published in: Meteorologicke Zpravy, vol 49 (1996), p.97-105
language: Czech + English abstract; includes 5 figures and 1 table with English captions.

Met&Roll: the stochastic generator of daily weather series for the crop growth model.

Martin Dubrovsky
Institute of Atmospheric Physics, Hradec Kralove, Czech Republic

visit my home page      

ABSTRACT:
Use of crop growth models (fig. 1) for estimating potential impacts of climate change on agriculture is discussed in the second chapter. The impacts are assessed based on comparison of model simulations with present climate and changed climate daily weather series (figs. 2-3). The weather series for the changed climate conditions may be obtained either by direct modification of observed weather series or by weather generator which generates synthetic series stochastically similar with the observed series. The review of the weather generators - approaches and applications - is given in chapter 3. The single-site four-variate weather generator `Met&Roll', which was developed for the purpose of climate change impact studies in the Czech Republic, is described in chapter 4. The model variables are: daily sum of global solar radiation (SRAD), daily maximum and minimum temperatures (TMAX and TMIN) and daily precipitation amount (RAIN). Precipitation occurrence is modelled by first-order Markov chain model, precipitation amount by Gamma distribution. The standardized deviations of SRAD, TMAX and TMIN from their means are modelled by first-order autoregressive model with means and standard deviations being conditioned on precipitation occurrence. Annual courses of parameters of the weather generator are smoothed by robust locally weighted regression (fig. 5). The main services provided by Met&Roll include (fig. 4): (1) estimating parameters of the weather generator from the observed weather series, (2) generating synthetic weather series, (3) modifying parameters of the weather generator in accordance with climate change scenario, (4) direct modification of observed daily weather series in accordance with climate change scenario. The results of the verification tests are subject of the forthcoming paper.

Figures and table:

Fig.1 Output from the CERES-Maize model simulating growth of the maize. Legend: (1) growth stage [x 0.001] (step-wise solid line); (2) grain weight [kg/ha] (steep dotted), (3) biomass [kg/ha] (slow dashed), (4) grain number [m^-2] (step-wise dashed - single step), (5) potential evapotranspiration [x 0.001 mm/d] (rugged solid), (6) cumulative precipitation [x 0.1 mm] (slow dotted).

/MW=MD; CGM=RSM/)
Fig.2 The conceptual model of the research into the climate change impacts on crop production (taken from [47] and revised). A: analysis of daily weather series; MP: modification of climatic characteristics (parameters of the weather generator) in accordance with climate change scenario; G: generation of synthetic weather series; RSM: crop growth model. The procedures inside the dashed area are supported by Met&Roll (see fig.4).

Fig. 3. Augmentation of the biomass [kg/ha] simulated by crop model CERES-Maize. The curves relate to various versions of the input daily weather series (from below in Sep-3): (1) RAINx1.1; (2) baseline weather data; (3) SRAD+1; (4) TMIN+1, (5) TMAX+1, (6) (SRAD+1) AND (TMAX+1) AND (TMIN+1) AND (RAINx1.1).

Fig. 4. The main procedures of Met&Roll. A: analysis of daily weather series (1 - calculation of raw annual cycles, 2 - smoothing annual cycles, 3 - analysis of standardized deviations from the smoothed means), B: modification of the daily weather series, C: modification of the generator's parameters, D: generation of synthetic daily weather series, E: reconstruction of daily climatic statistics from the monthly ones. Names of the procedures are in double boxes, the main data files are in 3D boxes and the auxiliary files are in rectangles. Symbol X in the file names stands for the optional station index. The main data files are: X.wtd = the daily weather series, X.day = smoothed annual cycles of parameters of the generator, X.mon = the same characteristic as in X.day but for individual months and matrices of the autoregression model, A.mdf = climate change scenario - the set of monthly (or daily) additive (or multiplicative) increments.

Fig. 5. Annual cycles of the mean, E(SRAD), and standard deviation, s(SRAD), of the daily sum of global solar radiation. Symbols ': raw values; thick lines: values smoothed by robust locally weighted regression.

Tab.1. Parameters of the stochastic generator and number of values representing annual cycle.