Updates to experimental default distribution for tasmin and tasmax for QDM and QM.

The beta distribution can, especially in running window settings, lead to unstable fits. Therefore, we replace the experimental default setting through a normal distribution.

Author: jakobwes

ibicus/debias/_quantile_delta_mapping.py

@@ -43,8 +43,8 @@
     psl: {"distribution": scipy.stats.beta, "trend_preservation": "absolute"},
     rlds: {"distribution": scipy.stats.beta, "trend_preservation": "absolute"},
     sfcwind: {"distribution": scipy.stats.gamma, "trend_preservation": "relative"},
-    tasmin: {"distribution": scipy.stats.beta, "trend_preservation": "absolute"},
-    tasmax: {"distribution": scipy.stats.beta, "trend_preservation": "absolute"},
+    tasmin: {"distribution": scipy.stats.norm, "trend_preservation": "absolute"},
+    tasmax: {"distribution": scipy.stats.norm, "trend_preservation": "absolute"},
 }
 
 

ibicus/debias/_quantile_mapping.py

@@ -68,12 +68,12 @@
         "mapping_type": "parametric",
     },
     tasmin: {
-        "distribution": scipy.stats.beta,
+        "distribution": scipy.stats.norm,
         "detrending": "additive",
         "mapping_type": "parametric",
     },
     tasmax: {
-        "distribution": scipy.stats.beta,
+        "distribution": scipy.stats.norm,
         "detrending": "additive",
         "mapping_type": "parametric",
     },