We examine H -band number counts determined using new photometry over two fields with a combined solid angle of 0.30 deg ^ { 2 } to H \approx 19 , as well as bright data ( H \leq 14 ) from the 2 Micron All Sky Survey ( 2MASS ) .
First , we examine the bright number counts from 2MASS extracted for the \approx 4000 deg ^ { 2 } APM survey area situated around the southern galactic pole .
We find a deficiency of \approx 25 per cent at H = 13 with respect to homogeneous predictions , in line with previous results in the B -band and K _ { s } -band .
In addition we examine the bright counts extracted for |b| > 20 ^ { \circ } ( covering \approx 27 000 deg ^ { 2 } ) ; we find a relatively constant deficit in the counts of \approx 15-20 per cent to H = 14 .
We investigate various possible causes for these results ; namely , errors in the model normalisation , unexpected luminosity evolution ( at low and high redshifts ) , errors in the photometry , incompleteness and large-scale structure .
In order to address the issue of the model normalisation , we examine the number counts determined for the new faint photometry presented in this work and also for faint data ( H \sim < 20 ) covering 0.39 deg ^ { 2 } from the Las Campanas Infra Red Survey ( LCIRS ) .
In each case a zeropoint is chosen to match that of the 2MASS photometry at bright magnitudes using several hundred matched point sources in each case .
We find a large offset between 2MASS and the LCIRS data of 0.28 \pm 0.01 magnitudes .
Applying a consistent zeropoint , the faint data , covering a combined solid angle of 0.69 deg ^ { 2 } , is in good agreement with the homogeneous prediction used previously , with a best fit normalisation a factor of 1.095 _ { -0.034 } ^ { +0.035 } higher .
We examine possible effects arising from unexpected galaxy evolution and photometric errors and find no evidence for a significant contribution from either .
However , incompleteness in the 2MASS catalogue ( < 10 per cent ) and in the faint data ( likely to be at the few per cent level ) may have a significant contribution .
Addressing the contribution from large-scale structure , we estimate the cosmic variance in the bright counts over the APM survey area and for |b| > 20 ^ { \circ } expected in a \Lambda CDM cosmology using 27 mock 2MASS catalogues constructed from the \Lambda CDM Hubble Volume simulation .
Accounting for the model normalisation uncertainty and taking an upper limit for the effect arising from incompleteness , the APM survey area bright counts are in line with a rare fluctuation in the local galaxy distribution of \approx 2.5 \sigma .
However , the |b| > 20 ^ { \circ } counts represent a 4.0 \sigma fluctuation , and imply a local hole which extends over the entire local galaxy distribution and is at odds with \Lambda CDM .
The increase in faint near infrared data from the UK Infrared Deep Sky Survey ( UKIDSS ) should help to resolve this issue .