One of the first things we came across was the inability of many humidity sensors to reliably report Relative Humidity (RH) above 75%. In particular the DHT21 style sensors saturated at about that point and were rendered useless until the RH dropped to the 50% range for an hour or more. That is completely unacceptable for a weather station that must operate in a coastal, tropical, or semi-coastal area.
Thus a search was mounted to find a reasonably priced sensor without that problem. The Sensirion line were good candidates but their relatively high cost was problematic. However recently they have begun making a more value sensitive line and the SHT21 was selected for testing.
At about the same time, TI began selling a good value sensor that looked appropriate, the HDC1080.
We began using the ESP8266 platform that is planned for our Bee Hive Monitor and modified the software to allow a compile time switch to select between either of these sensors. Since then they have been tested outside in our particularly challenging Texas Gulf Coast weather. In general the RH hits between 95-100% each night and the following day rarely sees it below 60%. At the same time the temperature ranges from about 77F (25C) at night to 99F (37C) in the afternoon. This puts the heat index at around 113F (45C).
You can see from that range that a good RH number is vital to calculating the heat index to determine the stress on bees, livestock, and humans.
Our initial testing has discovered no major differences in "out of the box" performance of the SHT21 and HDC1080. A simple function routine was written to have both types present their data to the main loop in the same format, so there is little difference from a software perspective. Both have readily available ESP libraries and datasheets.
We will be moving ahead with our Bee Hive monitors and use some of each type in the various initial prototype installation. That should allow a bit better "A to B" comparisons.