Precipitation forecast over western Himalayas using k‐nearest neighbour method

• Published in: International journal of climatology 2008-11, Vol.28 (14), p.1921-1931
• Main Authors: Dimri, A. P., Joshi, P., Ganju, A.
• Format: Article
• Language: English
• Subjects: k‐nearest neighbour; precipitation; probability
• ISSN: 0899-8418; 1097-0088
• DOI: 10.1002/joc.1687

Knowledge of precipitation over the western Himalayas is of utmost importance during winter season (December, January, February, and March—DJFM) due to extreme weather conditions. During winter, eastward moving synoptic weather systems called western disturbances (WDs) yield enormous amount of precipitation over this region. This amount of precipitation by a number of WDs keeps on accumulating and poses an avalanche threat to the habitat. Not only this, low temperature conditions along with precipitation form extremely hostile winter conditions to live with. In the present study, the nearest neighbour (NN) method is used to forecast probability of precipitation (PoP) occurrence/non‐occurrence and its quantity. The method of NN is introduced to visualize the results. The NN forecast technique attempts to compare similar situations in the past with current data and assumes that similar events are likely to occur under similar conditions. At present, only nine important weather variables are considered for generating a 3‐day advance forecast of PoP occurrence/non‐occurrence and quantity at eight representative observatories in Jammu and Kashmir (J&K), the northmost state of India. This state receives maximum amount of solid precipitation in the form of snow during winter. Past data of 8–14 years (between 1988 and 2003) is considered, from which the nearest days are looked for and tested with data of 2–4 years (between 2003 and 2005). PoP occurrence/non‐occurrence is well predicted by k‐Nearest Neighbour (k‐NN) sampling method. It is evident from the results that the model is able to give the projections 3 days in advance, with an accuracy of 71–88%. Probability forecast of occurrence of precipitation is predicted well by the present model setup, but the quantitative precipitation forecast (QPF) is an important issue that still needs improvement, keeping in view the topographical and land‐use heterogeneity of the area under study. Copyright © 2008 Royal Meteorological Society