A Percentage of Possible Sunshine Forecasting Experiment at Albany, New York

Alan M. Cope Department of Atmospheric Science, State University of New York at Albany, Albany 12222

Search for other papers by Alan M. Cope in
Current site
Google Scholar
PubMed
Close
and
Lance F. Bosart Department of Atmospheric Science, State University of New York at Albany, Albany 12222

Search for other papers by Lance F. Bosart in
Current site
Google Scholar
PubMed
Close
Restricted access

Abstract

The results of two regression experiments to predict percentage of possible sunshine (PoPS) one day in advance at Albany, New York are described. For the one experiment, predictors are derived from Albany radiosonde observations, while the other experiment uses predictors obtained from the daily 1200 GMT run of the operational Limited Area Fine Mesh (LFM) model at the National Meteorological Center (NMC). The most frequently chosen predictors from the radiosonde observations include layer-mean moisture and temperature information, 850 and 700 mb meridional wind components and stability indices. Similar predictors emerge from the LFM forecasts with the 700 mb vertical velocity replacing the meridional wind components. Ten years of twice-daily Albany radiosonde data and four years of LFM prognoses for Albany constituted the dependent data sample for both experiments.

Skill levels are assessed relative to a persistence climatology through the ranked probability score for forecasts of POPS as a continuous variable. PoPS forecasts based upon sounding predictors exhibit skill levels of 14,35 and 45% for predictors obtained from 1)the previous evening (0000 GMT), 2) the previous evening (0000 GMT) and this morning (1200 GMT) and 3) this morning (1200 GMT) and this evening (0000 GMT), respectively. The 24–36 h PoPS forecasts from 1200 GMT LFM-derived predictors show a 25% skill level. Categorical forecasts in both experiments underforecast the extremes.

Inspection of particularly erroneous individual PoPS forecasts suggests that errors arise from regression equation inadequacies and to a lesser extent from poor numerical model forecasts. Overforecasting cases are usually associated with a persistent, thin layer of low clouds in an otherwise dry atmosphere, while most underforecasting cases result from the inability of the winter season equations to recognize warm. dry situations in late autumn and early winter.

Abstract

The results of two regression experiments to predict percentage of possible sunshine (PoPS) one day in advance at Albany, New York are described. For the one experiment, predictors are derived from Albany radiosonde observations, while the other experiment uses predictors obtained from the daily 1200 GMT run of the operational Limited Area Fine Mesh (LFM) model at the National Meteorological Center (NMC). The most frequently chosen predictors from the radiosonde observations include layer-mean moisture and temperature information, 850 and 700 mb meridional wind components and stability indices. Similar predictors emerge from the LFM forecasts with the 700 mb vertical velocity replacing the meridional wind components. Ten years of twice-daily Albany radiosonde data and four years of LFM prognoses for Albany constituted the dependent data sample for both experiments.

Skill levels are assessed relative to a persistence climatology through the ranked probability score for forecasts of POPS as a continuous variable. PoPS forecasts based upon sounding predictors exhibit skill levels of 14,35 and 45% for predictors obtained from 1)the previous evening (0000 GMT), 2) the previous evening (0000 GMT) and this morning (1200 GMT) and 3) this morning (1200 GMT) and this evening (0000 GMT), respectively. The 24–36 h PoPS forecasts from 1200 GMT LFM-derived predictors show a 25% skill level. Categorical forecasts in both experiments underforecast the extremes.

Inspection of particularly erroneous individual PoPS forecasts suggests that errors arise from regression equation inadequacies and to a lesser extent from poor numerical model forecasts. Overforecasting cases are usually associated with a persistent, thin layer of low clouds in an otherwise dry atmosphere, while most underforecasting cases result from the inability of the winter season equations to recognize warm. dry situations in late autumn and early winter.

Save