Deaer Srijith Balan, the concept of variogram analysis applies the concept of standard deviation to spatial distance and direction. The higher spatial autocorrelation is in a certain direction the lower the variogram values are in this direction. This concept has been developed and used at first for ore reserve estimation by an south African mine engineer called Krige in the fourties of the last century. The French Mathematician Matheron (Ecole des mines de Paris) created a theory of regionalized variables. This variogram analysis can be used for a spatial prediction method, called Kriging. In the statistical sense Kriging is spatial BLUP - the best linear unbiased prediction. Computationally Kriging has similarities to the Kalman filter. For your practical application this means, that Kriging always works relatively well, when the measured data have relatively much autocorrelation, which also means that the data should have enough measurement points. This can work well in cases of precipitation or for example in the sediment, when diffusion brings low spatial variation. It also can be well used in the case of interpolation of radiation. On the other hand in the case of soil data not seldomly suddenly high spatial variations of measured variables are observed. Plain Kriging then, because of its similarity to the kalman filter assumes that sudden variation is an erroneous value and has the tendency to wipe out this local variation. For many cases of real world problems universal kriging is recommendable.
Program packages for variogram analysis and Kriging are for example: GEOPACK, http://www2.epa.gov/water-research/geostatistical-software-package-geopack or R.
In the literature, there are different sources, such as: Handbook of Applied Spatial Analysis: Software Tools, Methods and Applications
herausgegeben von Manfred M. Fischer, et al., or Variowin: Software for Spatial Data Analysis in 2D,
Kriging analysis of mean annual precipitation, Powder River Basin, Montana and Wyoming, US Geological Survey, Scrivan and Karlinger
Semi-variogram Estimation and Universal Kriging Program, Scrivan and Karlinger
Deaer Srijith Balan, the concept of variogram analysis applies the concept of standard deviation to spatial distance and direction. The higher spatial autocorrelation is in a certain direction the lower the variogram values are in this direction. This concept has been developed and used at first for ore reserve estimation by an south African mine engineer called Krige in the fourties of the last century. The French Mathematician Matheron (Ecole des mines de Paris) created a theory of regionalized variables. This variogram analysis can be used for a spatial prediction method, called Kriging. In the statistical sense Kriging is spatial BLUP - the best linear unbiased prediction. Computationally Kriging has similarities to the Kalman filter. For your practical application this means, that Kriging always works relatively well, when the measured data have relatively much autocorrelation, which also means that the data should have enough measurement points. This can work well in cases of precipitation or for example in the sediment, when diffusion brings low spatial variation. It also can be well used in the case of interpolation of radiation. On the other hand in the case of soil data not seldomly suddenly high spatial variations of measured variables are observed. Plain Kriging then, because of its similarity to the kalman filter assumes that sudden variation is an erroneous value and has the tendency to wipe out this local variation. For many cases of real world problems universal kriging is recommendable.
Program packages for variogram analysis and Kriging are for example: GEOPACK, http://www2.epa.gov/water-research/geostatistical-software-package-geopack or R.
In the literature, there are different sources, such as: Handbook of Applied Spatial Analysis: Software Tools, Methods and Applications
herausgegeben von Manfred M. Fischer, et al., or Variowin: Software for Spatial Data Analysis in 2D,
Kriging analysis of mean annual precipitation, Powder River Basin, Montana and Wyoming, US Geological Survey, Scrivan and Karlinger
Semi-variogram Estimation and Universal Kriging Program, Scrivan and Karlinger
Srijith Balan. In addition to the advice from Ludwig, you need to decide whether you are interested in the correlation length of precipitation, or the correlation length of precipitation anomalies from a long-term climatology. There is usually a larger correlation length for precipitation anomalies. In either analysis, it is necessary to allow for the problem that precipitation is not normally distributed (daily values are far from normally distributed in many climates). One method is to express raw precipitation values as a percentage of the monthly, seasonal or even annual average. This method needs careful testing for approximate normality. Another common raw precipitation data transformation is to take cube roots of the raw values, analyse them for spatial coherence, and then transform back. Sometimes the correlation length will vary significantly with direction. I could imagine that being true of the Western Ghat mountains for instance. So this also needs testing.