Homepage of Martin
Dubrovsky Martin Dubrovsky, 1995: Preparing meteorological data for the
crop growth model "CERES". In: Contemporary
Agroclimatology (proceedings of the conference). Velke Bilovice,
September 6-9, 1995.
includes: 5 figures and 5 tables
Martin Dubrovsky
Institute of Atmospheric Physics, Hradec Kralove, Czech
Republic
In 1994-1995 the "Czech Republic's Country Study" project was accomplished by National Climatic Program of the Czech Republic under sponsorship of U.S. Country Studies Program. The project's goals included assessment of the climate change impacts on conditions for the establishment and evolution of crop stands and the magnitude of the crop yields. For the purpose of the project, we were provided with a software package DSSAT v.3 which includes crop growth model CERES (see, e.g., MEARNS et al., 1992; HUNKAR, 1994) and the WeatherMan utility program (KOHUT, 1994). WeatherMan mediates services of two weather generators (WGEN and SIMMETEO) which allow to analyze and generate time series of four daily weather characteristics (SRAD - sum of global solar radiation, TMIN - minimum temperature, TMAX - maximum temperature and RAIN - precipitation amount) required by the crop growth model.
The crop growth model simulates daily incrementing taking into account plant genetics, daily weather conditions, soil properties and management factors. The climate change impacts are assessed based on comparison of crop model runs with present-climate weather series and changed-climate weather series.
The present contribution gives a description of the procedures aiming at building up the time series of daily weather data necessary for the crop growth models to assess impacts of potetntial climate change on crop productivity. The required series include (a) observed daily weather data representing baseline climate conditions (1961-1990) and (b) synthetic daily weather data for the changed climate conditions. The methodology was already published in DUBROVSKY (1994). The procedures concerning estimation of missing radiation data are in detail described in VANICEK (1994).
Sixteen climatic stations (hereafter referred to as
reference climatic stations) were selected to represent
climate conditions of two important crop regions of the
Czech Republic - central Bohemia and southern Moravia (fig.).
The available weather data included:
(a) Daily observations of TMAX, TMIN and RAIN
in the 16 reference climatic stations (squares in fig.1)
during period 1961-1990. Selected climatic
characteristics derived from the 30-year series are
listed in tab.I.
(b) Daily totals of global solar radiation, SRAD,
measured at stations of the radiation network (horizontal
lines in fig.1) of the Czech Hydrometeorological
Institute. Of these stations, series from Hradec Kralove
are available since 1964 to 1993. The series from other
stations cover period 1984-1993. The mean annual courses
of SRAD measured at seven radiation stations used
in the project are displayed in fig.2.
(c) Average monthly sums of global solar radiation
calculated from the sunshine duration measured during
1971-1980 (CHMI, 1984). The stations whose sunshine data
were used in the project are marked by vertical lines in
fig.1.
More details on climate conditions in the two regions may be found in: BRAZDIL, ROÎNOVSKY et al. (1995) and KALVOVA (1995).
Since the radiation data were not available for the
desired period in reference climatic stations, it was
necessary to estimate daily radiation sums from the
observations in the actinometric network of CR (VANICEK,
1995). In a first step, the smoothed annual courses of
daily sums were calculated from 1984-1993 observations by
robust locally weighted regression (DUBROVSKY, 1993;
KALVOVA and DUBROVSKY, 1995). Then the linear regression
analysis was applied on the deviations of daily radiation
sums from the smoothed annual course in Hradec Kralove (independent
variable) and other stations (dependent variables) (see
tab.II for the between-stations correlations). In a third
step, the daily radiation sums for the `independent
stations' for period 1964-1983 were estimated from
measurements in Hradec Kralove with use of equation:
SRAD(j,t) = <SRAD(j,t)> + D(j,SRAD(649,t))
where t is a time series index, j is the
identifier, <.> represents the smoothed annual
course and D(j,SRAD(649,t)) is a regression
estimate of the deviation from the annual course. The
error due to the regression estimate is illustrated in
fig.3.
The daily radiation sums for the stations for which no
radiation measurements exist were spatially interpolated
from the radiation network. Regarding the coarseness of
the radiation network and great spatial variability of
solar radiation, the daily sums were adjusted with use of
the mean monthly sums of solar radiation derived from
sunshine measurements avilable in much denser station
network (mean monthly sums were taken from CHMI, 1984):
SRAD(i,t) = SRAD(i',t) + [<SRAD*(i'',t)>
- <SRAD*(i',t)>]
where SRAD(i,t) is the series of daily radiation
sums for i-th station, SRAD(i',t) are
values interpolated from daily sums measured in stations
of the radiation network, <SRAD*(i--,t)> is
a spatially interpolated smooth annual course derived
from the monthly means at nearest `sunshine stations' and
<SRAD*(i-,t)> is a spatially interpolated
smooth annual course derived from the sunshine
observations in or close to the stations of the radiation
network. The terms in the square brackets thus embody the
adjustment for the spatial variability of the solar
radiation.
The missing values of the four variables during the 30-year period (including missing radiation data during 1961-1963) were filled by weather generator WGEN.
The weather series for changed climate conditions is typically obtained either by direct modification of the existing time series or stochastically generated by the weather generator (application of the direct GCM output is not recommended because of the inaccuracy of their present versions). The former approach (illustrated in fig.4) is simple and guarantees reproduction of selected features of the stochastic structure of the weather series without need to deal with its model representation. With use of a suitable modification formula one may change, e.g., annual course of the mean and/or variance of selected quantity in a desired manner (MEARNS et al., 1992, BACSI and HUNKAR, 1994). The drawback of the direct modification method is, that the length of the changed-climate series is limited by length of the observed series which need not suffice to perform a profitable statistical assessment. In a latter approach (illustrated in fig.5), the data are stochastically generated according to the model the parameters of which are estimated from the learning data and modified in accordance with the climate change scenario. The weather generator may produce arbitrarily long weather series for given climate conditions, which allows one to perform detailed sensitivity analysis of climate-crop relations. The necessary condition of applicability of weather generator is that it sufficiently well reproduce stochastic structure of the weather series. For the purpose of the project, the PC program Met&Roll was developed (DUBROVSKY, 1995) which allows both direct modification approach and stochastic generation. The model of the stochastic generator is the same as those involved in DSSAT package. However, it differs from the two DSSAT's generators in number of parameters derived from the learning series and in more detailed representation of the annual courses of individual characteristics which both together results in better reproduction of the stochastic structure of the weather time series. Since the latter approach (employment of the weather generator) is considerably more time-consuming and since tests of capability of the weather generators (both of Met&Roll and of those included in the DSSAT software) to reproduce the stochastic structure of the weather series were not finished in due time, it was decided that the direct modification approach would be used to synthesize daily weather data for the changed climate conditions.
In a Met&Roll's direct modification approach, the
series of the four weather quantities may be modified
according to the formula the general form of which reads:
xi'(t) = xi(t)
* [dai(d(t)) + e(t).dbi(d(t))]
where xi(t) and xi'(t),
i = 1,..,4, stand for unmodified and modified
values of SRAD, TMAX, TMIN and RAIN
respectively, dai((d(t)) and
dbi(d(t)) are the
deterministic and random components of the modification
function, e(t) is a series of uncorrelated random
numbers with normal distribution, e(t) ~ N(0,1),
(white noise), and * is either additive or multiplicative
operator. The modification functions may be optionally
given either in a form of the tables (daily or monthly
coefficients) or in form of the parameters of the
harmonics. All parameters of the modification function (including
operators) may be set independently for all four
variables and separately for dry and wet days. The shape
of the modification function used in the project conforms
the following conditions:
Construction of the climate change scenarios was based on a detailed analysis of outputs of several GCMs (KALVOVA, 1995; KALVOVA and NEMESOVA, 1995). To account for the uncertainties in climate change estimates in a regional scale, the set of climate change scenarios was constructed by combining GCM outputs and expert approach (see the previously mentioned works for the details). Seven scenarios selected for the study are listed in tab.III with modification coefficients being given in tab.IV.
The present contribution describes the process of preparing the series of the daily weather characteristics - sum of global solar radiation, maximum and minimum temperature and precipitation amount - for 16 reference climatic stations in the Czech Republic. The weather series were required as an input to the crop growth model CERES to estimate impacts of potential climate change on crop production which was studied within the frame of the "Czech Republic's Country Study" project.
In a first step, the daily weather series for the baseline (1961-1990) climate conditions were assembled and then the series were modified according to the incremental climate change scenarios. The first step consisted of (a) estimating solar radiation data for the reference climatic stations from measurements at the stations of the Czech Republic's radiation network and (b) filling gaps in the weather time series by stochastic weather generator WGEN. In a second step, the series for the changed climate conditions were synthesized by direct modification of the observed weather series with coefficients being based on seven incremental scenarios.
Fig.1. Position of the stations the data of which were
used in the project. Squares with station index numbers:
reference climatic stations (see tab.I for the list);
horizontal lines with arrows and station index numbers:
stations of the radiation network (see the text below fig.2
for the list); vertical lines: stations whose
climatological sunshine data were used to localize
radiation data for the reference climatic stations. (Note:
+'s are stations with both radiation and sunshine data.
Fig.2. The smoothed annual courses of mean daily sums of
global solar radiations based on measurements during 1984-1993
at selected stations of the actinometric network of CHMI:
Kocelovice (487), Praha-Karlov (519), Kosetice (628),
Hradec Kralove (649), Svratouch (683), Kucharovice (698)
and Luka (710).
Fig.3 Distribution function of relative errors of SRAD
for Kucha«ovice estimated by linear regression from
Hradec Kralove. Line/+'s: distribution of errors for
whole year/March-July on assumption that the linear
regression function was derived from the all year data.
Squares: distribution of errors on assumption that the
estimated daily radiation sums follow the mean annual
course.
Fig.4 Direct modification of the observed series. Legend:
x(t) is an observed series (line), d(t) is
a modification function (typically periodical with length
of the period being 1 year), circle is a modification
operator (typically additive for temperature and
multiplicative for precipitation and solar radiation) and
x'(t) is a new series (line with x's).
Fig.5 Generation of the synthetic series by stochastic
generator: (i) Structure of the daily weather series is
derived from a multiple year observational data (x(t)
thin lines). (ii) Characteristics of the structure of the
series are modified (the thick/dotted line represent the
original, x-(t), and
modified, x-'(t), means).
(iii) The synthetic series, x'(t), is
stochastically generated using modified characteristics
of the series structure.
Table I. The climatic characteristics of the 16
reference climatic stations.
ind station ALT RAIN Rmax Rmin #Rmax #Rmin Tavg Tmax Tmin dT
m mm prob m mm m mm m prob m prob deg m deg m deg deg
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19)
--------------------------------------------------------------------------------------------------------------------------
11518 Ruzyne 380 526 .46 5 77.2 1 23.5 11 .53 10 .38 8.1 7 17.5 1 -2.5 20.0
11523 Hostomice 345 543 .50 6 74.8 12 25.6 6 .55 10 .41 8.5 7 17.8 1 -1.9 19.7
11561 Semcice 234 579 .53 7 72.0 2 30.2 11 .65 4 .46 8.8 7 18.2 1 -2.0 20.2
11563 Stara Boleslav 179 576 .42 6 75.2 1 28.8 6 .46 10 .35 9.2 7 18.7 1 -1.5 20.2
11572 Ondrejov 526 675 .45 6 84.2 2 37.4 6 .50 10 .34 7.7 7 17.2 1 -2.9 20.1
11627 Cechtice 490 716 .51 6 85.0 10 42.8 1 .55 10 .46 7.8 7 16.8 1 -2.5 19.3
11636 Kostelni Myslova 569 583 .48 6 79.3 3 29.7 11 .56 10 .44 7.2 7 16.7 1 -3.4 20.1
11649 Hradec Kralove 278 617 .46 8 83.1 3 33.8 12 .55 10 .40 8.6 7 18.0 1 -2.2 20.2
11659 Pribyslav 530 677 .48 6 91.2 3 36.9 12 .58 10 .37 6.9 7 16.1 1 -3.6 19.7
11687 Velke Mezirici 452 594 .47 6 74.1 10 32.6 1 .57 10 .35 7.2 7 16.6 1 -3.6 20.2
11698 Kucharovice 334 471 .41 6 74.9 1 20.1 6 .47 10 .30 8.7 7 18.5 1 -2.4 20.9
11723 Brno - Turany 241 490 .39 6 72.2 3 23.4 12 .45 10 .28 8.9 7 18.6 1 -2.5 21.1
11754 Stare Mesto 235 536 .39 6 77.5 3 26.1 6 .47 9 .33 9.1 7 18.6 1 -2.2 20.8
11755 Straznice 176 537 .36 6 86.2 1 23.6 6 .43 10 .29 9.0 7 18.3 1 -2.0 20.3
11774 Holesov 224 613 .45 6 86.3 1 26.8 12 .52 9 .38 8.6 7 18.1 1 -2.7 20.8
11779 Strani 421 799 .47 6 94.6 3 44.5 12 .56 10 .36 7.9 7 17.2 1 -3.0 20.2
Legend: ind (column 1): station index number; ALT
(c.3) = altitude (m a.s.l.); RAIN: (c.4) = total annual
precipitation, (c.5) = mean annual probability of
occurrence of the wet day; Rmax: (c.6) = month with max.
amount of precipitation, (c.7) = max. monthly
precipitation amount; Rmin: (c.8) = month with minimum
amount of precipitation, (c.9) = minimum monthly
precipitation amount; #Rmax: (c.10) = month with max.
probability of occurrence of the wet day, (c.11) =
probability; #Rmin: (c.12) = month with minimum
probability of occurrence of the wet day, (c.13) =
probability; Tavg: (c.14) = mean annual temperature (annual
mean of the daily mean temperature, which is here defined
as an average of daily minimum and daily max. temperature);
Tmax: (c.15) = month with highest mean temperature, (c.16)
= highest mean monthly temperature; Tmin: (c.17) = month
with lowest mean temperature, (c.18) = lowest mean
monthly temperature; dT (c.19) = annual temperature
amplitude (= Tmax Ä Tmin).
Table II. Between-stations correlation coefficients of deviations of the daily solar radiation sums from the mean annual course (the numbers in the left column and upper row are station indeces).
| 11487 | 11519 | 11628 | 11683 | 11698 | 11710 --------------------------------------------------------------------------- 11649 | 0.76 | 0.84 | 0.85 | 0.89 | 0.73 | 0.81
Table III. The list of incremental scenarios of climate change (see tab.IV for the values of the coefficients for individual months).
scenario | D(srad) | D(temp) | D(rain)
----------------------------------------------------------------
1 | GISS30 | GISS30 | -5%
2 | GISS30 | GISS30 | =
3 | GISS30 | GISS30 | +5%
4 | GISS30 | GISS30 | =|X
5 | GISS30 | GISS30 | GISS30
6 | GISS30 | GISS30 | CCCM75
7 | CCCM30 | CCCM30 | CCCM30
Table IV. Parameters of the incremental scenarios.
| D(srad) | D(temp) | D(rain)
|------------------------------------------------------------------------------------------------
| GISS | CCCM | GISS | CCCM | -5% | = | +5% | =|X | GISS | CCCM | CCCM
| 30 | 30 | 30 | 30 | | | | | 30 | 75 | 30
==========================================================================================================
YEAR | 0.99 | 1.00 | 1.97 | 1.44 | 0.95 | 1 | 1.05 | 1.000| 1.082 | 1.10 | 1.047
JAN | 0.96 | 0.95 | 2.90 | 1.54 | 0.95 | 1 | 1.05 | 1.051| 1.085 | 1.32 | 1.151
FEB | 0.94 | 0.96 | 2.70 | 2.01 | 0.95 | 1 | 1.05 | 1.257| 1.188 | 1.28 | 1.132
MAR | 0.99 | 1.01 | 2.30 | 1.65 | 0.95 | 1 | 1.05 | 1.247| 1.094 | 0.96 | 0.981
APR | 0.98 | 0.98 | 1.90 | 1.14 | 0.95 | 1 | 1.05 | 1.129| 1.193 | 1.35 | 1.165
MAY | 1.01 | 1.03 | 1.50 | 1.11 | 0.95 | 1 | 1.05 | 0.962| 1.094 | 0.83 | 0.920
JUN | 1.01 | 1.00 | 1.30 | 1.33 | 0.95 | 1 | 1.05 | 0.815| 1.020 | 0.90 | 0.953
JUL | 1.02 | 1.03 | 1.20 | 1.59 | 0.95 | 1 | 1.05 | 0.707| 1.047 | 0.82 | 0.915
AUG | 1.00 | 1.00 | 1.30 | 1.33 | 0.95 | 1 | 1.05 | 0.805| 1.042 | 1.06 | 1.028
SEP | 1.02 | 1.03 | 1.50 | 1.58 | 0.95 | 1 | 1.05 | 0.953| 1.010 | 0.95 | 0.976
OCT | 0.99 | 0.99 | 1.90 | 1.45 | 0.95 | 1 | 1.05 | 1.100| 1.089 | 1.08 | 1.038
NOV | 0.97 | 0.98 | 2.40 | 1.55 | 0.95 | 1 | 1.05 | 0.982| 1.089 | 1.30 | 1.141
DEC | 0.95 | 0.99 | 2.70 | 1.04 | 0.95 | 1 | 1.05 | 0.992| 1.038 | 1.35 | 1.165
GISS30: GISS scenario for 2030 /// CCCM30: CCCM
scenario for 2030
CCCM75: CCCM scenario for 2075 (2xCO2)
=|X: annual precipitation amount is unchanged, annual
course of precipitation amount is modified according to
the observed precipitation changes during past 50 years