Gao el al. • MARSH GRASSBIRD AND REEDBED MANAGEMENT 
757 
TABLE 1. Differences (Kruskal-Wallis tests) in height and number of reeds in three types of reedbeds, Yalu River 
Estuary. China. Type 1 = lowland reed marsh. Type 2, = highland reed marsh, and Type 2 2 = highland reed-grass marsh. 
Reeds 
X 2 
p 
Type 
Mean rank 
Number 
105.183 
<0.001 
1 
40 
78.36 
2 , 
43 
109.38 
2-, 
53 
27.89 
Height 
97.197 
<0.001 
1 
47 
114.19 
2 , 
46 
80.34 
2 , 
53 
31.48 
reeds or grasses nearest to the point were selected 
if there was no reed or grass at each point. Green 
and dry reeds (some with inflorescence) height 
was measured at the point where the reed started 
to droop above the ground. 
Measurement of green grass height was focused 
on the dominant grass species. The tallest leaf 
(droop) height for green Co lama a rout is epigejos, 
■Iuncus grttcillimus, and Scirpus planiculmis 
without inflorescence was measured at the portion 
where the bulk of the mass occurred in the leafy 
portion of the grass. The inflorescence represents 
the hulk of the mass for dry Calamagwstis and 
the tallest portion of the inflorescence was 
measured. However, dry reeds were often broken 
by wind, and J. grctcillimus and S. planiculmis 
often decayed. Thus, the lowest height values 
were deleted, and the mean dominant height was 
used. Counting of reeds occurred in the field, 
whereas counting of grasses occurred after 
clipping the nest quadrants. Measurement of the 
height of green reeds and grasses was in late July. 
We used random sampling methods to locate 
sampling circles (radius = 1.5 m) to count reeds 
in Type 1 (lowland, n = 55) and Type 2\ 
highland, /? = 46) reed marsh. We also measured 
reed height in the two habitat types with regular 
sampling methods. 
We used GPS to locate dense grass areas and a 
table of random numbers for delining the 
sampling point at the intersection of the grids. 
Cover, height, and number of total grasses 
between nesting and dense-grass areas were 
included as there were no nests in dense grass 
areas. 
Statistical Analyses.— Three-independent sam¬ 
ples tests were used to compare height and density 
°l reeds among the three types of reedbeds. Two- 
independent samples tests were used to compare 
•1 habitat variables between before deep irriga¬ 
tion (>30 cm depth) (2006 and 2007) and after 
(2009). and to compare cover, height, and number 
of total grasses between nest and dense-grass 
areas. 
Multiple linear regression analysis was used to 
evaluate which habitat characteristics influenced 
nest height. Program MARK was used to estimate 
the daily survival rate and overall nest success. 
Akaike Information Criterion (AIC) values were 
calculated in EViews 5 (Quantitative Micro 
Software 2004) to identify which models best 
explained the relationship. Nest-entrance orienta¬ 
tion data were examined in Oriana 3.21 (Kovach 
Computing Services 2010), and other data were 
examined with SPSS 10.0 (SPSS 1999). 
RESULTS 
Nesting. —Three types of reedbed from the 
Dayang River to uplands were classified. Type 
1, without Marsh Grassbirds, was lowdand reed 
marsh with the tallest reeds and highest reed 
productivity. Type 2 was highland irrigated reed 
marsh with shorter reeds and lower productivity 
separated from the Yellow Sea and the Dayang 
River by dikes, among which, two subtypes of 
reed marsh were classified. Type 2j had the 
densest reed, medium reed height, and productiv¬ 
ity. Type 2 2 had the fewest reeds, the lowest reed 
height, and productivity (Table 1) as it contained 
> eight grass species Marsh Grassbirds selected 
as breeding habitats. 
Forty-six of the 53 Marsh Grassbird nests (5 in 
2006, 35 in 2007, 13 in 2009) over 4 years had 
sufficient information for use in nest survival 
analyses. No nests were found in 2008. Mean 
(SD) clutch size was 4.5 ± 0.83 (range = 3-6 
eggs). The earliest date of laying of the first egg 
was 22 May and daily survival rate (± SE) was 
0.923 ± 0.0125, suggesting overall nest success 
of 12.5% (95% Cl = 0.055^0.223) across the 26- 
day nesting cycle. 
Twenty-one (39.6%) nests had nest cover with 
