1
Using a gridded global data set to characterize regional hydroclimate in central Chile
2
3
E.M.C. Demaria1, E. Maurer2*, J. Sheffield3, E. Bustos1, D. Poblete1, S. Vicuña1, F. Meza1
4
5
1
Centro de Cambio Global, Pontificia Universidad Católica de Chile, Santiago, Chile
6
2
Civil Engineering Department, Santa Clara University, Santa Clara, CA, USA
7
3
Department of Civil and Environmental Engineering, Princeton University, Princeton, NJ, USA
8
9
*Corresponding author, emaurer@engr.scu.edu, 408-554-2178.
10
11
Proposed submission to J. Hydrometeorology
12
13
1
14
Abstract
15
Central Chile is facing dramatic projections of climate change, with a consensus for declining
16
precipitation, negatively affecting hydropower generation and irrigated agriculture. Rising from
17
sea level to 6,000 meters within a distance of 200 kilometers, precipitation characterization is
18
difficult due to a lack of long-term observations, especially at higher elevations. For
19
understanding current mean and extreme conditions and recent hydroclimatological change, as
20
well as to provide a baseline for downscaling climate model projections, a temporally and
21
spatially complete data set of daily meteorology is essential. We use a gridded global daily
22
meteorological data set at 0.25 degree resolution for 1948-2008, and adjust it using monthly
23
precipitation observations interpolated to the same grid using a cokriging method with elevation
24
as covariate. For validation, we compare daily statistics of the adjusted gridded precipitation to
25
station observations. For further validation we drive a hydrology model with the gridded 0.25-
26
degree meteorology and compare stream flow statistics with observed flow. We validate the high
27
elevation precipitation by comparing the simulated snow extent to MODIS images. Results show
28
that the daily meteorology with the adjusted precipitation can accurately capture the statistical
29
properties of extreme events as well as the sequence of wet and dry events, with hydrological
30
model results displaying reasonable agreement for observed flow statistics and snow extent. This
31
demonstrates the successful use of a global gridded data product in a relatively data-sparse
32
region to capture hydroclimatological characteristics and extremes.
33
34
35
2
36
Introduction
37
38
Whether exploring teleconnections for enhancing flood and drought predictability or assessing
39
the potential impacts of climate change on water resources, understanding the response of the
40
land surface hydrology to perturbations in climate is essential. This has inspired the development
41
and assessment of many large scale hydrology models for simulating land-atmosphere
42
interactions over regional and global scales [e.g.,Lawford et al., 2004; Milly and Shmakin, 2002;
43
Nijssen et al., 2001a; Sheffield and Wood, 2007].
44
45
A prerequisite to regional hydroclimatological analyses is a comprehensive, multi-decadal,
46
spatially and temporally complete data set of observed meteorology, whether for historic
47
simulations or as a baseline for downscaling future climate projections. In response to this need,
48
data sets of daily gridded meteorological observations have been generated, both over
49
continental regions [e.g.,Cosgrove et al., 2003; Maurer et al., 2002] and globally [Adam and
50
Lettenmaier, 2003; Sheffield et al., 2006]. These have benefited from work at coarser time scales
51
[Chen et al., 2002; Daly et al., 1994; Mitchell and Jones, 2005; New et al., 2000; Willmott and
52
Matsuura, 2001], with many products combining multiple sources, such as station observations,
53
remotely sensed images, and model reanalyses.
54
55
While these large-scale gridded products provide opportunities for hydrological simulations for
56
land areas around the globe, they are inevitably limited in their accuracy where the underlying
57
station observation density is low, the station locations are inadequate to represent complex
58
topography, or where the gridded spatial resolution is too large for the region being studied.
3
59
Central Chile is an especially challenging environment for characterizing climate and hydrology
60
since the terrain exhibits dramatic elevation changes over short distances, and the orographic
61
effects this drives produce high spatial heterogeneity in precipitation in particular. In general, the
62
observation station density in South America is inadequate for long-term hydroclimate
63
characterization [de Goncalves et al., 2006]. While some of South America is relatively well
64
represented by global observational datasets [Silva et al., 2007], regions west of the Andes are
65
much less so [Liebmann and Allured, 2005].
66
67
In this study, we utilize a new high-resolution global daily gridded dataset of temperature and
68
precipitation, adjust it with available local climatological information, and assess its utility for
69
representing river basin hydrology. Recognizing the value in simulating realistic extreme events,
70
we assess the new data product for its ability to produce reasonable daily streamflow statistics.
71
We evaluate the potential to reproduce climate and hydrology in a plausible manner, such that
72
historical statistics are reproduced.
73
74
The principal aim of this study is to produce a gridded representation of the climate and
75
hydrology of central Chile, are demonstrate a methodology for producing a reasonable set of data
76
products that can be used for future studies of regional hydrology or climate. Given these
77
regional results, we assess the potential to export the method to other relatively data-sparse
78
regions, where representative climatological average information is available but long-term daily
79
data are inadequate. The paper is organized as follows: Section 2 describes the study area. In
80
Section 3 we describe the data, the hydrological model and the methodological approach. Results
4
81
of the adjusted data set validation and model simulations are discussed in Section 4. Finally, the
82
main conclusions of the study are presented in Section 5.
83
84
Region
85
86
The focus area of this study is central Chile (Figure 1), encompassing the four major river basins
87
(from north to south, the Rapel, Mataquito, Maule, and Itata Rivers) between latitudes 35.25º S
88
and 37.5º S. The climate is Mediterranean, with 80% of the precipitation falling in the rainy
89
season from May-August [Falvey and Garreaud, 2007]. The terrain is dramatic, rising
90
approximately 6000 meters within a horizontal distance of approximately 200 km, producing
91
sharp gradients in climate [Falvey and Garreaud, 2009].
92
93
Driven by the terrain, the area exhibits a dramatic climate gradient, with mean precipitation of
94
approximately 500 mm per year at the North end of the study domain, and as much as 3000 mm
95
per year in the high elevations at the Southern end of the domain. It is evident from Figure 1 that
96
the high elevation areas are under-represented by any of the observation stations.
97
98
The region of Central Chile is especially important from a hydroclimatological standpoint, as it
99
contains the largest proportion of irrigated agriculture and reservoir storage of any region in the
100
country and provides water supply for some of Chile's largest cities. A changing climate is
101
evident in recent hydroclimate records [Rubio-Álvarez and McPhee, 2010], and future climate
102
projections for the region indicate the potential for very large impacts [Bradley et al., 2006]. The
103
vulnerability of Central Chile to projected climate change is high, with robust drying trends in
5
104
General Circulation Model (GCM) projections, and a high sensitivity to changing snow melt
105
patterns [Vicuna et al., 2010], who also discuss the challenges in characterizing climate in a
106
Chilean catchment with few precipitation observations, and none at high elevations.
107
108
Methods and data
109
110
Gridded data set development
111
112
We begin with a gridded global (land surface) dataset of daily precipitation and minimum and
113
maximum temperatures at 0.25º spatial resolution (approximately 25 km), prepared following
114
Sheffield et al. [2006]. To summarize, the forcing dataset is based on the NCEP–NCAR
115
reanalysis [Kalnay et al., 1996] for 1948-2008, from which daily maximum and minimum
116
temperature and daily precipitation are obtained at approximately 2º spatial resolution.
117
Reanalysis temperatures are based on observations, though precipitation is a model output and
118
thus exhibits significant biases.
119
120
The reanalysis temperatures are interpolated to a 0.25º spatial resolution, lapsing temperatures by
121
-6.5ºC/km based on the elevation difference between the large reanalysis spatial scale and the
122
elevation in each 0.25º grid cell. Precipitation is interpolated to 0.25º using a product of the
123
Tropical Rainfall Measuring Mission (TRMM) [Huffman et al., 2007] following the methods
124
outlined by Sheffield et al. [2006]. To ensure large-scale correspondence between this data set
125
and the observationally-based monthly 0.5º data from the Climate Research Unit [CRU, Mitchell
126
and Jones, 2005], precipitation is scaled so the monthly totals match the CRU monthly values at
6
127
the CRU spatial scale. Maximum and minimum temperatures are also scaled to match the CRU
128
time series, using CRU monthly mean temperature and diurnal temperature range.
129
130
While the incorporation of multiple sources of extensively reviewed data provides an invaluable
131
data product for global and continental scale analyses, as discussed by Mitchell and Jones [2005]
132
ultimately much of the local characterization is traceable to a common network of land surface
133
observations [Peterson et al., 1998], which is highly variable in station density for different
134
regions. For example, for the region of study shown in Figure 1, an average of 3-4 observation
135
stations are included in the CRU precipitation data product, and none are in high-elevation areas.
136
This results in a few low elevation meteorological stations in Chile on the western side of the
137
Andes, and the next observation station to the east is in a more arid area in Argentina. Thus, the
138
resulting precipitation fields in the gridded product for this region showed a spatial gradient
139
opposite to that published by the Dirección General de Aguas [DGA, 1987]. Figure 2a shows the
140
spatial distribution of gridded global total annual precipitation that displays a notable decrease of
141
rainfall with a elevation. Conversely the DGA precipitation map is able to capture the
142
climatological orographic enhancement of precipitation by the Andes (Figure 2b) . The
143
precipitation lapse rates for the latitudinal bands -35.125º S and -36.125º S show a negative
144
gradient of precipitation with elevation in the global gridded data set whereas the DGA
145
precipitation shows a positive gradient for the period 1951-89 (Figures 2c and 2d, respectively).
146
147
Local data from the DGA of Chile, some monthly and some daily, were obtained to characterize
148
better the local climatology. While still biased toward low elevation areas, the stations (Figure 1)
149
do cover a wider range and include altitudes up to 2400 m. These stations were filtered to include
7
150
those that had at least 90% complete monthly records for the 25-year period 1983-2007. The
151
monthly average precipitation for the 25-year period for these 40 DGA stations was interpolated
152
onto the same 0.25º grid using cokriging, with elevation being the covariate. This method of
153
cokriging has been shown to improve kriging interpolation to include orographic effects induced
154
by complex terrain [Diodato and Ceccarelli, 2005; Hevesi et al., 1992].
155
156
This process produced 12 monthly mean precipitation maps for the region. The same 1983-2007
157
period was extracted from the daily gridded data set, and monthly average values were calculated
158
for each grid cell. Ratios (12, one for each month) of observed climatology divided by the
159
gridded data set average were then calculated for each grid cell. Daily values in the gridded data
160
set were adjusted to create a new set of daily precipitation data, Padj, which matches the
161
interpolated observations produced with cokriging, using a simple ratio:
Padj i, j, t   Pgrid i, j, t  
Pobs,mon i, j 
Pgrid,mon i, j 
(1)
162
where Pgrid is the original daily gridded 0.25º data at location (i,j), Pobs is the interpolated
163
observed climatology, overbars indicate the 25-year mean, and the subscript “mon” indicates the
164
month from the climatology in which day t falls.
165
166
This same method was applied to a global dataset of daily meteorology in a data sparse region in
167
Central America, resulting in improved characterization of precipitation and land surface
168
hydrology [Maurer et al., 2009]. In addition, this new adjusted data set includes the full 1948-
169
2008 period, despite the fact that local observations are very sparse before 1980.
170
8
171
To validate the adjusted precipitation data set, we computed a set of statistical parameters widely
172
used to describe climate extremes [dos Santos et al., 2011; X. Zhang and Yang, 2004].
173
Additionally to evaluate the temporal characteristics of rainfall events we computed the wet, dry
174
and transition probabilities. Table 1 shows a description of the statistics used.
175
176
To evaluate if the adjusted precipitation data set was capturing the orographic gradient of
177
precipitation we compared VIC simulated Snow Water Equivalent (SWE) to the MODIS/Terra
178
Snow Cover data set, which is available at 0.05 degree resolution for 8-day periods starting from
179
the year 2000. MODIS snow cover data are based on a snow mapping algorithm that employs a
180
Normalized Difference Snow Index [Hall et al., 2006]. To estimate snow cover from the
181
meteorological data, a hydrological model was employed.
182
183
Hydrologic Model Simulations
184
185
To assess the ability of the daily gridded meteorology developed in this study to capture daily
186
climate features across the watersheds, we simulate the hydrology of river basins in the region to
187
obtain streamflow and snow cover estimates. The hydrologic model used is the Variable
188
Infiltration Capacity (VIC) model [Cherkauer et al., 2003; Liang et al., 1994]. The VIC model is
189
a distributed, physically-based hydrologic model that balances both surface energy and water
190
budgets over a grid mesh. The VIC model uses a “mosaic” scheme that allows a statistical
191
representation of the sub-grid spatial variability in topography, infiltration and vegetation/land
192
cover, an important attribute when simulating hydrology in heterogeneous terrain. The resulting
193
runoff at each grid cell is routed through a defined river system using the algorithm developed by
9
194
Lohmann et al. [1996]. The VIC model has been successfully applied in many settings, from
195
global to river basin scale [e.g.,Maurer et al., 2002; Nijssen et al., 2001b; Sheffield and Wood,
196
2007].
197
198
For this study, the model was run at a daily time step at a 0.25º resolution (approximately 630
199
km2 per grid cell for the study region). Elevation data for the basin routing are based on the 15-
200
arc-second Hydrosheds dataset [Lehner et al., 2006], derived from the Shuttle Radar Topography
201
Mission (SRTM) at 3 arc-second resolution. Land cover and soil hydraulic properties were based
202
on values from Sheffield and Wood [2007], though specified soil depths and VIC soil parameters
203
were modified during calibration. The river systems contributing to selected points were defined
204
at a 0.25º resolution, following the technique outlined by O’Donnell et al. [1999].
205
206
Results and Discussion
207
208
The adjusted data set was validated in several ways. First, daily statistics were compared
209
between the adjusted global daily data set and local observations, where available. Second,
210
hydrologic simulation outputs were compared to observations to investigate the plausibility of
211
using the new data set as an observational baseline for studying climate impacts on hydrology.
212
213
Gridded meteorological data development and assessment
214
215
The quality of daily gridded precipitation fields was improved using available monthly observed
216
precipitation. Rain gauge records from DGA were selected using two criteria: stations with
10
217
records of twenty-five years and with no more than 10% missing daily measurements. Based on
218
those two constraints the period 1983-2007 was identified as that with the largest number of
219
reporting stations. From the pool of 70 available stations, 40 stations met the two criteria (Figure
220
1). Except for the Itata river basin, which had two stations located at 1200 and 2400 meters
221
above sea level, most of the selected stations were located in the central part of the region at
222
elevations below 500 meters. Mean precipitation was computed for each month and for each
223
selected station, resulting in 12 mean values for the 25-year climatological period.
224
225
Cokriging was then applied to produce a set of 12 maps of climatological precipitation at 0.25º
226
spatial resolution. A scatter plot between observed and predicted monthly precipitation for July,
227
the middle of the rainy season, is shown in Figure 3. Cokriged monthly totals match
228
observations quite closely for the region with a bias equal to -0.8 % with respect to the observed
229
values and a relative RMSE of 0.50 %.
230
231
Figure 4 shows the adjusted gridded annual precipitation fields and the difference from the
232
original gridded observed data set for the period 1950-2006. It is evident that in the more humid
233
southern mountainous portion of the study area there has been a marked increase in precipitation
234
with the adjustment, incorporating the more detailed information embedded in the rain gauge
235
observations. Differences between original and adjusted gridded precipitation indicates the
236
existence of a band along the Andes where annual precipitation is greater in the adjusted
237
precipitation data set (Figure 4b).
238
11
239
To verify how the adjusted daily precipitation relates to observations, we compared daily rainfall
240
at selected 0.25º grid points with the day-by-day means of three rain gauge stations located in
241
approximately a 50 km diameter circle (Figure 5). Rain gauge stations were selected from the
242
pool of 40 stations used to perform the cokriging interpolation, hence they had record of 25 years
243
with not more than 10% missing values. Selected stations were located, when possible, not more
244
that 50% higher or lower elevations than that of the 0.25º grid cell. Four 0.25º grid points were
245
selected for the comparison. The locations of the four grid points are listed in Table 2. For these
246
four locations, we computed basic statistics, bias, RMSE and correlation coefficient for daily
247
observed (OBS) and daily adjusted gridded precipitation (ADJ) for Austral summer (DJF) and
248
Austral winter (JJA) for the period 1983-2007. Summary statistics are shown in Table 3. The
249
bias is defined as the sum of the differences between ADJ and OBS and the RMSE is equal to
250
the root mean squared error between the ADJ and OBS daily precipitation values.
251
252
Mean daily values are very close for the observed and adjusted datasets for both seasons, which
253
is expected given the adjustment process. The variability of daily precipitation within each
254
season, represented by the standard deviation, also compares relatively well, though the adjusted
255
gridded data show greater variability than the observations during the rainy winter season. A
256
high RMSE and low correlation values indicate that temporal sequencing differs between the two
257
data sets. This is not unexpected, since the daily precipitation in the original 0.25º gridded data
258
was derived from reanalysis, and as such it is a model output that does not incorporate station
259
observations [Kalnay et al., 1996]. Thus, while important characteristics of daily precipitation
260
variability are represented in the 0.25º gridded data, and monthly totals should bear resemblance
12
261
to observations (at least as represented by the underlying monthly data such as CRU),
262
correspondence with observed daily precipitation events is not anticipated.
263
264
With this limitation in mind, we focus on the statistics of daily hydroclimatology, rather than
265
event-based statistics. Especially given the rising interest in characterizing extreme events in the
266
context of a changing climate [IPCC, 2011], the ability of the adjusted daily gridded dataset to
267
characterize extreme statistics is important. We compute a set of statistical variables frequently
268
used to describe climate extremes, using the RClimDex software [X. Zhang and Yang, 2004;
269
Xuebin Zhang et al., 2005]. Additionally we computed the probability of occurrence of wet and
270
dry days and transition probabilities. Figure 6 shows box plots of six statistical parameters listed
271
in Table 1 for the four locations. Extreme precipitation events (R95p) are well captured in the
272
adjusted gridded data set at most locations. The agreement between adjusted total annual
273
precipitation (PRCPTOT) and observations is good with an average bias of -9% from the
274
observed station mean (not shown), although this is constrained by design, as noted above. The
275
Simple Daily Intensity Index (SDII), which is a measure of the mean annual intensity of rainfall,
276
also shows good agreement at the four locations, indicating that the number of rainy days is well
277
represented in the adjusted dataset. The number of days with intensities larger than the 20 mm
278
(R20mm) compares well between observations and adjusted gridded precipitation, though the
279
adjusted gridded data slightly underestimate observations. The maximum consecutive number of
280
dry days and wet days in a year is lower for the adjusted gridded observations compared to
281
observations suggesting the durations of wet and dry events are shorter in the adjusted gridded
282
data set.
283
13
284
Figure 7 shows that the probabilities of a day being wet or dry are comparable between both
285
datasets (panels 7a and 7b). Conversely the adjusted precipitation data shows an average
286
transition probability of a wet day followed by a wet day of 0.21 compared to 0.50 obtained for
287
the observations suggesting that the duration of storm events is shorter in the adjusted gridded
288
dataset. This could partially explain the underestimation of maximum consecutive wet days
289
(CWD) in the adjusted gridded precipitation as well.
290
291
The statistical quantities presented in Figures 6 and 7 were compared statistically using the
292
correlation coefficient and a two-sample unpaired Student’s t-test. These are summarized in
293
Table 4. Statistics linked to high intensity events (R99p and R95p and maximum 1-day
294
precipitation (RX1day) have statistically indistinguishable means for all four locations. The
295
mean annual precipitation and intensity parameters (Prcptot and SDII) show equal means for
296
three out of the four locations. Conversely the parameters, R5mm R20mm, albeit strongly
297
correlated, were found to have statistically different mean values. This phenomenon of a gridded
298
precipitation data produce having lower extreme precipitation values than station observations
299
was also noted in the South American study of Silva et al. [2007] and is consistent with the effect
300
of spatial averaging, i.e., comparing the average of a 630 km2 0.25º grid cell to the smaller, more
301
discrete area represented by the three averaged stations [Yevjevich, 1972]. The statistics related
302
to duration of wet and dry spells showed statistically different population means at all 4
303
locations.
304
305
Hydrologic Model Validation of Adjusted Meteorology
306
14
307
To assess the representation in the new meteorological data set of basin-wide and high elevation
308
areas, the adjusted gridded data developed and assessed in the previous sections were then used
309
to drive the VIC hydrologic model. Since the precipitation was shown to be comparable to
310
observations (where available) in many important respects, another validation of the driving
311
meteorology would be the successful simulation of observed streamflow and snow cover.
312
Records of observed streamflow in the region tend to be incomplete or for short periods, and
313
since most of the rivers are affected by reservoirs and diversions the flows often do not reflect
314
natural streamflow as simulated by the VIC model. For this project, we focused on three sites,
315
which are shown in Figure 1.
316
317
For the site on the Mataquito River, the VIC model was calibrated to monthly stream flows for
318
the period 1990-1999 using the Multi-Objective Complex Evolution (MOCOM-UA) algorithm
319
[Yapo et al., 1998]. The three optimization criteria used in this study were the Nash-Sutcliff
320
model efficiency [Nash and Sutcliffe, 1970] using both flow (NSE) and the logarithm of flow
321
(NSElog), and the bias, expressed as a percent of observed mean flow. This provides a balance
322
between criteria that penalize errors at high flows and others that are less sensitive to a small
323
number of large errors at high flows [Lettenmaier and Wood, 1993]. Figure 8 shows the VIC
324
simulation results for the calibration period and for a validation period of 2000-2007. The flows
325
for both periods generally meet the criteria for “satisfactory” calibration based on the criteria of
326
Moriasi et al. [2007], with a NSE > 0.50 and absolute bias < 25% (the third criterion of Moriasi
327
was not calculated for this experiment). While during the validation period several of the
328
maximum annual flow peaks are overestimated, resulting in a lower NSE score compared to the
15
329
calibration period, the reasonable peaks, low flows, and satisfactory calibration and validation do
330
serve to provide further validation of the driving meteorology as plausible.
331
332
Despite the highly variable precipitation across the study region, we applied the same VIC
333
calibrated parameters from the Mataquito basin to the entire domain and used the VIC model to
334
generate streamflow at the other two gage sites. This avoids the possibility of allowing extensive
335
calibration to hide meteorological data deficiencies. The simulated flows for the period 2000-
336
2007 for each site, and the associated statistics, are in Figures 9 and 10. The simulated flows on
337
average show little bias in both locations. The Claro River NSElog value is low, reflecting the
338
underestimation of low flows and overestimation of peak flows during the simulation period,
339
though the higher NSE value suggests the errors at the high flows are not as systematic. The
340
Loncomillo River displays a general overestimation by VIC of low flows, though both NSE and
341
NSElog are above the “satisfactory” threshold. While these are not demonstrations of the best
342
hydrologic model that could be developed for each basin, or the best that the VIC model could
343
produce (since no calibration was performed for two of the three basins), they do provide some
344
further validation that the driving meteorology appears plausible, and does not appear to show
345
any systematic biases.
346
347
A comparison of four streamflow properties are shown in Figure 11 for the three simulated
348
basins. We calculate the center timing (CT), defined as the day when half the annual (water year)
349
flow volume has passed a given point [Stewart et al., 2005], where the water year runs from
350
April 1 through March 31. CT values lie within the -11 to 17 day window compared to observed
351
values, indicating the snow melting season is reasonably captured by the model (Figure 11a).
16
352
The unpaired Student’s t-test indicates the distributions have equal means at a 5% significance
353
level. The water year volume and the 3-day peak flow are systematically overestimated by VIC
354
simulations, however their means are found to statistically equal with the exception of the Rio
355
Claro 3-day peak flow. Low flows are over and underestimated by VIC simulations but only the
356
Loncomillo River has means that are statistically different (Figure 11d).
357
358
Recognizing the high dependence of this region on snow melt and thus the importance of this
359
process being well represented, we validate the high elevation meteorology of the new data set
360
by comparing VIC simulated SWE to MODIS 8-day global snow coverage for six events
361
between 2002 and 2007. The satellite images were selected in mid August to capture the
362
maximum snow accumulation in the region. Following Maurer et al. [2003] a snow depth of
363
25.4 mm was used as threshold to indicate the presence of snow on the ground. MODIS snow
364
coverage was interpolated to a 0.25º grid using triangle-based cubic interpolation. VIC simulated
365
SWE was averaged to match the MODIS eight-day period. Strong similarities in the spatial
366
extent is found between MODIS and VIC simulated the snow coverage for the period August
367
21-28, 2002 (Figure 12). The average area covered by snow in the six years is 172,320 and
368
167,050 km2 in VIC simulations and MODIS, respectively. This represents a 3% error in the
369
SCA simulated by the VIC model, which is very small.
370
371
Table 5 is a contingency table of relative frequencies of snow/no snow in MODIS and VIC
372
simulated SWE. We include all the pixels for the six selected periods (total 1530). The number
373
of pixels classified as snow or no snow are similar in VIC and MODIS with frequencies of 0.65
374
and 0.24% for no snow and snow classification, respectively. Conversely the occurrence of
17
375
misclassified snow/no snow events is quite low, in the order of 0.06% indicating an excellent
376
agreement between both data sources.
377
378
Conclusions
379
380
In this study an adjusted gridded daily precipitation data set is developed for Central Chile for
381
the period 1948-2008. Rain gauge data are used to correct the inaccuracies in the representation
382
of orographic distribution of precipitation existent in the available global gridded data set.
383
Adjusted gridded data are validated using station observations and hydrological model
384
simulations.
385
386
In data-sparse regions, a simple cokriging method that incorporates topographic elevation as
387
covariate can be successfully used to improve the spatial representation of gridded precipitation
388
in areas with complex terrain. A month-to-month adjusting can effectively remove biases in
389
precipitation values hailing from few or nonexistent rain gauge observations.
390
391
The adjusted gridded precipitation is able to capture precipitation enhancement due to orography
392
in the region with a good representation of annual totals and precipitation intensity. However the
393
duration of storm events is slightly shorter than observed perhaps as a result of comparing a 630
394
km2 grid cell to the smaller, more discrete, areal precipitation represented by three averaged rain
395
gauges. The statistics of extreme precipitation events are well captured by the adjusted gridded
396
data set which encourages its use for climate change applications.
397
18
398
Streamflow simulations in three basins realistically capture high and low flows statistical
399
properties indicating that the driving meteorology in the adjusted gridded data set is well
400
represented. Simulated SWE closely resembles satellite observations which can be linked to a
401
good depiction of winter rainfall at higher elevations, despite the driving meteorological dataset
402
including no high elevation station observations.
403
404
Based on our results, the adjusted daily gridded precipitation data set is very useful for
405
hydrologic simulations of climate variability and change in Central Chile. However there are two
406
caveats. First, we assume the period 1983-2007 is representative of a longer time period, hence
407
long term variability of precipitation is assumed properly captured. Second, the sensitivity of the
408
results to the number of rain gauges used to obtain plausible adjusted values was not determined.
409
Despite those, the methodology presented in this paper can be implemented in numerous data-
410
sparse basins located in mountainous regions around the globe.
411
412
Acknowledgements
413
This study was funded by CORFO INNOVA grant to the Centro de Cambio Global and the
414
Departamento de Ingeniería Hidráulica y Ambiental at the Pontificia Universidad Católica de
415
Chile. A Fulbright Visiting Scholars Grant also provided partial support to the second author.
416
19
417
References
418
Adam, J. C., and D. P. Lettenmaier (2003), Adjustment of global gridded precipitation for
419
420
systematic bias, J. Geophys Res., 108(D9), 1-14.
Bradley, R. S., M. Vuille, H. F. Diaz, and W. Vergara (2006), Threats to Water Supplies in the
421
422
Tropical Andes, Science, 312(5781), 1755-1756.
Chen, M., P. Xie, J. E. Janowiak, and P. A. Arkin (2002), Global Land Precipitation: A 50-yr
423
424
Monthly Analysis Based on Gauge Observations, J. Hydrometeorology, 3(3), 249-266.
Cherkauer, K. A., L. C. Bowling, and D. P. Lettenmaier (2003), Variable infiltration capacity
425
cold land process model updates, Global Plan. Change, 38, 151-159.
426
Cosgrove, B. A., D. Lohmann, K. E. Mitchell, P. R. Houser, E. F. Wood, J. C. Schaake, A.
427
Robock, C. Marshall, J. Sheffield, Q. Duan, L. Luo, R. W. Higgins, R. T. Pinker, J. D.
428
Tarpley, and J. Meng (2003), Real-time and retrospective forcing in the North American
429
Land Data Assimilation System (NLDAS) project, J. Geophys. Res., 108(D22), 8842.
430
Daly, C., R. P. Neilson, and D. L. Phillips (1994), A statistical-topographic model for mapping
431
climatological precipitation over mountainous terrain, Journal of Applied Meteorology, 33,
432
140-158.
433
de Goncalves, L. G. G., W. J. Shuttleworth, B. Nijssen, E. J. Burke, J. A. Marengo, S. C. Chou,
434
P. Houser, and D. L. Toll (2006), Evaluation of model-derived and remotely sensed
435
precipitation products for continental South America, J. Geophys. Res., 111(D16), D16113.
436
DGA (1987), Balance hidrico de ChileRep., Ministerio de Obras Publicas, Direccion General de
437
Aguas, Santiago, Chile.
20
438
Diodato, N., and M. Ceccarelli (2005), Interpolation processes using multivariate geostatistics
439
for mapping of climatological precipitation mean in the Sannio Mountains (southern Italy),
440
Earth Surface Processes and Landforms, 30(3), 259-268.
441
dos Santos, C. A. C., C. M. U. Neale, T. V. R. Rao, and B. B. da Silva (2011), Trends in indices
442
for extremes in daily temperature and precipitation over Utah, USA, Int. J. Climatol., 31(12),
443
1813-1822.
444
Falvey, M., and R. Garreaud (2007), Wintertime Precipitation Episodes in Central Chile:
445
Associated Meteorological Conditions and Orographic Influences, J. Hydrometeorology,
446
8(2), 171-193.
447
Falvey, M., and R. D. Garreaud (2009), Regional cooling in a warming world: Recent
448
temperature trends in the southeast Pacific and along the west coast of subtropical South
449
America (1979&#8211;2006), J. Geophys. Res., 114(D4), D04102.
450
Hall, D. K., G. S. Riggs, and V. V. Salomonson (2006), Updated daily. MODIS/Terra Snow
451
Cover 8-Day L3 Global 0.05deg CMG V005, Digital media, edited, National Snow and Ice
452
Data Center, Boulder, Colorado USA.
453
Hevesi, J. A., J. D. Istok, and A. L. Flint (1992), Precipitation Estimation in Mountainous
454
Terrain Using Multivariate Geostatistics. Part I: Structural Analysis, Journal of Applied
455
Meteorology, 31(7), 661-676.
456
Huffman, G. J., D. T. Bolvin, E. J. Nelkin, D. B. Wolff, R. F. Adler, G. Gu, Y. Hong, K. P.
457
Bowman, and E. F. Stocker (2007), The TRMM Multisatellite Precipitation Analysis
458
(TMPA): Quasi-Global, Multiyear, Combined-Sensor Precipitation Estimates at Fine Scales,
459
J. Hydrometeorology, 8(1), 38-55.
21
460
IPCC (2011), Intergovernmental Panel on Climate Change Special Report on Managing the
461
Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, Summary for
462
Policymakers, Cambridge University Press, Cambridge, United Kingdom and New York,
463
NY, USA.
464
Kalnay, E., M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G.
465
White, J. Woollen, Y. Zhu, A. Leetmaa, and B. Reynolds (1996), The NCEP/NCAR 40-year
466
reanalysis project, Bull. Am. Met. Soc., 77(3), 437-472.
467
Lawford, R. G., R. Stewart, J. Roads, H. J. Isemer, M. Manton, J. Marengo, T. Yasunari, S.
468
Benedict, T. Koike, and S. Williams (2004), Advancing Global-and Continental-Scale
469
Hydrometeorology: Contributions of GEWEX Hydrometeorology Panel, Bull. Am. Met. Soc.,
470
85(12), 1917-1930.
471
Lehner, B., K. Verdin, and A. Jarvis (2006), HydroSHEDS Technical DocumentationRep.,
472
473
[http://hydrosheds.cr.usgs.gov] pp, Washington, DC. .
Lettenmaier, D. P., and E. F. Wood (1993), Hydrologic Forecasting, in Handbook of Hydrology,
474
edited by D. R. Maidment, pp. 26.21-26.30, McGraw-Hill Inc., New York, NY, USA.
475
Liang, X., D. P. Lettenmaier, E. Wood, and S. J. Burges (1994), A simple hydrologically based
476
model of land surface water and energy fluxes for general circulation models, J. Geophys
477
Res., 99(D7), 14415-14428.
478
Liebmann, B., and D. Allured (2005), Daily Precipitation Grids for South America, Bull. Am.
479
480
Met. Soc., 86(11), 1567-1570.
Lohmann, D., R. Nolte-Holube, and E. Raschke (1996), A large-scale horizontal routing model
481
to be coupled to land surface parameterization schemes, Tellus, 48A, 708-721.
22
482
Maurer, E. P., J. C. Adam, and A. W. Wood (2009), Climate model based consensus on the
483
hydrologic impacts of climate change to the Rio Lempa basin of Central America, Hydrol.
484
Earth System Sci., 13(2), 183-194.
485
Maurer, E. P., J. D. Rhoads, R. O. Dubayah, and D. P. Lettenmaier (2003), Evaluation of the
486
snow-covered area data product from MODIS, Hydrol. Processes, 17(1), 59-71.
487
Maurer, E. P., A. W. Wood, J. C. Adam, D. P. Lettenmaier, and B. Nijssen (2002), A long-term
488
hydrologically-based data set of land surface fluxes and states for the conterminous United
489
States, J. Climate, 15(22), 3237-3251.
490
Milly, P. C. D., and A. B. Shmakin (2002), Global Modeling of Land Water and Energy
491
Balances. Part I: The Land Dynamics (LaD) Model, J. Hydrometeorology, 3(3), 283-299.
492
Mitchell, T. D., and P. D. Jones (2005), An improved method of constructing a database of
493
monthly climate observations and associated high-resolution grids, Int. J. Climatol., 25(6),
494
693-712.
495
Moriasi, D. N., J. G. Arnold, M. W. V. Liew, R. L. Bingner, R. D. Harmel, and T. L. Veith
496
(2007), Model evaluation guidelines for systematic quantification of accuracy in watershed
497
simulations, Trans. of ASABE, 50(3), 885-900.
498
Nash, J. E., and J. V. Sutcliffe (1970), River flow forecasting through conceptual models part I --
499
A discussion of principles, J. Hydrol., 10(3), 282–290.
500
New, M. G., M. Hulme, and P. D. Jones (2000), Representing twentieth-century space-time
501
climate variability. Part II: development of 1961-90 monthly grids of terrestrial surface
502
climate, J. Climate, 12, 829-856.
503
Nijssen, B., R. Schnur, and D. P. Lettenmaier (2001a), Global retrospective estimation of soil
504
moisture using the VIC land surface model, 1980-1993, J. Climate, 14(8), 1790-1808.
23
505
Nijssen, B., G. M. O'Donnell, D. P. Lettenmaier, D. Lohmann, and E. F. Wood (2001b),
506
Predicting the discharge of global rivers, J. Climate, 14, 1790-1808.
507
O'Donnell, G., B. Nijssen, and D. P. Lettenmaier (1999), A simple algorithm for generating
508
streamflow networks for grid-based, macroscale hydrological models, Hydrol. Processes,
509
13(8), 1269-1275.
510
Peterson, T. C., R. Vose, R. Schmoyer, and V. Razuvaëv (1998), Global historical climatology
511
network (GHCN) quality control of monthly temperature data, Int. J. Climatol., 18(11),
512
1169-1179.
513
Rubio-Álvarez, E., and J. McPhee (2010), Patterns of spatial and temporal variability in
514
streamflow records in south central Chile in the period 1952–2003, Water Resour. Res., 46,
515
W05514, doi:05510.01029/02009WR007982.
516
Sheffield, J., and E. F. Wood (2007), Characteristics of global and regional drought,
517
1950&#8211;2000: Analysis of soil moisture data from off-line simulation of the terrestrial
518
hydrologic cycle, J. Geophys. Res., 112(D17), D17115.
519
Sheffield, J., G. Goteti, and E. F. Wood (2006), Development of a 50-yr high-resolution global
520
521
dataset of meteorological forcings for land surface modeling, J. Climate, 19(13), 3088-3111.
Silva, V. B. S., V. E. Kousky, W. Shi, and R. W. Higgins (2007), An Improved Gridded
522
523
Historical Daily Precipitation Analysis for Brazil, J. Hydrometeorology, 8(4), 847-861.
Stewart, I. T., D. R. Cayan, and M. D. Dettinger (2005), Changes toward earlier streamflow
524
525
timing across western North America, J. Climate, 18(8), 1136-1155.
Vicuna, S., R. D. Garreaud, and J. McPhee (2010), Climate change impacts on the hydrology of
526
a snowmelt driven basin in semiarid Chile, Climatic Change, (in press).
24
527
Willmott, C. J., and K. Matsuura (2001), Terrestrial air temperature and precipitation: monthly
528
and annual time series (1950–1999) (Version 1.02)Rep., Center for Climatic Research,
529
University of Delaware, Newark, DE, USA.
530
Yapo, P. O., H. V. Gupta, and S. Sorooshian (1998), Multi-objective global optimization for
531
532
hydrologic models, J. Hydrol., 204, 83-97.
Yevjevich, V. (1972), Probability and Statistics in Hydrology, Water Resources Publications, Ft.
533
534
Collins, CO, USA.
Zhang, X., and F. Yang (2004), RClimDex (1.0) User GuideRep., Climate Research Branch,
535
Environment Canada, Downsview, Ontario, Canada.
536
Zhang, X., E. Aguilar, S. Sensoy, H. Melkonyan, U. Tagiyeva, N. Ahmed, N. Kutaladze, F.
537
Rahimzadeh, A. Taghipour, T. H. Hantosh, P. Albert, M. Semawi, M. Karam Ali, M. H. Said
538
Al-Shabibi, Z. Al-Oulan, T. Zatari, I. Al Dean Khelet, S. Hamoud, R. Sagir, M. Demircan,
539
M. Eken, M. Adiguzel, L. Alexander, T. C. Peterson, and T. Wallis (2005), Trends in Middle
540
East climate extreme indices from 1950 to 2003, J. Geophys. Res., 110(D22), D22104.
541
542
543
25
544
Table 1 - List of statistical quantities and descriptions.
Name
Description
R95p
Annual total precipitation when rainfall > 95th prctile
R99p
Annual total precipitation when rainfall > 99th prctile
PRCPTOT
Annual total precipitation in wet days (RR >=1mm)
CWD
Consecutive wet days: largest number of consecutive wet days with RR >=1mm
CDD
Consecutive dry days: largest number of consecutive dry days with RR <=1mm
SDII
Simple daily intensity index: mean annual intensity for RR > = 1 mm
R5mm
Annual count of days with Precipitation >= 5 mm
R20mm
Annual count of days with Precipitation >= 20 mm
RX1d
Maximum 1 day precipitation in the year
RX5d
Maximum 5 days precipitation in the year
PW
Probability of Wet days
PD
Probability of Dry days
PWW
Probability of a wet day followed by a wet day
PDD
Probability of a dry day followed by a dry day
545
26
546
Table 2 - Location of adjusted gridded precipitation grid cells used in daily precipitation
547
validation.
0.25º Grid Cell Abbreviation
Grid Cell Center Latitude
Grid Cell Center Longitude
Loc1
-34.375
-70.875
Loc2
-34.875
-71.125
Loc3
-35.875
-71.125
Loc4
-36.125
-71.625
548
549
27
550
551
Table 3 - Daily precipitation statistics for summer (DJF) and winter (JJA) periods, 1983-2007. OBS are
552
observations, ADJ are adjusted gridded meteorology.
Summer (DJF) Statistics
OBS
ADJ
OBS Std ADJ Std Mean
RMSE
Mean
Mean
(mm)
(mm)
(mm)
(mm)
Loc1
0.08
0.08
1.21
0.63
0.00
1.37
-0.01
Loc2
0.14
0.11
1.61
0.59
-0.03
1.71
0.02
Loc3
0.64
0.60
5.44
2.29
-0.04
5.88
0.01
Loc4
0.60
0.54
4.38
2.34
-0.07
4.99
-0.01
(mm)
Bias
Correlation
(mm)
Winter (JJA) Statistics
Loc1
3.80
3.56
10.88
13.96
-0.23
17.16
0.06
Loc2
5.64
4.72
14.17
16.72
-0.92
20.99
0.09
Loc3
13.63
12.58
30.46
37.39
-1.05
46.18
0.08
Loc4
7.24
7.36
14.51
22.08
0.12
25.64
0.06
553
554
555
28
556
557
Table 4 - Correlation coefficients between observed and adjusted daily precipitation statistical parameters.
558
Shaded values indicate the null hypothesis of equal means cannot be rejected at the 5% level based on a t-test.
R99p
R95p
PRCTOT SDII
R5mm R20mm CWD
CDD
RX1d
RX5d
Loc1
0.10
0.70
0.96
0.76
0.58
0.82
0.37
0.45
0.78
0.61
Loc2
0.40
0.65
0.92
0.65
0.61
0.84
0.43
0.38
0.79
0.53
Loc3
0.18
0.47
0.83
0.53
0.61
0.65
0.25
0.73
0.45
0.48
Loc4
0.31
0.29
0.87
0.57
0.71
0.73
0.00
0.28
0.56
0.37
559
560
29
561
562
Table 5 - Contingency table summarizing the comparisons of MODIS and VIC simulated snow cover. Values
563
are relative frequencies calculated as the total number of occurrences in each category divided by the number
564
of pixels (1530).
No Snow
Snow
Total
No Snow
0.65
0.05
0.69
Snow
0.06
0.24
0.31
Total
0.71
0.29
1.0
565
566
30
567
List of Figures
568
569
Figure 1 - Geographic location of the study area in Central Chile. From north to south the basins
570
are: Rapel, Mataquito (Mataquito river at Licanten), Maule (Claro river at Rauquen and
571
Loncomilla river at Bodega) and Itata river basins. Circles indicate the location of DGA rain
572
gauges and stars the location of the three stream gauges used in VIC simulations
573
Figure 2 - Maps of annual precipitation for the period 1951-1980. Source a) gridded global
574
observations and b) DGA. Precipitation lapse rates for latitudinal bands -35.125 S and -36.125 S
575
for c) global gridded precipitation data set and d) DGA data set.
576
Figure 3 - Scatterplots of observed and predicted monthly precipitation for the month of July.
577
Figure 4 - a) Annual adjusted global precipitation for the period 1950-2006 and b) differences
578
between the original global gridded and the adjusted global precipitation data sets.
579
Figure 5 - Location of DGA rain gauge stations and adjusted global precipitation grid points used
580
for validation of daily rainfall.
581
Figure 6 - Boxplots of statistical parameters, green represents observations and purple represents
582
adjusted precipitation for each geographic location. The bottom and top lines represent the 25th
583
and 75th percentiles and the middle line represents the median. Whiskers extend from each end
584
of the box to the adjacent values in the data within 1.5 times the Inter Quartile Range. The Inter
585
Quartile Range is the difference between the third and the first quartile, i.e., 25th and 75th
586
percentiles. Outliers are displayed with a plus sign.
587
Figure 7 - Probabilities of a) wet and b) dry days, and transition probabilities c) and d). Daily
588
observed (black) and adjusted gridded (grey) precipitation for the four selected locations.
31
589
Figure 8 - Observed and Simulated monthly flows for the Mataquito river at Licanten for the
590
calibration period (top panel) and validation period (bottom panel). Summary statistics are
591
shown in each panel.
592
Figure 9 - Monthly observed and simulated flows for the Claro river at Rauquen.
593
Figure 10 - Same as Figure 9 but for Loncomilla river at Bodega.
594
Figure 11 - Statistical properties of observed and VIC simulated streamflows in three basins:
595
Mataquito river, Claro river and Loncomilla river. (a) Center timing, (b) water year volume, (c)
596
3-day peak flows and (d) 7-day low flows.
597
Figure 12 - Comparison of snow coverage for the period August 21-28, 2002. Shaded areas
598
indicate snow coverage. a) MODIS and b) VIC simulated Snow Water Equivalent.
599
32
600
601
Figure 1 - Geographic location of the study area in Central Chile. From north to south the basins are: Rapel,
602
Mataquito (Mataquito river at Licanten), Maule (Claro river at Rauquen and Loncomilla river at Bodega)
603
and Itata river basins. Circles indicate the location of DGA rain gauges and stars the location of the three
604
stream gauges used in VIC simulations
605
33
606
607
Figure 2 - Maps of annual precipitation for the period 1951-1980. Source a) gridded global observations and
608
b) DGA. Precipitation lapse rates for latitudinal bands -35.125 S and -36.125 S for c) global gridded
609
precipitation data set and d) DGA data set.
610
611
612
34
613
614
Figure 3 - Scatterplots of observed and predicted monthly precipitation for the month of July.
615
616
35
617
618
Figure 4 - a) Annual adjusted global precipitation for the period 1950-2006 and b) differences between the
619
original global gridded and the adjusted global precipitation data sets.
620
621
36
622
623
Figure 5 - Location of DGA rain gauge stations and adjusted global precipitation grid points used for
624
validation of daily rainfall.
625
626
37
627
628
Figure 6 - Boxplots of statistical parameters, green represents observations and purple represents adjusted
629
precipitation for each geographic location. The bottom and top lines represent the 25th and 75th percentiles
630
and the middle line represents the median. Whiskers extend from each end of the box to the adjacent values
631
in the data within 1.5 times the Inter Quartile Range. The Inter Quartile Range is the difference between the
632
third and the first quartile, i.e., 25th and 75th percentiles. Outliers are displayed with a plus sign.
633
38
634
635
Figure 7 - Probabilities of a) wet and b) dry days, and transition probabilities c) and d). Daily observed
636
(black) and adjusted gridded (grey) precipitation for the four selected locations.
637
638
39
639
640
Figure 8 - Observed and Simulated monthly flows for the Mataquito river at Licanten for the calibration
641
period (top panel) and validation period (bottom panel). Summary statistics are shown in each panel.
642
643
40
644
645
Figure 9 - Monthly observed and simulated flows for the Claro river at Rauquen.
646
41
647
648
Figure 10 - Same as Figure 9 but for Loncomilla river at Bodega.
649
650
42
651
652
Figure 11 - Statistical properties of observed and VIC simulated stream flows in three basins: Mataquito
653
river, Claro river and Loncomilla river. (a) Center timing, (b) water year volume, (c) 3-day peak flows and
654
(d) 7-day low flows.
655
656
43
657
658
Figure 12 - Comparison of snow coverage for the period August 21-28, 2002. Shaded areas indicate snow
659
coverage. a) MODIS and b) VIC simulated Snow Water Equivalent.
660
44