Presentation is loading. Please wait.

Presentation is loading. Please wait.

Towards the Development of a Spatial Decision Support Systems for the Application of Climate Forecasts in Uruguayan Rice Production System Alvaro Roel.

Similar presentations


Presentation on theme: "Towards the Development of a Spatial Decision Support Systems for the Application of Climate Forecasts in Uruguayan Rice Production System Alvaro Roel."— Presentation transcript:

1 Towards the Development of a Spatial Decision Support Systems for the Application of Climate Forecasts in Uruguayan Rice Production System Alvaro Roel – INIA Uruguay Synthesis Workshop of the Advanced Institute on Climatic Variability and Food Security Geneva, 9-10 May 2005

2 INTRODUCTION I In selected regions of the world, climate anomalies are linked to the onset and intensity of a warm or cold event of the El Niño-Southern Oscilation (ENSO) phenomenon. ENSO is the main source of inter- annual climate variability in these regions (Uruguay).Uruguay  Current scientific level allow forecasters to provide probabilistic forecasts that give information about the likely characteristics of the seasonal climate, but they cannot indicate the exact timing, spatial distribution and eventual averages or totals of specific variables.

3 INTRODUCTION II  Although forecasts make predictions of climate variable behaviors for large regions of the world, these regions are not uniform.  The heterogeneity of these regions could determine very different spatial reactions to a given forecast. We believe that these spatial differences should be understood in order to improve the applications of climate predictions.  Therefore, we propose that an effective application of a seasonal climate forecast would need to take in consideration the natural spatial variability in biotic and abiotic conditions that regulate productivity in agricultural ecosystems.

4 A pilot project was proposed to evolve a system for the effective application of a seasonal climate forecast, which can address the natural spatial and temporal variability in growing conditions that control productivity in a rice ecosystem in Uruguay. OVERALL OBJECTIVE

5  ha of irrigated long-grain rice  More than 90% of production is exported  Rice is the second largest agricultural export after beef  Average Farm size: 400 Hectares (450 farmers)  Farmers - Industry - Research integration  Production under contract  No government intervention in commercialization  No subsidies Rice production in Uruguay

6 GIS  Crop Modeling  Forecast  Spatial Statistics TOOLS SPATIAL DECISION SUPPORT SYSTEM (SDSS)

7 OBJECTIVES 1)Evaluate ENSO effects on Uruguayan Rice Production. 2)Evaluate the capability of the DSSAT 3.5 Ceres-Rice crop simulation model in recreating the observed yield spatial and temporal variability. 3)Simulate rice yield spatial variability under different seasonal forecast scenarios: El Niño, La Niña and Neutral years.

8 MATERIALS AND METHODS OBJECTIVE 1 Evaluate ENSO effects on Uruguayan Rice Production The relationship between ENSO 3.4 average total Sea Surface Temperatures (SST) anomalies in October November and December (OND) and rice yield Yield data were obtained from the Uruguayan Rice Growers Association for the period. Yields for any given year were expressed as the relative difference (RYD) between the observed yield for that year and the yield predicted by the regression model: RYD = Yld(n) - PYld(n) * 100/PYld(n) RYD = relative yield deviation expressed in (%), Yld(n) = Observed crop yield for year n, PYld(n)= Yield predicted by regression model for year n

9 MATERIALS AND METHODS OBJECTIVE 1 Evaluate ENSO effects on Uruguayan Rice Production The SST anomalies were calculated relative to the period and aggregated into three-month period means. In order to have a more comprehensive analysis of ENSO impacts on rice production the distribution shifts of crop yields were studied using the same approach as the one used by Baetghen (1986). The detrended National average crop yield data from 1973 to 2003 were divided into quartiles and any given value was defined as being "high" if it was greater than the third quartile (upper 75% of the data), "low" if it was less than the first quartile (lower 25%), and "normal" if its value fell between the first and the third quartile (central 50% of the data). Using these values the shift in the distribution of crop yields were studied for the different ENSO phases (El Niño, La Niña and Neutral).

10 RESULTS – OBJECTIVE 1

11 RESULTS OBJECTIVE 1 Evaluate ENSO effects on Uruguayan Rice Production Figure 1. National rice yield ( ). The year correspond to the planting season

12 RESULTS OBJECTIVE 1 Evaluate ENSO effects on Uruguayan Rice Production National average yield deviations ( ) Vs Average SST anomalies for October, November and December. Green dots La Niña Years, Blue Dots Neutral years and Red dots El Niño years

13 High Yields Medium Low Yields Upper Quartile Central Quartiles Lower Quartile < % RESULTS OBJECTIVE 1 Evaluate ENSO effects on Uruguayan Rice Production RYD

14 RESULTS OBJECTIVE 1 Evaluate ENSO effects on Uruguayan Rice Production

15 National Rice Yield Distribution and ENSO phases ( )

16 MATERIALS AND METHODS OBJECTIVE 2 Evaluate the capability of the DSSAT 3.5 Ceres-Rice crop simulation model in recreating the observed yield spatial and temporal variability The study was conducted at a 12 hectares rice field located at Paso de la Laguna Experimental Unit of the National Institute of Agricultural Research (INIA), Treinta y Tres -Uruguay. Cultivar: El Paso 144, Planting Date: November Ten locations were selected in this rice field in which recording data loggers (Hobo H8 Pro) were fitted. Water and canopy temperatures were measured hourly throughout the growing season. Data logger locations were georeferenced using a back-pack differential global positioning system (DGPS) receiver (Trimble AG 132).

17 MATERIALS AND METHODS II OBJECTIVE 2 Evaluate the capability of the DSSAT 3.5 Ceres-Rice crop simulation model in recreating the observed yield spatial and temporal variability At harvest, yield, yield components, and percent blanking were recorded in the vicinity of each sensor. Interpolated yield maps of the field were created using a geographic information system (Arcview, ESRI, Redlands, CA). Yield data from each of the ten locations were spatially interpolated to a fixed 5×5 m grid using inverse distance weighted interpolation with power 2 and number of neighbors 12. Soil samples were extracted at three different depths: 0-10 cm, cm and cm at the same locations where sensors were installed. Yield was predicted at each sensor location using the DSSAT 3.5 Ceres-Rice model.

18 RESULTS – OBJECTIVE 2 Spatial Variability

19 RESULTS - OBJECTIVE 2 Spatial Variability

20 ** P ≤ * P ≤0.05 ns = indicates probability levels ≥ 0.05

21 RESULTS - OBJECTIVE 2 Spatial Variability

22

23 RESULTS – OBJECTIVE 2 Temporal Variability

24 RESULTS - OBJECTIVE 2 Temporal Variability

25 MATERIALS AND METHODS OBJECTIVE 3 Simulate rice yield spatial variability under different seasonal forecast scenarios: El Niño, La Niña and Neutral years. The crop simulation model was run in each of the ten selected locations in the field using the weather data from a series of years ( ). The same soil data, management practices (planting date, seeding rate, fertilization, etc) and genetic data (El Paso 144) that were used in the studied field for growing season were applied at each of the ten locations through out all of these years. Yield data from each of the ten locations and for each of the 31 growing seasons were spatially interpolated in order to generate yield maps for each growing seasons. To be able to display the yield range variability along these 31 growing seasons with a common legend the whole data set of yield outcomes was divided in quartiles. These quartiles defined the range of variability of the different four yield classes.

26 MATERIALS AND METHODS OBJECTIVE 3 Simulate rice yield spatial variability under different seasonal forecast scenarios: El Niño, La Niña and Neutral years. Temporal patterns were determined based on a cluster analysis of the standardized annual yields. K-means clustering algorithm in Statistica software package was used. The Moran Coefficient was used to test the if the cluster detected in the previous step were spatially auto correlated or random distributed.

27 RESULTS – OBJECTIVE 2 Spatio-Temporal Variability

28 Yield Spatial Variability. Red years correspond to El Niño, Blue years correspond to La Niña Black years to Neutral conditions.

29

30

31 CONCLUSIONS This study showed that the distribution of national rice yield averages varied with ENSO phases. The frequency of high national rice yields was more than two times higher in La Niña years than in Neutral years. The DSSAT 3.5 Ceres-Rice model was able to capture satisfactorily well rice yield spatial and temporal variability. The study conducted in the 12 hectares rice field showed that this field presented certain yield spatial pattern with high yielding areas at the north and center portion of the field and a low yielding area at the south portion of the field. When the model was run spatially, at the different locations within the field and temporally along the different growing seasons, the same pattern of yield spatial variation can be observed. Overall, this suggest that there is no interaction between temporal and spatial effects, there were no climatic conditions (temporal variability) that could make that the south portion of the field would achieved a higher yields than the north portion.

32 CONCLUSIONS II This study was able to demonstrate that for this rice field although we were able to characterize its yield spatial variability very precisely this pattern of spatial variability did not changed with different climatic conditions. Therefore, what regulate yield spatial variability within this field are factors related with the soil and not with the climatic conditions. Consequently, at the scale of this field forecast output scale did not constitute a problem. We believe that the approach used in this study can be implemented at larger spatial scale to evaluate at which level of spatial resolution forecast output scale start to become a problem.

33 FUTURE STEPS Apply same procedure at a larger scale: -Larger Rice Field -Different regions within the country

34 18% 12% 70% Rice Production in Uruguay

35

36 URUGUAY 30 o 35 o back


Download ppt "Towards the Development of a Spatial Decision Support Systems for the Application of Climate Forecasts in Uruguayan Rice Production System Alvaro Roel."

Similar presentations


Ads by Google