update publications and downscaling methodology

downscaling2/Cov/cov.html

@@ -52,7 +52,7 @@
   <p class="big">Covariability between Grid Points and Variables</p>
 
   <p class="normal"> 
-  Under Construction
+  The end result of the previous steps are spatially and temporally varying PDF parameters on a grid. To create "real data" random numbers are drawn from the PDFs. Because real precipitation and temperature fields have spatial and temporal auto-correlation, the random numbers must be correlated in space and time to reproduce the proper spatial and temporal scales of observations. In addition, the large-scale wind field is used to advect the random noise for precipitation so that anisotropy in the precipitation spatial decorrelation is better captured. Also, the noise for minimum and maximum temperature need to be correlated to capture the covariability between these two variables. An iterative process is use to fit the random noise parameters.
   </p>
 
   <hr>

downscaling2/Data/bias.html

@@ -45,7 +45,6 @@
 
   <p class="indent"><a class="small" href="source.html">Data Sources</a></p>
   <p class="indent"><a class="small" href="hour.html">Hour of Observation</a></p>
-  <p class="indent"><a class="small" href="day.html">Day of Observation</a></p>
   <p class="indent"><a class="small" href="bias.html">Precipitation Biases</a></p>
 
   <p class="normal"><a href="../Interp/interp.html">Interpolating to a Grid</a></p>
@@ -64,7 +63,7 @@
   <p class="medium">Precipitation Biases</p>
 
   <p class="normal"> 
-  Under Construction
+  Many COOP stations report days with weak precipitation as zero precipitation leading to an under-reporting of "wet days" and an artificial precipitation amount histogram. These same stations also tend to round precipitation to the nearest 5 hundredths of an inch instead of the nearest hundredth of an inch (see <a href="https://journals.ametsoc.org/view/journals/bams/88/6/bams-88-6-899.xml" target="_blank">Daly et al. 2007</a>). For the CMIP5 downscaling, poor stations were identified by calculating the ratio of days with precipitation amounts from 1 to 4 hundredths of an inch to the days with 5 hundredths of an inch. The poor station are then discarded from any further calculations. For the new CMIP6 downscaling, a new method for correcting the bias that keeps the station is developed.
   </p>
 
   <hr>

downscaling2/Data/data.html

@@ -35,7 +35,6 @@
 
   <p class="indent"><a class="small" href="source.html">Data Sources</a></p>
   <p class="indent"><a class="small" href="hour.html">Hour of Observation</a></p>
-  <p class="indent"><a class="small" href="day.html">Day of Observation</a></p>
   <p class="indent"><a class="small" href="bias.html">Precipitation Biases</a></p>
 
   <p class="normal"><a href="../Interp/interp.html">Interpolating to a Grid</a></p>

downscaling2/Data/hour.html

@@ -45,7 +45,6 @@
 
   <p class="indent"><a class="small" href="source.html">Data Sources</a></p>
   <p class="indent"><a class="small" href="hour.html">Hour of Observation</a></p>
-  <p class="indent"><a class="small" href="day.html">Day of Observation</a></p>
   <p class="indent"><a class="small" href="bias.html">Precipitation Biases</a></p>
 
   <p class="normal"><a href="../Interp/interp.html">Interpolating to a Grid</a></p>
@@ -64,9 +63,14 @@
   <p class="medium">Hour of Observation</p>
 
   <p class="normal"> 
-  Under Construction
+  For many stations the hour of observation given in the Metadata is incorrect. This fact is obvious when looking at temporally lagged correlations between nearby stations. An iterative method to correct the hour of observation errors was developed based on the observed versus the expected lagged correlation. The station with worst agreement between observed and expected is corrected first. Then all lagged correlations are completed again, and the next worst station is corrected until some threshold. 
   </p>
 
+  <p class="normal">
+  In some cases the day of observation is incorrect due to the volunteer observer's misunderstanding. For example, a morning observation is sometimes incorrectly labeled as the previous day because the majority of the 24-hour period is in the preceding day. These type errors are also corrected using the above method.
+  </p>
+
+
   <hr>
   <a href="../../">Home</a>
 

downscaling2/Data/source.html

@@ -45,7 +45,6 @@
 
   <p class="indent"><a class="small" href="source.html">Data Sources</a></p>
   <p class="indent"><a class="small" href="hour.html">Hour of Observation</a></p>
-  <p class="indent"><a class="small" href="day.html">Day of Observation</a></p>
   <p class="indent"><a class="small" href="bias.html">Precipitation Biases</a></p>
 
   <p class="normal"><a href="../Interp/interp.html">Interpolating to a Grid</a></p>