#### breeding parameters

Relative sample sizes within habitats varied among sites because of different population sizes and habitat use at each. Of the breeding attempts sampled, 9% and 54% were from Ouse washland and arable, respectively, and it should therefore be noted that the power to detect inter-site differences was low.

The first date of incubation was determined for 115 clutches (23 on arable). The average date on which incubation started on washland was 26 April (SD 9 days) and the average stop date was 20 May (SD 9). On arable, the start and stop dates were 1 May (SD 5) and 15 May (SD 4), respectively. The average duration of the laying season was therefore 25 days on washland and 15 on arable.

The fates of 156 clutches were recorded (22 on arable). Daily survival rates varied significantly among wet grassland and arable habitats ( = 15·1, *P* < 0·001). There were no effects of year ( = 2·9, *P* > 0·5), site ( = 0·76, *P* > 0·3) or calendar date ( = 0·5, *P* > 0·4) once the habitat-specific deviance was explained. The daily nest survival rates were 98·5% (Lower 95% Confidence Interval (LCI) 97·9, Upper 95% Confidence Interval (UCI) 98·9) on washland and 94·4% (LCI 90·0, UCI 96·9) on arable. Over the 26-day average laying and incubation period, the hatching success on washland was 67·4% and that on arable was 22·1%. Of the daily losses on washland, predation accounted for 96·2% and desertion for 3·8%, while on arable these figures were 66·6% and 33·3%, respectively. Predation accounts for 56% and desertion for 44% of the difference in overall nest survival among habitats.

The survival of 116 chicks in 90 broods and 11 chicks in seven broods was recorded on washland and arable, respectively. Daily chick survival rate was 94·7% (LCI 93·6, UCI 95·6) on washland and 77·6 (LCI 66·7, UCI 82·3) on arable, with the lack of overlap between these estimates and their bootstrapped confidence intervals showing the differences to be significant. There were no effects of year, site, calendar date or chick age on survival rates within habitats. Over the 27-day average fledging period, the fledging success on wet grassland was 22·9% and that on arable was 0·1%. None of the chicks that hatched on arable subsequently moved to washland. Of the daily losses on washland, predation accounted for 94·6% and starvation for 5·4%, while on arable these were 81·9% and 18·1%, respectively. Predation accounted for 79% and starvation for 21% of the overall difference in chick survival among habitats.

#### productivity

The productivity of black-tailed godwits nesting solely on washland in years when flooding did not disrupt nesting was 0·65 (SE 0·23) chicks pair^{−1}. The productivity of pairs nesting exclusively on arable was 0·01 chicks pair^{−1} (SE 0·002). The simulated annual productivity of black-tailed godwits breeding at the Ouse and Nene Washes between 1975 and 2003 are presented in Fig. 1. Floods caused productivity at the Ouse to be significantly lower than the level of 0·395 chicks pair^{−1} required to maintain a stable population in 59% of years (*n* = 29) and complete failure occurred in 76% of these. At the Nene, productivity was suppressed to approximately this level in only 10% of years (*n* = 20).

The productivity indices at the Ouse in relation to flood mitigation options are presented in Table 1. Productivity under current management was 42% of that in the absence of flooding. Changing a single component of the Ouse Washes or the catchment produced only marginal improvements in the productivity index of between 4% and 10%. Options involving pairing of changes to the Ouse Washes or the catchment caused the productivity index to improve to between 57% and 63%. Combining all available flood mitigation options resulted in the productivity index rising to 89%, which was more than double that resulting from current management.

Table 1. The effect of flood mitigation options (see Appendix S1 for details) on the productivity and population multiplication rate (λ) of black-tailed godwits at the Ouse Washes. Productivity is expressed as a percentage of that which would have been attained if no floods had occurred during the 20 years of study Option | Productivity | Open population | Closed population |
---|

Without option 12 | With option 12 | λ | SD | λ | SD |
---|

Option 0 | 42·1 | 93·6 | 1·033 | 0·023 | 0·902 | 0·048 |

Option 1 | 49·5 | 93·6 | 1·046 | 0·019 | 0·923 | 0·044 |

Option 2a | 49·1 | 95·5 | 1·040 | 0·021 | 0·916 | 0·044 |

Option 2b | 49·4 | 96·7 | 1·040 | 0·022 | 0·917 | 0·043 |

Option 3 | 48·7 | 96·1 | 1·040 | 0·021 | 0·921 | 0·044 |

Option 4 | 48·5 | 96·5 | 1·039 | 0·022 | 0·923 | 0·043 |

Option 5 | 48·9 | 96·5 | 1·039 | 0·021 | 0·921 | 0·042 |

Option 6 | 46·5 | 95·4 | 1·038 | 0·021 | 0·912 | 0·045 |

Option 7 | 51·9 | 96·8 | 1·042 | 0·021 | 0·924 | 0·043 |

Option 8 | 60·1 | 96·2 | 1·059 | 0·017 | 0·938 | 0·039 |

Option 9 | 57·4 | 96·5 | 1·049 | 0·019 | 0·939 | 0·038 |

Option 10 | 63·0 | 96·2 | 1·057 | 0·019 | 0·947 | 0·039 |

Option 11 | 88·9 | 96·3 | 1·087 | 0·013 | 1·000 | 0·029 |

Option 12 | – | – | 1·092 | 0·012 | 1·010 | 0·029 |

Creating wet meadows outside the Ouse Washes resulted in the productivity index rising further to between 93% and 97% (Table 1). Combining habitat creation with the various flood mitigation options produced only marginal increases in the productivity index and so further modelling of habitat creation was conducted in isolation.

#### population trends

The population trends at the Ouse and Nene Washes differed markedly (Fig. 2). Numbers at the Ouse declined from 55 pairs in 1975 to five pairs in 1992, with fluctuations associated with the occurrence of spring flooding. The population recovered to a peak of 19 pairs in 1997 during a flood-free period, but declined to four pairs in 2004 as floods recurred. The Nene population increased at an average rate of 11% pa, with a more rapid increase of 32% between 1999 and 2000 and between 2003 and 2004.

The predictions of the population models incorporating constant adult and immature survival and flood-dependent variation in annual productivity and breeding likelihood broadly described the contrasting population trends at the Ouse and Nene (Fig. 2). This provided strong evidence that flood-induced breeding failure was responsible for the decline in godwit numbers at the Ouse Washes. It also suggested that the parameters used in the model were reasonably accurate and provided a quantitative link between flood patterns and population growth. The use of the models to predict future population trends under various management options is therefore justifiable.

Population growth rates at the Nene Washes were 5·1% pa (SD 0·7) and 4·1% pa (SD 0·8) for the closed and open models, respectively. These rates of growth resulted in the population in 30 years time being 195 (SD 40) and 145 (SD 31) pairs, respectively. The slower growth at the Nene Washes in the open model was because of net emigration of birds to the smaller and less productive Ouse Washes population.

Population multiplication rates of godwits at the Ouse under various management options are presented in Table 1. The closed population model resulted in the population declining at a rate of −10·3% pa under current conditions. Single flood mitigation options resulted in numbers declining at slower rates of around −8% pa, and paired options resulted in growth rates of −6% pa. Option 11 resulted in a population that was on average stable, while option 12 produced a population that increased at 1% pa. For the open population model, numbers increased at a rate of 3·2% pa under current conditions. Single *in situ* flood mitigation options produced population growth rates in the order of 4% pa. Paired mitigation options produced population growth in the order of 5% p.a. Options 11 and 12 produced population growth rates of 8·3% and 8·8% pa, respectively. The contrasting predictions of the closed and open models suggested that the Ouse population was a sink, dependent on immigration from the Nene Washes source in order to persist, under all management options other than 11 and 12.

#### likelihood of population persistence and meeting objectives

The likelihood of a closed population persisting over varying time periods in relation to the flood mitigation options employed are presented in Fig. 3a. The persistence likelihood was 91% for all options up to 5 years but decreased rapidly under most options thereafter. Only 8% of simulated populations persisted for 30 years under current conditions, 15% under single flood mitigation options and 30% under paired options. The rate of decline in population persistence with time under options 11 and 12 was slower, at 52% and 60% after 30 years, respectively.

Allowing immigration from the Nene improved persistence likelihoods considerably for all options examined (Fig. 3b). Persistence likelihoods over 30 years were in excess of 75% under current conditions, 80% for single flood mitigation options, 90% for paired options and approached 100% for options 11 and 12.

The likelihood of meeting the target of 46 pairs over any time period up to 30 years was less than 0·03 when employing options 0–10 for either the open or closed population models. The closed population model showed that the likelihood of the target being met was zero under all management options. For open models under options 11 and 12 the likelihood of meeting the target was zero up to 20 years but this increased rapidly thereafter to 0·56 and 0·71 (respectively) by 30 years.