratégias regenerativas das diferentes espécies (Keeley, 1986; Correia & Catarino, 1994). Todos os indivíduos de C. crispus, C. salvifolius e de L. luisieri (regeneração exclusiva por semente) apresentavam sistemas radicais relativamente superficiais enquanto todas as outras espécies (regeneração vegetativa) apresentavam sistemas radicais mais profundos. De entre as espécies estudadas as plantas de L. luisieri e de D. gnidium representam duas tendências opostas. O crescimento das raízes é essencialmente horizontal no primeiro caso e essencialmente vertical no segundo. Em termos da ecologia das espécies, os valores significativamente mais altos encontrados para a biomassa radical, para o

119 diâmetro médio das raízes e para a relação raíz-parte aérea das plantas de regeneração vegetativa, podem ser interpretados apenas como uma consequência da existência de raízes profundas, apesar da fraca correlação evidenciada pela PCA. Estudos de outras regiões mediterrânicas têm confirmado a relação entre as características das raízes e as estratégias regenerativas, nomeadamente na Austrália (Bell, 2001), na California (Hellmers et al., 1955) e na África do Sul (Higgins et al., 1987). A tendência decrescente observada para a razão raíz-parte aérea e para a SRL relativamente a valores crescentes de secção basal, está de acordo com as expectativas. Diferentes autores têm descrito a existência de uma estratégia no sentido de atribuir uma fracção cada vez menor de recursos às raízes à medida que o desenvolvimento das plantas prossegue (Nobel, 1996; Bazzaz, 1997; Grace, 1997). Esta fenómeno é particularmente importante nas plantas mediterrânicas, nas quais o desenvolvimento precoce das raízes é um factor crítico para sobreviver à primeira estação seca após a germinação. A tendência decrescente observada para a SRL (Fitter, 1985), foi mais importante para as duas espécies com regeneração exclusiva por semente, o que pode revelar diferentes estratégias de crescimento radical. Estas diferenças poderão ser atribuídas a um investimento preferencial em raízes estruturais e orgãos de armazenamento de hidratos de carbono das duas espécies de regeneração vegetativa durante os primeiros estádios de desenvolvimento, contrastando com o investimento preferencial em raízes finas por parte das duas espécies de regeneração exclusiva por semente durante estas mesmas fases de crescimento. Apesar da existência de uma grande amplitude de valores da relação raíz-parte aérea correspondendo a diferentes espécies e diferentes estádios de desenvolvimento, a biomassa das raízes e a biomassa da parte aérea apresentaram uma relação semelhante e igualmente consistente com a secção basal das plantas. Este aspecto pôde ser confirmado tanto pela regressão linear como pela PCA, apesar da diversidade de espécies e de estádios de desenvolvimento das plantas utilizadas nas análises. A existência deste tipo de relação tem sido relatada em diferentes estudos com espécies arbóreas (e.g. Santantonio et al., 1977; Drexhage & Colin, 2001; Hoffmann & Usoltsev, 2001; Ranger & Gelhaye, 2001). O estabelecimento de uma única relação alométrica entre medições no caule e medições na raiz utilizando indivíduos de diferentes espécies (Santantonio et al., 1977) está de acordo com o amplamente difundido pipe model, apresentado por (Shinozaki et al., 1964). A secção do tronco parece de facto limitar o

120 desenvolvimento da raiz e da parte aérea de forma semelhante para diferentes espécies e para diferentes estádios de desenvolvimento. Algumas observações devem ser feitas relativamente à interpretação dos resultados obtidos, tendo em conta as duas estratégias regenerativas. As semelhanças encontradas ao nível dos parâmetros caracterizadores dos sistemas radicais das espécies de regeneração exclusiva por semente reflectem também a existência de relações filogenéticas estreitas entre algumas plantas (mesmo género ou mesma espécie). Deste modo, a existência de características comuns pode dever-se apenas aos laços filogenéticos existentes entre as plantas. Para contornar esta questão foram desenvolvidos esquemas de delineamento de amostragem (Nicotra et al., 2002) os quais no entanto não podem ser aplicados a situações como aquela aqui estudada, com uma diversidade de espécies disponíveis relativamente limitada. Deste modo as diferenças significativas encontradas entre os dois grupos de plantas representantes de duas estratégias regenerativas distintas, devem ser interpretadas também tendo em consideração os laços filogenéticos existentes. No entanto os nossos resultados são apoiados por resultados semelhantes noutras regiões mediterrânicas do Mundo, o que nos leva a assumir que os mesmos mecanismos adaptativos encontrados noutras regiões são também válidos para as comunidades de plantas aqui estudadas. Um dos aspectos não abordados no âmbito do presente estudo foi o estudo da influência dos factores ambientais nas características das raízes. É consensual a influência determinante dos factores ambientais nas características dos sistemas radicais das plantas (e.g. Spurr & Barnes, 1980; Kummerow, 1981; Fitter, 1996; Atkinson, 2000). No nosso caso a ausência de limitações evidentes ao desenvolvimento das raízes e a proximidade do locais onde as plantas se encontravam levaram-nos a assumir a existência de condições ambientais relativamente homogéneas, permitindo a comparação dos sistemas radicais com base na respectiva espécie e estado de desenvolvimento. No entanto devemos admitir que a existência de diferenças em termos de densidade de plantas ou em termos de micro-topografia poderão ter contribuído de forma importante para a variabilidade geral da amostra.

Distribuição das raízes de algumas plantas mediterrânicas. Proposta de um novo modelo De entre os quatro modelos testados, o MLDR revelou uma considerável flexibilidade e capacidade para se ajustar aos dados relativos à distribuição das raízes de

121 um leque alargado de plantas lenhosas. A utilização desta função matemática permitiu descrever de forma muito aproximada a distribuição vertical das raízes das plantas amostradas, em termos de biomassa e de comprimento. Da mesma forma, a utilização conjunta de um coeficiente de biomassa e de um coeficiente de comprimento de raízes, permitiu reflectir a diversidade de sistemas radicais existentes no conjunto de plantas amostradas. Muito embora esta abordagem pudesse potencialmente conduzir a um sistema de classificação empírica de sistemas radicais, tal como a apresentada por Guowei et al. (1997) tal não foi feito devido à necessidade de um conjunto de dados bastante mais extenso. Uma classificação abrangente apenas seria possível dispondo de repetições para cada espécie, para cada estádio de desenvolvimento e para diferentes condições ambientais. O enorme esforço exigido para a escavação de raízes completas em condições naturais fez com que não fosse possível conseguir este objectivo. Na utilização do modelo MLDR para o ajustamento de dados relativos à distribuição de raízes de uma comunidade de plantas com uma profundidade máxima de enraizamento desconhecida, deverá ter-se em conta o facto de que os dados referentes aos valores de maior profundidade correspondem a Yr=1. Se este aspecto for tido em conta, a utilização do modelo proposto para o ajustamento de dados de distribuição de raízes ao nível da comunidade de plantas é perfeitamente possível (ver capítulo 4). Algumas observações devem ser feitas relativamente à metodologia utilizada para obter dados de distribuição de raízes, em particular no que se refere à determinação do comprimento das raízes. A distribuição vertical do comprimento das raízes foi com certeza afectada, embora não se sabendo até que ponto, pelo processo de escavação utilizado, dada a inevitável perca de raízes finas (Wallace et al., 1974; Böhm, 1979; Caldwell & Virginia, 1989; Bengough et al., 2000) as quais são responsáveis pela maior parte do comprimento de raízes no solo. Muito embora possamos assumir que este erro se repartiu de forma idêntica em cada sistema radical, a perca de raízes finas poderá eventualmente ter contribuído para esconder parte da acumulação normal de raízes nas camadas superficiais do solo. Um outro aspecto a ter em conta é a possibilidade de terem ocorrido alterações importantes no arranjo arquitectural das raízes, em particular das mais finas, relativamente à disposição original no solo, antes da escavação. No entanto pensamos que muito poucas alterações substanciais foram introduzidas dado se tratarem de plantas lenhosas com raízes estruturais sólidas e não deformáveis. Por outro lado durante o trabalho de escavação tentou-se evitar este tipo de erros através do registo fotográfico das plantas no campo e de uma cuidadosa disposição das raízes para

122 a determinação estratificada da biomassa e do comprimento radicais. Uma outra fonte possível de erro é todo o processo de captação e processamento das imagens digitais. No entanto, os erros derivados deste processo podem resultar tanto num aumento como de uma diminuição dos valores reais a determinar (Richner et al., 2000). Apesar das óbvias limitações da metodologia utilizada, foi assumido que os erros referidos terão sido igualmente distribuídos por todas as plantas estudadas e ao longo de cada planta. Por outro lado, a utilização de valores normalizados e a natureza comparativa do presente estudo fazem-nos considerar que os erros existentes serão aceitáveis tendo em vista os objectivos em vista. De um modo geral os dados de distribuição de raízes revelaram uma separação entre espécies de regeneração exclusiva por semente e espécies de regeneração vegetativa. De entre as plantas do último grupo aquelas com raízes profundas revelaram um padrão de distribuição de raízes distinto (Daphne, Erica, Myrtus, Ulex). Em particular as espécies com tubérculo lenhoso ou lignotuber (E. lusitanica e E. scoparia) apresentaram uma elevada concentração de biomassa junto à superfície do solo. Estes resultados reflectem bem a importância destas estruturas enquanto órgãos de armazenamento de hidratos de carbono sob a forma de amido. No entanto alguns autores questionam a decisão de considerar os tubérculos lenhosos como partes integrantes do sistema radical, dado que este funciona essencialmente como órgão de armazenamento, normalmente localizado muito perto ou mesmo acima da superfície do solo, e não exactamente como uma raiz estrutural (Kummerow, 1981; Canadell & Zedler, 1995). Esta decisão é obviamente determinante nos resultados obtidos em termos da distribuição vertical da biomassa das plantas e da determinação da relação raíz-parte aérea. Duas outras espécies de regeneração vegetativa com uma elevada concentração de biomassa junto à superficie, D. gnidium e M. communis, não apresentavam um tubérculo lenhoso típico mas antes intumescências não lenhificadas, possivelmente desempenhando o mesmo papel de armazenamento de hidratos de carbono. Com a excepção do único exemplar recolhido de M. communis, todas as plantas com raízes profundas apresentavam raízes pivotantes com orientação aproximadamente vertical. Tal como já foi referido relativamente ao estudo anterior, estas raízes pivotantes têm um papel fundamental para a captação de água durante a estação seca, muito embora representem uma pequena fracção da biomassa e do comprimento total das raízes (Kummerow et al., 1990; Canadell & Zedler, 1995). A consequência deste padrão de distribuição é uma mais elevada concentração relativa de

123 raízes junto à superfície, para maiores profundidades máximas de enraizamento. A maior parte das plantas com raízes profundas apresentavam uma expansão lateral dos respectivos sistemas radicais, constituindo um sistema típico de outras regiões mediterrânicas (Canadell & Zedler, 1995) composto por raízes pivotantes e por raízes superficiais ou pastadeiras. Estudos diversos têm demonstrado que este padrão de distribuição pode ser vantajoso em regiões sujeitas a períodos de seca, estando presente em várias espécies destas regiões (Hellmers et al., 1955; Specht & Rayson, 1957; Kummerow, 1981; Krämer et al., 1996). No caso particular das duas espécies do género Erica, apesar das semelhanças do padrão de distribuição das raízes, algumas diferenças podem ser observadas devido à menor expansão lateral das raízes do único exemplar adulto de E. scoparia. As razões para esta diferença podem ter a ver com o facto de esta planta ter sido escavada numa comunidade com elevada densidade, ao passo que a planta de E. lusitanica se encontrava num local aberto. Muito embora não nos fosse possível testar esta hipótese, é sabido que as interacções entre as plantas têm um efeito determinante na distribuição das raízes no solo, tal como foi demonstrado por (Atkinson, 2000). Os sistemas radicais das espécies de regeneração vegetativa C. monogyna e R. ulmifolius, apresentam estruturas distintas quando comparadas com as restantes espécies deste grupo. No primeiro caso existem apenas raízes estruturais à superfície e quase nenhumas raízes finas, o que explica a diferença entre as curvas de biomass e de comprimento das raízes. No segundo caso apenas existem raízes finas ao longo de todo o perfil ocupado o que explica a semelhança entre as curvas de biomassa e de comprimento de raízes. O facto de se tratar de uma espécie escandente deverá explicar em parte a ausência de raízes estruturais para a sustentação da planta. Relativamente às espécies de regeneração exclusiva por semente, existe uma semelhança notável nos padrões de distribuição exibidos pelas plantas do género Cistus, todas elas com raízes pouco profundas tal como é comum nas espécies exibindo esta estratégia regenerativa (Keeley, 1986; Bell, 2001). O caso particular da L. luisieri é paradigmático já que se trata de uma espécie altamente adaptada à secura apesar de apenas dispor de um sistema de raízes superficial. A estratégia desta espécie parece ser o desenvolvimento de um sistema de raízes finas com uma ramificação extrema na profundidade máxima de enraizamento, o que deverá permitir um aumento da eficiência na extracção de água. Em termos dos diferentes estádios de desenvolvimento, apenas a distribuição do comprimento das raízes parece ter sofrido alterações como consequência do aumento da

124 profundidade máxima de enraizamento (Maxd). A ausência de diferenças consistentes em termos da distribuição da biomassa das raízes ao longo dos estádios de desenvolvimento parece dever-se em parte à grande concentração da biomassa em órgãos de armazenamento por parte das espécies de regeneração vegetativa. Pelo contrário os valores do coeficiente dl50 apresentaram um aumento consistente para todas as espécies ao longo dos diferentes estádios de desenvolvimento, como consequência dos valores mais elevados da profundidade máxima de enraízamento. 7.2 Estudo das raízes ao nível da comunidade de plantas

Distribuição das raízes de um matagal mediterrânico na Tapada Nacional de Mafra Os resultados obtidos demonstraram a existência de padrões de distribuição de raízes bastante distintos para as diferentes espécies presentes na comunidade de plantas. Estes diferentes padrões de distribuição podem estar associados a diferentes estratégias para a captação de água e nutrientes. De facto os resultados sugerem a existência de uma separação de nichos (Casper & Jackson, 1997) entre as quatro espécies associado a diferentes estratégias adaptativas. Os padrões de distribuição de raízes obtidos através do mapeamento das paredes das trincheiras reflectem sobretudo os padrões de distribuição das plantas individuais tal como puderam ser observados no local de estudo. As plantas da espécie Erica exibiam raízes pivotantes mas também raízes laterais junto à superfície. As plantas de Ulex embora também como raízes profundas não apresentavam este padrão devido à ausência de raízes laterais à superfície. Este facto fez com que o pico de contagens para esta espécie não fosse atingido junto à superfície mas sim abaixo dos primeiros 20 cm com o consequente aumento do valor de do coeficiente D50. A disposição em grupos das raízes mapeadas no fundo das trincheiras está também de acordo com a arquitectura das plantas individuais de Erica e de Ulex. Na realidade as raízes pivotantes destas plantas foram observadas crescendo praticamente na vertical, desta forma dando origem a uma disposição das raízes no plano horizontal formando grupos, supostamente referentes a uma única planta. Tal como foi referido anteriormente, estas raízes profundas são fundamentais para a absorção de água durante a estação seca. Este padrão de distribuição de raízes pivotantes é associado a climas com um Verão seco e com bastante chuva no Inverno, como é o caso do local estudado (Schenk & Jackson, 2002b). No entanto no nosso caso específico o elevado número de raízes pivotantes é provavelmente o resultado da intensa

125 competição (Atkinson, 1978) originada pela elevada densidade de plantas. Para além da competição intra-específica há que considerar igualmente a competição entre espécies diferentes. Apesar de não termos abordado este aspecto específico, vários trabalhos referem a existência de competição assimétrica em povoamentos mistos (e.g. McKay, 1988; Casper & Jackson, 1997; Leuschner et al., 2000) como resultado da diferente eficiência na exploração de recursos do solo evidenciada pelas espécies presentes. No caso das plantas de Rubus os indivíduos observados evidenciavam uma elevada densidade de raízes finas junto ao solo, mas também algumas raízes profundas. Esta disposição das raízes explica o baixo valor de D50 e a tendência decrescente obtida com as contagens verticais. No caso das raízes de Pteridium estas dispunham-se de acordo com uma rede horizontal muito densa de rizomas, junto à superfície do solo, aos quais se encontravam ligadas raízes finas relativamente dispersas. Em relação a esta espécie deve ter-se em conta a sua fenologia, caracterizada por grandes alterações sazonais (Pakeman & Marrs, 1994). Esta sazonalidade traduz-se fundamentalmente na morte das frondes durante a estação fria e no seu reaparecimento na Primavera. Estas flutuações em termos da biomassa viva da parte aérea parecem ter correspondência abaixo da superfície do solo, dada a elevada percentagem de rizomas em diferentes estádios de decomposição. Desta forma há que ter em conta, relativamente a esta espécie, que os nossos resultados dizem respeito ao período de crescimento (Junho-Agosto) e que por isso apenas reflectem as características da espécie durante este período. As diferenças na distribuição das raízes entre classes de diâmetro são em parte uma consequência da localização preferencial das raízes estruturais nas camadas mais superficiais do solo. No caso particular das raízes médias (5-10 mm) e grossas (>10 mm) existiu uma contribuição relativamente importante dos rizomas de Pteridium para os resultados obtidos. As raízes estruturais localizadas junto à superfície do solo têm uma importância fundamental na sustentação da parte aérea (Coutts, 1983; Fitter & Ennos, 1989) e na estabilidade do solo (Ziemer, 1981). Deste modo a distribuição mais uniforme e mais profunda das raízes mais finas e a distribuição mais superficial e heterogénea das raízes mais grossas não surpreende, mesmo tendo em conta que os resultados foram obtidos com quatro espécies distintas. Dado que este é, tanto quanto sabemos, o único estudo realizado sobre a distribuição das raízes de uma comunidade dominada por plantas do género Erica, é difícil dispor de outros estudos para comparação. Por outro lado, por entre a panóplia de métodos que podem ser encontrados na literatura, a maior parte baseiam-se em medições de biomassa e não em

126 contagens de raízes. Este facto é bem patente na compilação de dezassete estudos realizados em comunidades arbustivas esclerófilas levada a cabo por Schenk & Jackson

(2002a) onde quase todos os estudos se baseiam em medições de biomassa. O D50 médio destes estudos é de 19 cm o que representa um valor inferior ao do presente estudo (26 cm). Uma outra compilação de seis estudos em vegetação esclerófila presente em Jackson et al. (1997) incluindo apenas raízes finas (neste caso raízes maiores que 2 mm) forneceu um valor médio de D50 igual a 14 cm, o que é bastante inferior à nossa estimativa de 27 cm baseada em contagens, mas exactamente igual à nossa estimativa baseada em comprimento das raízes. Na verdade pudemos verificar uma diferença considerável entre os resultados obtidos com as amostragens de biomassa e o comprimento de raízes, quando comparadas com a contagem. A verdade é que dificilmente se encontra uma relação directa entre contagens e medições de biomassa ou comprimento radical (Van Noordwijk et al., 2000). Para além das limitações da metodologia utilizada, a distribuição final das contagens de raízes foi influenciada de forma decisiva pelas trincheiras 4 e 5, as quais apresentavam um horizonte A espesso. Nestas trincheiras as raízes estavam em geral mais uniformemente distribuídas, aparentemente acompanhando a distribuição da matéria orgânica ao longo do perfil. No que diz respeito à estimativa da biomassa total (viva e morta) de raízes finas, as nossas estimativas são mais elevadas que a média de seis estudos de 0.52 kg/m2 referidos por Jackson et al. (1997) e mais elevada que o valor de 0.78 kg/m2 referido por Martinez et al. (1998) referido para uma comunidade arbustiva mediterrânica em dunas no Sul de Espanha. Aparentemente a densa rede de raízes lenhificadas de Erica representa uma biomassa mais elevada que o relatado nos outros estudos. Devemos também admitir que uma pequena percentagem de raízes com diâmetro acima de 1mm possa ter sido incluída na amostragem influenciando dessa forma o resultado. Por outro lado, a estimativa do comprimento de raízes por unidade de superfície (18.7 km/m2) é comparável com a estimativa de 17.5 km/m2 (Jackson et al., 1997) referente a comunidades esclerófilas lenhosas. É notável o facto de a nossa estimativa ser mais elevada que os valores referidos em alguns estudos realizados em povoamentos florestais (e.g. Jackson et al., 1997; Vande Walle et al., 1998; Wiesenmüller, 1998). Tal não surpreende se considerarmos que a área basal da comunidade arbustiva estudada é maior que a de um povoamento florestal com uma lotação normal conduzido para produção de madeira (20-30 m2/ha para a maior parte das espécies da região temperada). Muitos autores expressam a ocupação radical do solo através do cálculo da

127 distância média entre raízes em vez de utilizar o comprimento por unidade de área ou de volume. Esta distância pode ser calculada através do inverso da densidade de raízes por unidade de volume (Miller & Ng, 1977). No nosso caso a distância entre raízes estimada para o primeiro metro de solo foi de 0.9 cm. Trata-se de um valor muito inferior aos 2.0 cm estimados por Kummerow et al. (1977) para um chaparral na California, inferior aos 2.8 cm estimados por Hoffmann & Kummerow (1978) para um matorral no Chile e ainda inferior aos valores de 1.9 cm e 1.7 cm estimados por Martinez & Rodriguez (1988) e por Martinez et al. (1998) para uma comunidade arbustiva no Sul de Espanha. A maior ocupação do solo verificada neste estudo revela diferenças importantes em termos da massa e do comprimento de raízes de um matagal tipo maquis quando comparado com outras comunidades arbustivas com menor densidade de plantas tais como as dos estudos referidos. Apesar da diferente composição florística da comunidade de plantas estudada, relativamente aos restantes trabalhos consultados, podemos especular que a existência de uma tão elevada ocupação do solo apenas é possível devido à existência de condições mais favoráveis ao crescimento, nomeadamente em termos de nutrientes e de água no solo. Deste modo parece existir um padrão distinto de ocupação do solo pelas raízes em condições mediterrânicas mais mésicas tal como no nosso caso. As nossas estimativas de profundidade máxima de enraizamento para as duas espécies de Erica são razoávelmente próximas dos valores referidos por Canadell et al. (1996) para outras ericáceas (Arbutus unedo, 350 cm; Erica arborea, 200 cm). A distribuição estimada para a profundidade máxima de enraizamento das diferentes plantas sugere que a grande maioria das raízes atinge profundidades localizadas numa faixa de 50 cm entre os 125 cm e os 175 cm, e que apenas algumas raízes atingem profundidades superiores. As razões para esta distribuição não foram analisadas no âmbito do presente trabalho, mas podemos especular que as raízes mais profundas deverão estar associadas a plantas de maiores dimensões e maior biomassa, tal como sugerido por Schenk & Jackson (2002b). Dado não termos conhecimento de estudos semelhantes com espécies do género Ulex, não podemos obter informações sobre a profundidade máxima de enraizamento destas plantas ou mesmo de outras filogenéticamente próximas. No entanto os nossos resultados sugerem de forma consistente a existência de sistemas radicais pelo menos tão profundos como os de Erica. Se aplicarmos a mesma equação obtida para ajustar os dados das plantas de Erica, às raízes de Ulex, obtemos uma profundidade máxima absoluta de 349 cm e uma

128 média para as 28 raízes mais profundas de cada trincheira, igual a 221 cm. A importância destas raízes profundas para a captação de água no Verão é confirmada nos capítulos 5 e 6 desta tese. As medições de humidade revelaram uma perca de água pelo solo no final da estação seca igual a 18 mm na camada 160-180 cm e igual a 12 mm na camada 120-140 cm. 7.3 Estudo da dinâmica da água no solo

Efeitos do fogo na dinâmica da água do solo De acordo com os resultados obtidos, a ocorrência do fogo alterou de forma notável a dinâmica da água no solo da comunidade estudada. Estas alterações foram detectadas não apenas na estação seca, durante a qual seria normal fazer-se sentir a redução drástica da evapotranspiração, mas em geral ao longo de todo o período estudado. A definição de uma estação seca e de uma estação húmida, baseada num limite de precipitação, teve como único objectivo o estabelecimento de uma a separação do período em que não existe recarga da água no solo, do período em que essa recarga se verifica. No entanto esta diferenciação teve um significado distinto para as diferentes camadas de solo consideradas, já que as camadas mais profundas mantiveram a mesma tendência na variação dos valores de humidade muito para além do fim da estação seca ao passo que a reacção das camadas mais superficiais foi muito mais imediata, como seria natural. Deste modo cada região do solo tem as suas épocas seca e húmida, dependendo da profundidade a que se encontra. Um outro aspecto a ter em conta tem a ver com a data em que foi realizado o fogo experimental. Os resultados teriam sido bastante diferentes caso o fogo tivesse tido lugar mais tarde na estação seca. Por exemplo um fogo em Setembro teria como consequências uma menor recuperação na vegetação devido ao menor tempo disponível para o crescimento e a uma menor disponibilidade de água para o crescimento. Neste caso os efeitos verificados neste estudo deveriam ocorrer essencialmente durante a estação seca seguinte tal como foi constatado por Klock & Helvey (1976) e Soto & Diaz-Fierros (1997) após fogos ocorridos em Agosto e Setembro, respectivamente. De um modo geral os diferentes padrões de dinâmica da água no solo observados durante a estação seca nas diferentes camadas de solo, podem ser explicados pelo efeito directo da evaporação, pela extracção de água pelas raízes e pelos movimentos de água no solo. Qualquer dos três processos contribui para uma diminuição do teor de água à superfície comparativamente com as outras camadas de

129 solo. Tal como pôde ser verificado através dos dados de distribuição das raízes, a maior parte destas ocorre nas camadas superficiais. Em particular a espécie com resposta mais imediata e vigorosa ao fogo, a Pteridium aquilinum, possui rizomas muito superficiais. Por outro lado as camadas mais profundas não sofreram o efeito da evaporação e as únicas raízes presentes eram raízes de Erica e Ulex (ver capítulo 4). Estes aspectos aparentemente podem contribuir para explicar o facto de nas camadas superiores de solo se verificar uma diminuição importante do teor de água no solo, especialmente após o crescimento súbito das frondes de P. aquilinum em Julho, ao passo que em profundidade o teor de humidade foi sofrendo um decréscimo muito gradual e quase imperceptível a partir de certa altura. Nas camadas superiores tal correspondeu a um decréscimo de Dbi, presumivelmente devido ao crescimento do coberto vegetal (P. aquilinum e Erica). Pelo contrário as camadas mais profundas revelaram valores progressivamente mais elevados de Dbi, o que é conclusivo acerca da importância da extracção de água pelas raízes profundas durante a estação seca do ano de referência. Tal está de acordo com o que é referido por diversos autores sobre o papel das raízes profundas no solo (Canadell & Zedler, 1995; Canadell et al., 1996). Durante a estação húmida, diferentes padrões de recarga da água no solo podem ser associados a diferentes camadas de solo. Estes padrões reflectem fundamentalmente o momento em que a frente de humedecimento demora a atingir as diferentes profundidades. Novamente se pôde observar uma muito mais rápida resposta á queda de precipitação por parte das camadas superiores (0-40 cm) em ambas as parcelas. No entanto as diferenças entre parcelas nestas camadas não foram tão consistentes durante a estação húmida, dada a oscilação entre valores positivos e negativos de Si. Por outro lado dada a diversidade de factores que actuam nas camadas superiores torna-se difícil tentar determinar qual a causa essencial para os mais baixos valores de Si. neste período e nesta região do solo. Muito embora não o tenhamos podido provar, podemos especular que a exposição do solo na parcela queimada terá sido responsável por uma mais elevada evaporação durante o período pós-fogo durante os invulgarmente secos meses de Novembro e Dezembro de 2001. Uma importante consequência dos nossos resultados tem a ver com a importância do teor de humidade no solo no crescimento das plantas durante o período pós-fogo. O elevado teor de água verificado após o fogo aparentemente permite explicar o rápido aumento do coberto vegetal na parcela queimada. Condições semelhantes pós- fogo poderão estar na origem dos elevados potencial hídrico e taxa de transpiração

130 verificados em espécies de regeneração por semente e em espécies de regeneração vegetativa (Clemente, em prep.) numa outra comunidade arbustiva estudada na Serra da Arrábida. No entanto os resultados deste estudo também sugerem que a invulgar seca verificada no final do Outono de 2001 está relacionada com teores comparativamente baixos de humidade junto à superfície, na parcela queimada, embora a mesma área se tenha caracterizado por um teor de humidade superior durante o período seco. Períodos invulgarmente secos têm sido referidos como factores críticos que explicam, em interacção com o fogo, a mortalidade das plantas durante o período de recuperação da vegetação (Mazzoleni & Pizzolongo, 1990). Experiências em situações controladas mostraram que as plantas de Erica arborea são sensíveis á existência de condições de humedecimento durante a rebentação, após perturbação (Mazzoleni & Esposito, 1993). Os nossos resultados confirmam o interesse das interacções entre o fogo e as condições climáticas relativamente à dinâmica da vegetação. Alguns trabalhos relatam uma diminuição no teor de água no solo devido ao fogo (Campbell et al., 1977; Redmann, 1978; Wells et al., 1979) durante o período imediatamente a seguir à sua ocorrência, e portanto apresentando conclusões exactamente opostas às do presente estudo. No entanto múltiplas razões podem estar na origem destes diferentes resultados. Na verdade nenhum destes trabalhos foi realizado nas mesmas condições que o nosso, nomeadamente no que diz respeito às profundidades estudadas, ao tipo de vegetação, ao clima e à época de queima. Um dos trabalhos confirmando um aumento no teor de água no solo após fogo é apresentado em Klock & Helvey (1976). O aumento estimado de 108.2 mm de água no solo em Setembro devido ao fogo, é semelhante aos 116 mm referidos por aqueles autores relativamente a uma floresta mista de coníferas, um ano após o fogo. No entanto é difícil tirar conclusões sólidas desta semelhança de resultados, na medida em que as condições em que os estudos decorreram foram bastante distintas, nomeadamente no que diz respeito ao delineamento experimental (sem medições antes do fogo), à época de queima (Agosto), ao clima e ao tipo de vegetação. Devemos referir que os valores de

Si obtidos no nosso estudo estão provavelmente sobrestimados na camada superficial (0- 20 cm) durante a estação seca e subestimados durante a estação húmida. Estas estimativas foram provavelmente afectadas pelo uso de um intervalo de 20 cm na integração da série de valores, devido à muito variável e heterogénea natureza desta região do solo. Por outro lado a estimativa para todo o perfil (0-180 cm) não deverá ser encarada com o impacto total do fogo na água do solo armazenada. De facto tanto a

131 extracção de água durante o período de referência por raízes abaixo do perfil estudado, como a possível percolação de água para camadas de solo mais profundas durante o período pós-fogo, podem ter contribuído para uma ainda maior diferença no armazenamento de água no solo em consequência do fogo. Os resultados deste estudo mostraram inequivocamente a existência de um muito maior teor de água no solo ao longo do perfil estudado durante a estação seca, provavelmente como consequência da redução da transpiração da vegetação devido ao fogo. Adicionalmente este efeito revelou-se consistente ao longo de todo o período estudado, incluindo a estação húmida. Deste modo, os repetidamente estudados efeitos negativos dos incêndios na taxa de infiltração não parecem ser a causa determinante no balanço final da água no solo, pelo menos para as condições estudadas neste trabalho.

O uso da modelação para simular a dinâmica da água do solo Os padrões de precipitação verificados durante o período estudado são bem o exemplo da irregularidade do clima mediterrãnico, em especial durante a estação chuvosa. A estação seca apresenta normalmente condições mais regulares, em que a escassa chuva que cai, pouco contribuí para a recarga do teor de água no solo, até à chegada das primeiras chuvas em Setembro. As espécies das comunidades arbustivas mediterrânicas evoluíram de acordo com estas condições de clima e desenvolveram adaptações para fazer face à secura estival. Uma das adaptações desenvolvidas por estas espécies consiste no desenvolvimento de sistemas radicais profundos. Esta característica é apontada como estando associada a climas com Invernos muito chuvosos (Schenk & Jackson, 2002b) e Verões secos, tal como acontece na região estudada. Os valores de humidade do solo obtidos estão de acordo com estas considerações, tendo revelado teores de água bastante elevados e relativamente constantes ao longo do ano, a 170 cm de profundidade. Este aspecto é crucial para a sobrevivência das plantas dado permitir às espécies com raízes profundas a manutenção de taxas relativamente elevadas de transpiração e de crescimento durante a estação seca. Deste modo, estas plantas têm a possibilidade de manter um fluxo sensivelmente constante de transpiração devido à existência de raízes profundas (Williams et al., 2001). A comunidade estudada é, tal como vimos anteriormente, essencialmente composta por plantas com raízes profundas dos géneros Ulex e Erica. Silva & Rego (não publ.) verificaram que estas plantas possuíam raízes com profundidades muito para além do perfil estudado (superior a 240 cm e 200 cm nas parcelas testemunha e queimada, respectivamente). Aparentemente

132 estas plantas utilizam o seu sistema radical profundo de forma a extrair água durante a estação seca, quando a água não está disponível na região mais superficial do solo. Logo que começa a estação húmida, estas plantas utilizam a elevada densidade de raízes que possuem junto à superfície para tirar então partido dos mais elevados teores de humidade e de minerais do solo. No nosso caso, a espécie Pteridium aquilinum, embora igualmente abundante, não apresentava um sistema radical profundo, (Silva & Rego, não publ.) estando a absorção de água limitada pela relativamente superficial rede de rizomas. Tal explica em parte a sazonalidade desta espécie a qual mantém frondes verdes apenas enquanto a água e a temperatura do ar se situam a níveis favoráveis, começando a secar antes do início da estação húmida. Os nossos dados confirmaram de certa forma o carácter mésico ou mediterrânico atenuado do clima. Na verdade os valores mais baixos de humidade no solo encontrados junto à superfície mantiveram-se sempre bastante afastados do ponto de emurchecimento. Estes teores de humidade verificados, deverão ser no entanto suficientemente baixos para implicar o recurso a raízes mais profundas. Aparentemente as camadas mais superficial e mais profunda do perfil explorado pelas raízes têm uma importância fundamental para as plantas mediterrânicas, dado que representam as zona preferenciais de extracção de água durante as estações húmida e seca, respectivamente (Canadell & Zedler, 1995). Este aspecto pôde ser confirmado pela distribuição de raízes encontrada na comunidade arbustiva dado que as plantas tendem a optimizar a sua distribuição radical em função da disponibilidade de água e também de nutrientes. Este facto tem levado alguns autores a estimar a distribuição das raízes a partir de modelos de água no solo (Musters & Bouten, 1999; Wijk & Bouten, 2001). As funções de pedotransferência e as equações de Mualem-Van Genuchten utilizadas em Wösten et al. (1999) para o estabelecimento da European Database of Soil Hydraulic Properties (HYPRES), revelaram uma subestimação da humidade do solo quando comparada com os valores obtidos em laboratório para valores semelhantes de pressão hidrostática equivalente. Há que referir que seria de esperar à partida alguma discrepância dado que as equações de pedotransferência utilizadas resultaram de relações empíricas destinadas a funcionar com um leque alargado de tipos de solo da Europa. Esta subestimação esteve na origem dos desvios verificados em termos da modelação da dinâmica da água no solo pelo modelo SWADY. É de supor que a aplicação do modelo a outro tipo de solo possa dar origem ou não a desvios de maior ou menor amplitude, dependendo do grau de adequação das funções empíricas utilizadas,

133 ao solo em questão. A este respeito o modelo aqui apresentado não constitui uma excepção dentro do universo de modelos de água no solo, já que quase todos apresentam alguma dependência relativamente à utilização de funções empíricas para estimar as propriedades hidráulicas do solo (Feddes et al., 2001). Em todo o caso os desvios verificados podem ser devidos a outro tipo de razões incluindo o próprio processo de medição (Musters, 1998). Os erros de medição podem ter sido originados de diversas formas, desde erros de medição, a alterações causadas na estrutura do solo (Rothe et al., 1997) até ao crescimento preferencial das raízes junto aos tubos(Maertens & Clauzel, 1982; Merril, 1992), ou ainda a erros de calibração. No que toca ao modelo, algumas das simplificações utilizadas contribuirão seguramente para uma parte dos desvios observados. Em particular há que referir a não contabilização dos processos de dinâmica radícular ao longo do ano e o facto de algumas propriedades hidráulicas do solo não serem contabilizadas tais como a condutividade hidráulica ou a histerese característica das curvas de retenção de humidade (Feddes & Koopmans, 1998). Para finalizar este aspecto, há que referir que o modelo poderia igualmente ter em conta as características particulares da camada de folhada/húmus nos processos de infiltração e evaporação. Na verdade os resultados (medidos e modelados) obtidos para a primeira profundidade estudada (15 cm) não podem ser considerados representativos dos fenómenos que ocorrem junto à superfície do solo. Para além da existência de mecanismos específicos que controlam os fluxos de água, esta zona do solo é directamente influenciada por todos os agentes meteorológicos. A existência de um leque alargado de mecanismos influenciando as entradas e as saídas de água, conduz a uma elevada variabilidade desta primeira camada em termos de teor de água, a qual não pôde ser simulada pelo modelo. Basicamente o papel da camada superficial foi o de actuar como uma camada tampão, relativamente às camadas abaixo, o que é sem dúvida uma considerável simplificação tendo em conta a sua especificidade em termos do balanço de água no solo. Este efeito tampão foi por exemplo responsável pelo atraso verificado, entre a primeira chuvada de Setembro e a chegada da frente de humedecimento a 15 cm de profundidade. No entanto, em termos gerais o modelo foi capaz de reproduzir a dinâmica da água no solo tal como foi obtida através das medições. Em particular são de referir os resultados obtidos na simulação do efeito do fogo no armazenamento de água no solo. Estes resultados basicamente confirmaram que, para situações semelhantes à estudada, o maior efeito de curto prazo de um incêndio em termos de água no solo, é a redução da transpiração pelas plantas. Deste

134 modo, tanto o modelo como os dados reais revelaram que os mais intensamente estudados efeitos do fogo na redução da infiltração, no aumento do escoamento superficial e no aumento da evaporação, não parecem ser a causa fundamental que controla a dinâmica da água no solo, no curto prazo. Muito embora possamos considerar encorajadores os resultados das simulações utilizando o modelo SWADY, estão programados alguns melhoramentos a introduzir futuramente. Estes melhoramentos deverão incluir uma abordagem mais baseada em aspectos fundamentais da física do solo. Tal permitirá uma melhor simulação dos fluxos de água, possibilitando a aplicação do modelo a um leque alargado de situações ao nível das características hidráulicas do solo. Com este objectivo, um dos desenvolvimentos futuros deverá ser a inclusão da amplamente utilizada equação de Richards (Feddes & Koopmans, 1998; Hillel, 1998). Uma outra importante alteração destinada a melhorar a performance do modelo consistirá na adopção de versões modificadas das funções empíricas de pedotransferência, de forma a permitir a obtenção de uma melhor estimativa das propriedades hidráulicas do solo.

Bibliografia

Atkinson D. (1978). The use of soil resources in high density planting systems. Acta Horticulturae 65: 75-90. Atkinson D. (2000). Root characteristics: Why and what to measure. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 1-32 pp. Bazzaz F.A. (1997). Allocation of resources in plants: state of the science and critical questions. In Bazzaz F.A. & Grace J. (eds.), Plant Resource Allocation. Academic Press, San Diego, 1-37 pp. Bell D.T. (2001). Ecological response syndromes in the flora of southwestern Western Australia: Fire resprouters versus reseeders. The Botanical Review 67: 417-440. Bengough A.G., Castrignano A., Pagès L. & Van Noordwijk M. (2000). Sampling strategies, scaling and statistics. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S. & van de Geijn S.C. (eds.), Root methods, a handbook. Springer Verlag, Berlin, 147-173 pp. Böhm W. (1979). Methods of studying root systems. Springer Verlag, Berlin. Caldwell M.M. & Virginia R.A. (1989). Root systems. In Pearcy R.W., Ehleringer J.R., Mooney H.A. & Rundel P. (eds.), Physiological plant ecology. Field methods and instrumentation. Chapman and Hall, London, 367-398 pp. Campbell R.E., Baker P.F., Ffolliot P.F., Larson F.R. & Avery C.C. (1977). Wildfire effects on a ponderosa pine ecosystem. An Arizona case study. U.S.D.A. Forest Service Research Paper RM- 191. Rocky Mountains Forest and Range Experimental Station. Fort Colins. Canadell J. & Zedler P.H. (1995). Underground structures of woody plants in mediterranean ecosystems of Australia, California and Chile. In Arroyo M.T., Zedler P.H. & Fox M.D. (eds.), Ecology and biogeography of mediterranean ecosystems in Chile, California and Australia. Springer Verlag, Berlin, 177-210 pp. Canadell J., Jackson R.B., Ehleringer J.R., Mooney H.A., Sala O.E. & Schulze E.-D. (1996). Maximum root depth of vegetation types at the global scale. Oecologia 108: 583-595.

135

Casper B.B. & Jackson R.B. (1997). Plant competition underground. Annual Review of Ecology and Systematics 28: 545-570. Correia O. & Catarino F. (1994). Seasonal changes in soil-to-leaf resistance in Cistus sp. and Pistacia lentiscus. Acta Oecologica 15: 289-300. Coutts M.P. (1983). Root architecture and tree stability. Plant and Soil 71: 171-188. Drexhage M. & Colin F. (2001). Estimating root system biomass from breast-height-diameters. Forestry 74: 491-497. Feddes R.A & Koopmans R.W. (1998). Agrohydrology. Wageningen Agricultural University, Wageningen. Feddes R.A., Hoff H., Bruen M., Dawson T.E., de Rosnay P. Dirmeyer P., Jackson R.B., Kabat P., Kleidon A., Lilly A ., & Pitman, A.J., (2001). Modeling root water uptake in hydrological and climate models. Bulletin of the American Meteorological Society 82: 2797-2809. Fitter A.H. & Ennos R.A. (1989). Architectural constrains to root system function. In Robinson D. (ed.), Roots and the soil environment. Aspects of Applied Biology 2:15-22. Fitter A.H. (1985). Functional significance of root morphology and root system architecture. In Fitter A.H., Atkinson D., Read D.J., Usher M.B. (eds.), Ecological interactions in soil. Special publication of the British Ecological Society, No. 4). Blackwell Scientific, Oxford, 87-106 pp. Fitter A.H. (1996). Characteristics and functions of root systems. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 1-20 pp. Grace J. (1997). Toward models of resource allocation by plants. In Bazzaz F.A. & Grace J. (eds.), Plant resource allocation. Academic Press, San Diego, 279-291. Guowei S., Coffin D.P. & Lauenroth W.K. (1997). Comparison of root distribution of species in North American grasslands using GIS. Journal of Vegetation Science 8: 587-596. Hellmers H. Horton J.S., Juhren G. & O’Keefe J. (1955). Root systems of some chaparral plants in Southern California. Ecology 36: 667-678. Higgins K.B., Lamb A.J. & Van Wilgen B.W. (1987). Root systems of selected plant species in mesic mountain fynbos in the Jonkershoek Valley, South-Western Cape Province. South African Journal of Botany 53: 249-257. Hillel D. (1998). Environmental soil physics. Academic Press, San-Diego, 2nd edition. Hoffmann A., & Kummerow J. (1978). Root studies in the Chilean Matorral. Oecologia 32: 57-69. Hoffmann C. & Usoltsev V.A. (2001). Modelling root biomass distribution in Pinus sylvestris forests of the Turgai depression of Kazakhstan. Forest Ecology and Management 149: 103-114. Jackson R.B., Mooney H.A. & Schulze E.-D. (1997). A global budget for fine root biomass, surface area, and nutrient contents. Proceedings of the National Academy of Science of the United States of America 94: 7363-7366. Keeley J.E. (1986). Resilience of Mediterranean shrub communities to fires. In Dell A., Hopkins J.M. & Lamont B.B. (eds.), Resilience in mediterranean-type ecosystems. Dr. Junk Publishers, Dordrecht, 95-112 pp. Klock G.O. & Helvey J.D. (1976). Soil water trends following wildfire on the Entiac Experimental Forests. Proceedings of the 15th Tall Timbers Fire Ecology Conference Tall Timber Research Station, Talahassee, U.S.A., 193-200 pp. Krämer S., Miller P.M. & Eddleman L.E. (1996). Root system morphology and development of seedling and juvenile Juniperus occidentalis. Forest Ecology and Management 86: 229-240. Kummerow J. (1981). Structure of roots and root systems. In Di Castri F., Goodal D.W. & Specht R.L. (eds.), Mediterranean-type Shrublands. Ecosystems of the World 11. Elsevier, Amesterdam, 269- 288 pp. Kummerow J., Krause D., & Jow W. (1977). Root system of chaparral shrubs. Oecologia 29: 163-177.

136

Kummerow J., Kummerow M. & Trabaud L. (1990). Root biomass, root distribution and the fine-root growth dynamics of Quercus coccifera L. in the garrigue of Southern France. Vegetatio 87: 37- 44. Leuschner C., Hertel D., Coners H. & Büttner V. (2000). Root competition between beech and oak: a hypothesis. Oecologia 126: 276-284. Maertens C. & Clauzel Y. (1982). Premières observations sur l’utilisation de l’endoscopie dans l’étude de l’enracinement in situ de plantes cultivées (Sorghum vulgare et Lolium multiflorum). Agronomie 2: 677-680. Martinez F. & Rodriguez J.M. (1988). Distribución vertical de las raices del matorral de Doñana. Lagascalia 15: 549-557. Martínez F., Merino O., Martín A., García Martín D., & Merino J. (1998). Belowground structure and production in a mediterranean sand dune shrub community. Plant and Soil 201: 209-216. Mazzoleni S. & Esposito A. (1994). Vegetation regrowth after fire and cutting of mediterranean macchia species. In Trabaud L., Prodon R. (eds.), Fire in Mediterranean Ecosystems. Commission of European Communities, Brussels, 87-99 pp. Mazzoleni S. & Pizzolongo P. (1990). Post-fire Regeneration Patterns of Mediterranean Shrubs in the Campania Region, Southern Italy. In Goldammer J.G., Jenkins M.J. (eds.), Fire in Ecosystems Dynamics. SPB Academic Publ., The Hague, 43-51 pp. McKay H.M. (1988) The influence of pine on the form of Sitka spruce fine roots. Journal of Experimental Botany 39: 1263-1266. Merril S. (1992). Pressurized-wall minirhizotron for field observation of root growth dynamics. Agronomy Journal 84: 755-758. Miller P.C. & Ng E. (1977). Root:shoot biomass relations in shrubs in Southern California and Central Chile. Madroño 24: 215-223. Musters P.A.D. & Bouten W. (1999). Assessing rooting depths of an Austrian pine stand by inverse modelling variability in maps of soil water contents. Water Resources Research 35: 3041-3048. Musters P.A.D. (1998). Temporal and spatial patterns of root water uptake in an Austrian pine stand on sandy soil. PhD thesis. University of Amsterdam. Nepstad D.C., Carvalho C.R., Davidson E.A., Jipp P.H., Lefebvre P.A., Negreiros G.H., Silva E.D. da, Stone T.A., Trumbore S.E. & Vieira S. (1994). The role of deep roots in the hydrological and carbon cycles of amazonian forests and pastures. Nature 372: 666-669. Nicotra A., Babicka N. & Westoby M. (2001). Seedling root anatomy and morphology: an examination of ecological differentiation with rainfall using phylogenetically independent contrasts. Oecologia 130: 136-145. Nobel P.S. (1996). Ecophysiology of roots of desert plants, with special emphasis on Agaves and Cacti. In Waisel Y., Eshel A. & Kafkafi U. (eds.), Plant roots. The hidden half. Marcel Dekker, New York, 823-844 pp. Pakeman R.J. & Marrs R.H. (1994). The effects of control on the biomass, carbohydrate, content and bud reserves of bracken (Pteridium aquilinum L. Kuhn), and an evaluation of a bracken growth model. Annals of Applied Biology 124: 479-493. Ranger J. & Gelhaye D. (2001). Belowground biomass and nutrient content in a 47-year-old Douglas-fir plantation. Annals of Forest Sciences 58: 423-430 Redmann R.E. (1978). Plant and soil water potentials following fire in a northern mixed grassland. Journal of Range Management 31: 443-445. Richner W., Liedgnens M., Bürgi H., Soldati A. & Stamp P. (2000). Root image analysis and interpretation. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 305-341 pp. Rothe A., Weis W., Kreutzer K., Mathies D., Hess U. & Ansorge B. (1997). Changes in soil structure caused by the installation of time domain reflectometrty probes and their influence on the measurement of soil moisture. Water Resources Research 33: 1585-1593.

137

Santantonio D., Hermann R.K. & Overton W.S. (1977). Root biomass studies in forest ecosystems. Pedobiologia 17: 1-31. Schenk H.J. & Jackson R.B. (2002a). The global biogeography of roots. Ecological Monographs 72: 311- 328. Schenk H.J. & Jackson R.B. (2002b). Rooting depths, lateral root spreads, and belowground/aboveground allometries of plants in water-limited ecosystems. Journal of Ecology 90: 480-494. Shinozaki K., Yoda K., Hozumi K. & Kira T. (1964). A quantitative analysis of plant form - the pipe model theory II. Further evidence of the theory and its application in forest ecology. Japanese Journal of Ecology 14: 133-139. Soto B. & Diaz-Fierros F. (1997) Soil water balance as affected by throughfall in gorse (Ulex euroapeus, L.) shrubland after burning. Journal of Hydrology195: 218-231. Specht R.L. & Rayson P. (1957). Dark Island heath (ninety mile plain, South Australia). Australian Journal of Botany 5: 103-114. Spurr S.H. & Barnes B.V. (1980). Forest Ecolgy. Krieger Publishing Company, Malabar. Van Noordwijk M., Brouwer F., Meijboom M., Oliveira M. do R., Bengough A.G. (2000). Trench profile techniques and core break methods. In Smit A.L., Bengough A.G., Engels C., van Noordwijk M., Pellerin S., van de Geijn S.C. (eds.), Root methods, a handbook. Springer-Verlag, Berlin, 211- 234 pp. Vande Walle I., Willems S., Lemeur R. (1998). Root length and distribution in the mineral soil of a mixed deciduous forest (experimental forest aelmoeseneie). Silva Gandeensis 63: 1-15. Wallace A., Bamberg S.A. & Cha J.W. (1974). Quantitative studies of roots of perennial plants in the Mojave desert. Ecology 55: 1160-1162. Wells C.G., Campbell R.E., DeBano L.F., Lewis C.E., Fredriksen R.L., Franklin E.C, Froelich R.C. & Dunn P.H. (1979). Effects of Fire on Soil. U.S.D.A. Forest Service General Technical Report WO-7. Wiesenmüller J., Santos W., Denich M. and Vlek P.L. (1998). Modelling of fine root distribution under secondary vegetation in NE Amazonia – a qualitative and quantitative assessment. In Proceedings of the Third SHIFT-Workshop. Manaus, Brazil, 185-189 pp. Wijk MT van, & Bouten W (2001) Towards understanding tree root profiles: simulating hydrologically optimal strategies for root distribution. Hydrology & Earth System Sciences 5: 629-644. Williams M., Law B.E., Anthoni P.M. & Unsworth, M.H. (2001). Use of a simulation model and ecosystem flux data to examine carbon-water interactions in ponderosa pine. Tree Physiology 21: 287-298. Wösten J.H., Lilly A., Nemes A., & Le Bas C. (1999). Development and use of a database of hydraulic properties of European soils. Geoderma 90: 169-185. Ziemer R.R. (1981). Roots and stability of forested slopes. In Proceedings of the Symposium on Erosion and Sediment Transport in Pacific Rim Wetlands. Christchurch, New Zealand, 343-361 pp.