Results Directory UMM :Data Elmu:jurnal:A:Agricultural & Forest Meterology:Vol101Issue4April2000:

H. Sinoquet et al. Agricultural and Forest Meteorology 101 2000 251–263 257 3.3. Model application to the multispecies canopies Light partitioning between species in all mixed canopies described earlier was computed using SIR- ASCA Sinoquet et al., 1990, ERIN Wallace, 1997 and the KMS model proposed in this work. SIRASCA was assumed to provide the reference computations, because it is a 1D version of a 3D model which has been tested against experimental data in a large number of contrasting canopies, e.g. grass–legume mixtures Sinoquet et al., 1990; Fau- rie et al., 1996, row canopies Sinoquet and Bon- homme, 1992; Andrieu and Sinoquet, 1993; Mabrouk et al., 1997, alley-cropping Tournebize and Sinoquet, 1995. SIRASCA was run using LAI and mean leaf inclination of each plant species in each canopy layer, as well as leaf scattering coefficient of each species. Incident radiation was assumed to be an overcast sky obeying the SOC luminance distribution Moon and Spencer, 1942. ERIN was run with some modifications to the orig- inal version proposed by Wallace 1997. The original extinction coefficients were replaced by those given by Eq. 6 i.e. as a function of mean leaf inclina- tion, and they were weighed by √ 1 − σ in order to account for leaf scattering as proposed by Goudri- aan, 1977. Such changes were aimed at harmonising the models for the calculation of the extinction coeffi- cients, in order to compare the models only with regard to their ability in computing light partitioning. In the case of the monolayer canopy, the two-species were assumed to have the same height, while plant height was entered as described earlier for the multilayer canopies. ERIN was not applied to the four-component canopy. With regard to the KM model proposed in this work, only the simplified equations as described in Section 2.3 were used, even in the case of the monolayer canopy.

4. Results

4.1. Two-species monolayer canopies Under low scattering conditions, linear regres- sion analysis shows that the fraction of absorbed radiation computed by the KMS model is very close to the reference model outputs, i.e. SIRASCA simulations ε K 1 = 1.006ε S 1 + 0.009 r 2 = 1.000 ε K 2 = 1.010ε S 2 − 0.023 r 2 = 0.999 where subscript 1 and 2 refers to the planophile and erectophile species, and superscript K and S refers to KMS and SIRASCA models, respectively. The max- imum deviation between absorption value computed by the two models is 0.02 and 0.03 for the planophile Species 1 and erectophile Species 2, respectively. Mean residuals are respectively +0.012 and −0.018, indicating that light sharing between canopy compo- nents slightly favours the planophile species when computed by the KMS model. Deviations between SIRASCA and ERIN outputs are much more important, as shown in Fig. 1 and the linear regression equations ε E 1 = 0.868ε S 1 + 0.081 r 2 = 0.971 ε E 2 = 0.864ε S 2 − 0.023 r 2 = 0.953 where superscripts E and S refer to the ERIN and SIRASCA models, respectively. Maximum difference between light absorption computed by the models reaches 0.10 and 0.12 for Species 1 and 2, respec- tively. Fig. 1 clearly shows that outputs computed by the ERIN model are biased: residuals are not ran- domly distributed, and the scatter diagram looks like a family of curves corresponding to the combinations of the two species. Fig. 2 shows simulated fraction of absorbed radia- tion as a function of LAI for each species, and for dif- ferent levels of the presence of the companion species. The deviations between SIRASCA and ERIN outputs are the lowest when the competing species exhibits a small LAI LAI=1 in Fig. 2. When competition in- creases, ERIN underestimates the difference in light absorption by the two species, i.e. the strength of light competition. Under high scattering conditions, linear regression analysis shows that the agreement between SIRASCA and KMS outputs is good only in the case of the planophile Species 1 ε K 1 = 0.995ε S 1 + 0.009 r 2 = 0.996 ε K 2 = 0.847ε S 2 − 0.012 r 2 = 0.999 258 H. Sinoquet et al. Agricultural and Forest Meteorology 101 2000 251–263 Fig. 3. Comparison between SIRASCA, KMS and ERIN models in the case of Trifolium repens–Lolium perenne canopies described by Faurie et al. 1996: fraction of PAR absorbed ε by Trifolium repens 1 a and Lolium perenne 2 b. The mixture was simu- lated as either a multilayer SIRASCA, ε S,n , reference run; KMS: ε K,n , d or a monolayer canopy SIRASCA: ε S,1 , h ; KMS: ε K,1 , s ; ERIN: ε E,1 , 1. See text for input parameters. where superscript K refers to the KMS model. KMS significantly underestimates by 15 light absorption by the erectophile Species 2: maximum deviation reaches –0.08 when light absorption is above 0.4. In contrast, ERIN largely overestimates light absorption Fig. 4. Comparison between SIRASCA, KMS and ERIN models in the case of Vicia sativa–Avena sativa canopies described by Ouknider and Jacquard 1989: fraction of PAR absorbed ε by Vicia sativa 1 a and Avena sativa 2 b. The mixture was simulated as either a multilayer SIRASCA, ε S,n , reference run; KMS: ε K,n , d or a two-layer canopy SIRASCA: ε S,2 , h ; KMS: ε K,2 , s ; ERIN: ε E,2 , 1. See text for input parameters. by the two species, i.e. by 50 and 28, respectively, as can be shown from linear regression lines ε E 1 = 1.505ε S 1 + 0.033 r 2 = 0.994 ε E 2 = 1.286ε S 2 + 0.008 r 2 = 0.992 H. Sinoquet et al. Agricultural and Forest Meteorology 101 2000 251–263 259 The maximum deviation is 0.30 and 0.18 for Species 1 and 2, respectively. 4.2. Multispecies multilayer canopies Fig. 3 shows the relationships between the fraction of PAR absorption computed by the three models in the T. repens–L. perenne mixture, i.e. where the height of both species was the same. The reference run was the SIRASCA simulation where the mixture was described as a multilayer canopy according to the measurements by Faurie et al., 1996. The KMS model with the same canopy description gave results very close to the reference run: the maximum devia- tion was −0.04 and 0.01 for T. repens and L. perenne, respectively. When the canopy was described as a monolayer canopy, both SIRASCA and KMS mod- els significantly deviated from the reference model: the maximum deviations were −0.21 and −0.23, respectively. However the observed discrepancies were very similar for the two models, as suggested by the good agreement between SIRASCA and KMS outputs when applied to a monolayer canopy Fig. 1. The monolayer models tend to smooth light compe- tition between the two species by overestimating the lowest light absorption efficiencies and underestimat- Fig. 5. Simulated fraction of PAR absorbed by rice and two weed groups described by Graf et al. 1990, as a function of growth stage in degree-days. Comparison between KMS filled symbols and Sirasca open symbols simulation. Rice r , tall Cyperaceae, j , and other Poaceae d . See text for input parameters. ing the highest ones. This is a result of disregarding the vertical stratification of foliage within the canopy thickness. Deviations of ERIN outputs from the ref- erence run were highest and reached 0.24 and −0.29 for T. repens and L. perenne, respectively. The ERIN model markedly smoothed light competition between the two components. This resulted from neglecting the inter-species differences in LAI, leaf inclination and vertical distribution of leaf area. The same general trends were found by applying the three models to the V. sativa–A. sativa mixture Fig. 4. First PAR absorption as calculated by KMS using a multilayer description i.e. according to the measurements by Ouknider and Jacquard, 1989 was very close to that given by the reference run. Second both SIRASCA and KMS models underestimated the strength of light competition when the mixture was represented as a two-layers canopy. Third ERIN simu- lations gave the greatest deviations from the reference run, with large underestimation of light competition between species. Application of the KMS model to the rice–weeds mixture confirmed that the KMS model was able to correctly summarise the more detailed model, even in a case of complex multilayer four and multispecies four canopies Fig. 5, where results for small dicots 260 H. Sinoquet et al. Agricultural and Forest Meteorology 101 2000 251–263 are not shown since both measured and simulated val- ues ranged between 0 and 0.02.

5. Discussion