diff --git a/conda-recipes/pcmdi_metrics/conda_upload.sh b/conda-recipes/pcmdi_metrics/conda_upload.sh
index 335ccdffc..ce188289b 100644
--- a/conda-recipes/pcmdi_metrics/conda_upload.sh
+++ b/conda-recipes/pcmdi_metrics/conda_upload.sh
@@ -15,9 +15,12 @@ cp conda-recipes/pcmdi_metrics/build.sh conda-recipes/pcmdi_metrics/meta.yaml.in
 cd temp/conda-recipes
 
 python ./prep_for_build.py -v $VERSION
+cp pcmdi_metrics/meta.yml.in pcmdi_metrics/meta.yml
 conda build pcmdi_metrics
+rm pcmdi_metrics/meta.yml
 anaconda -t $CONDA_UPLOAD_TOKEN upload -u $USER -l nightly $CONDA_BLD_PATH/$OS/$PKG_NAME-$VERSION-0.tar.bz2 --force
 
 python ./prep_for_build.py -v $VERSION -f nox
+cp pcmdi_metrics/meta.yml.in pcmdi_metrics/meta.yml
 conda build pcmdi_metrics
 anaconda -t $CONDA_UPLOAD_TOKEN upload -u $USER -l nightly $CONDA_BLD_PATH/$OS/$PKG_NAME-nox-$VERSION-0.tar.bz2 --force
diff --git a/conda-recipes/pcmdi_metrics/meta.yaml b/conda-recipes/pcmdi_metrics/meta.yaml
deleted file mode 100644
index b94a882cd..000000000
--- a/conda-recipes/pcmdi_metrics/meta.yaml
+++ /dev/null
@@ -1,34 +0,0 @@
-package:
-    name: pcmdi_metrics
-    version: {{ environ['VERSION'] }}
-
-source:
-    git_url: git://github.com/PCMDI/pcmdi_metrics
-
-build:
-    number: 0
-    script: python setup.py install
-    script_env:
-     - VERSION
-     - CONDA_BLD_PATH
-
-
-requirements:
-  build:
-    - python
-    - setuptools
-    - numpy
-  run:
-    - vcs >=2.6
-    - vcsaddons
-    - cdms2 >2.6
-    - genutil
-    - cdutil
-    - unidata
-    - cdp
-    - matplotlib
-    - numpy
-
-about:
-    home:  http://gitub.com/PCMDI/pcmdi_metrics
-    summary: The PCMDI metrics package is used to objectively compare results from climate models with observations using well-established statistical tests. Results are produced in the context of all model simulations contributed to CMIP5 and earlier CMIP phases. Among other purposes, this enables modeling groups to evaluate changes during the development cycle in the context of the structural error distribution of the multi-model ensemble. Currently, the comparisons are focused on large- to global-scale annual cycle performance metrics.