Improving Applications in Agriculture of ENSOBased Seasonal Rainfall Forecasts Considering Atlantic Ocean Surface Temperatures

G. O. Magrin • M. I. Travasso • W. E. Baethgen • R. T. Boca

23.1

Introduction

Climate uncertainties, derived from annual climatic variability, often lead to conservative crop management strategies that sacrifice some productivity to reduce the risk of losses in bad years. The availability of ENSO-based climate forecasts has led many to believe that such forecasts may benefit decision-making in agriculture. The forecasting capability may allow the mitigation of negative effects of ENSO-related climate variability as well as taking advantage of favorable conditions (Stern and Easterling 1999).

Benefits of using ENSO-based climate forecasts have been demonstrated in South America. Changing crop mix (Messina et al. 1999) or crop management options were proposed as adaptive measures to cope with climatic variability (Magrin et al. 2000; Jones et al. 2000). However, the large inconsistency of the precipitation signal within ENSO phases led to considerable overlap in yields and net returns for the various ENSO phases (Ferreira et al. 2001), decreasing the potential usefulness of the forecasts (Magrin and Travasso 2001; Podestá et al. 2002).

But ENSO is not the unique source of climatic variability in southeastern South America. Evidence of the influence of South Atlantic Ocean (SAO) on precipitation was presented for Uruguay and south Brazil by Díaz et al. (1998). Barros et al. (2000) signaled the influence of the South Atlantic Convergence Zone (SACZ) on midsummer interannual variability of the low-level circulation and precipitation in subtropical South America. Recently Berri and Bertossa (2004) reported that the Atlantic Ocean influences seasonal precipitation over the northwestern and southeastern parts of southern central South America.

Furthermore, in previous works significant relationships were found between SAO SST anomalies and crop yields or precipitation anomalies in the Pampas region of Argentina. In comparison to ENSO or SSTs from the Pacific, SAO SSTs presented a stronger signal on crop yields in the southern part of the region, especially for maize (Travasso et al. 2003a,b).

These antecedents encourage the consideration of SAO SST anomalies as a way to improve climate forecasting and decision-making in agriculture.

The aim of the present work was to explore the capability of considering SAO by itself and in conjunction with ENSO phases to optimize maize agronomic management practices and, to assess the additional economic value of including SAO information in an ENSO-based seasonal forecast.

23.2

Methods

A location placed in the southeastern part of the Argentina's Pampas region, Azul (latitude 36.8° S, long 59.9° W), was selected as case study. Daily climatic data for maximum and minimum temperature, precipitation and solar radiation were available since 1931 from the National Meteorological Service.

CERES-Maize model included in DSSAT V3.5 (Tsuji et al. 1994) was used to examine the benefits of tailoring crop production decisions to different types of climate forecasts. The model had been previously calibrated and validated in the region with estimation errors for yield predictions lower than 10% at the field level (Travasso and Magrin 2001).

Different types of climate forecasts were used based on: (a) ENSO phases (neutral, El Niño and La Niña) following the Japan Meteorological Agency classification, (b) three monthly (November-December-January) rainfall categories, and (c) South Atlantic Ocean SST anomalies (SAO).

Smoothing techniques (Cleveland et al. 1988) were used for isolating the low frequency variability in monthly precipitation record. Then, the anomalies (difference between observed and smoothed values) were classified in terciles obtaining three rainfall categories: wet (upper tercile), normal and dry (lower tercile).

South Atlantic Ocean SST anomalies (SAO) (0-20° S, 30° W-10° E) were obtained from the NOAA website. SAO values corresponding to August and September, which are significantly related to maize yield in this location (Travasso et al. 2003a) were used. SAO anomalies were classified in quartiles and 3 categories were used: warm (wSAO = upper quartile), neutral (between probability of 75 and 25%) and cold (cSAO = lower quartile).

Model runs were done for the period 1931-2002 considering the soil series predominant for Azul (Typic Argiudoll) and the most frequent farm management: planting on 30 October with a plant density of 7 plants m-2, and a nitrogen fertilizer rate of 60 kg N ha-1. These runs were taken as the baseline data and corresponded to expected yields when climate forecast is not considered. The crop's gross margin was calculated according to the prices presented in Table 23.1.

Optimal management options for each climate forecast method were obtained by varying planting dates (15-day intervals starting at 15 October) and nitrogen doses (0, 20, 40, 60, 80, 100, 120 kg N ha-1). The best option for each extreme phase (El Niño, La Niña, wet, dry, wSAO and cSAO) was defined as the one producing the highest gross margin. For the years classified as neutral or normal the management practices were always the same (typical farm management) assuming that those years, farmers would not use climate forecasts in their decision-making.

The economic value of climate forecast was calculated as the difference in gross margin between the best management option for each forecast and the typical management without considering forecast.

23.3 Results

The cumulative probability for grain yields under the typical farmer management for the different climate predictions methods (precipitation terciles, ENSO, and SAO

Table 23.1. Prices considered for gross margin calculation

Item

Price (U.S.$)

Labour

69 ha-1

Fertilizer application

12 ha-1

Seed

57 ha"1

Herbicide

34 ha-1

Insecticide 1 ha

Fertilizer (urea) 0.34 kg'

Fertlizer (diammonium phosphate) 0.35 kg Harvest 8%

Commercialization 27%

Maize price (April 2005) 771"1

Insecticide 1 ha

Fertilizer (urea) 0.34 kg'

Fertlizer (diammonium phosphate) 0.35 kg Harvest 8%

Commercialization 27%

Maize price (April 2005) 771"1

anomalies) are presented in Fig. 23.1. The best method allowing to discriminate among yield categories was "precipitation terciles" (i.e. assuming a "perfect" forecast). The use of "ENSO phases" was useful only in 50% of the years, while "SAO anomalies" clearly separated the highest yields.

This result suggests that maize yields are likely to be driven not only by the influence of ENSO phases but also by South Atlantic Ocean conditions. Figure 23.2 shows the relationship between maize yields and SAO temperature anomalies. Upper quartile SAO anomalies were consistently associated with mean or high yield levels, with only one exception. It is important to emphasize that our results suggest that even under La Niña or neutral years, high or normal maize yields could be expected if SAO anomalies in August and September are higher than normal. However, with normal or low SAO anomalies yield behavior was erratic.

ENSO phases were combined with SAO anomalies in an attempt to improve yield predictions. In Fig. 23.3 maize yields were regrouped as La Niña (all La Niña years except those with warm SAO anomalies), neutral (all neutral years except those with warm SAO anomalies) and a third group including El Niño years plus warm SAO years. Because this combination seems to be a better approach to separate yields categories, we decided to consider it as a fourth climate forecasting method.

Optimal management options, grain yields and gross margins for each one of the considered climate forecast are summarized in Table 23.2. Expected yields in Azul averaged 7.70, 8.48, 8.18, 8.02 and 8.39 t ha-1 for most common farmer management and management optimized by rainfall terciles, ENSO, SAO and ENSO + wSAO, respectively. For gross margin these figures were 140, 172, 155, 147, and 162 U.S.$ ha-1.

Optimal crop management options for less favorable years (La Niña, Dry) resulted in later planting dates and lower N rates. For more favorable years (El Niño, Wet and wSAO) higher N rates was a better option, although the optimal planting date differed among methods (Table 23.2). These differences in optimal crop management evidenced between El Niño and Warm SAO could be attributed to differences in their signal on precipitation. During El Niño years rainfall tends to be higher than normal in November-December (Barros et al. 1996; Magrin et al. 1998), while Warm SAO episodes are

Fig. 23.1. Cumulative probability for simulated grain yields for each weather category

Fig. 23.1. Cumulative probability for simulated grain yields for each weather category

positively correlated with October-February precipitations (Travasso et al. 2003b). Because maize crops are highly sensitive to water shortage during the pre-flowering period, for planting dates in mid October (like in El Niño years) water availability will be crucial during December, but late planting dates (wSAO) will be more dependant on January rainfall. As shown in Fig. 23.4 precipitation anomalies in Azul tended to be higher in January during the wSAO years.

The economic value (EV) of forecast (Table 23.3) was obviously the best when considering precipitation terciles (22.9%). The EV for individual ENSO phases (10.5%) or SAO anomalies (5%) was considerably lower. However using ENSO forecast and taking into account warm SAO anomalies during August and September could signifi-

Fig. 23.2. Relationship between simulated maize yield and South Atlantic Ocean surface temperature anomalies

Fig. 23.2. Relationship between simulated maize yield and South Atlantic Ocean surface temperature anomalies

cantly increase the incomes (15.9%). It is important to note that in dry years the EV attained 90% while in the wet years it ranged between 15 and 30% (Fig. 23.5).

Variability in precipitation within an ENSO phase is one of the most important obstacles for forecast's adoption. For example, if dry conditions are expected during a given ENSO event but do not materialize (as happened in 1999-2000 in the western Pampas), cold events will not appear to be very salient or memorable. (Podestá et al. 2002). In this particular year, classified as La Niña according to Pacific conditions, SAO temperatures were significantly higher than normal and, as mentioned above, warm SAO is associated with positive rain/yield anomalies in the southern Pampas. Precipitation in December, January and February in Azul was 25.0, 9.0 and 134.0 mm over the mean values.

Therefore combining both approaches (ENSO + SAO) could be promising for improving the applications of ENSO-based seasonal forecasts in agriculture.

Table 23.2. Optimal management options and expected outcomes for different climate forecasts

Years n Planting Total N applied Predicted yield Predicted margin date (kg ha"1) (tha-1) (US$ha~1)

All years (according to most frequent farmer management)

All 68 30 October 60 7.70 140 Optimized by three monthly precipitation terciles (November to January)

Wet 16 15 October 120 11.66 294

Normal 35 30 October 60 8.11 160

Dry 17 30 November 40 6.23 81

Mean 8.48 172 Optimized by ENSO phase

El Niño 14 15 October 120 10.31 226

Neutral 39 30 October 60 7.63 136

La Niña 15 15 November 60 7.63 136

Mean 8.18 155 Optimized by SAO temperature anomalies

Warm 20 15 November 80 9.62 221

Neutral 32 30 October 60 7.09 109

Cold 16 30 October 80 7.88 132

Mean 8.02 147 Optimized by ENSO phase and SAO temperature anomalies

El Niño 14 15 October 120 10.31 226

Warm SAO 14 15 November 80 9.65 223

Neutral 30 30 October 60 7.42 126

La Niña 10 15 November 60 6.85 97

Mean 8.39 162

SAO = South Atlantik Ocean.

Table 23.3. Absolute and relative value of optimal use of various types of perfect seasonal forecast for maize management in Azul

Forecast value

Rain terciles

ENSO phases

SAO anomalies

ENSO + SAO

Absolute (U.S.$ ha"1)

31.9

14.7

7.0

22.2

Relative (%)

22.9

10.5

5.0

15.9

Fig. 23.4. Precipitation anomalies during December, January and February for; a El Niño years; b warm SAO years

Fig. 23.4. Precipitation anomalies during December, January and February for; a El Niño years; b warm SAO years

Fig. 23.5. Predicted yields and 11

economic value of ENSO-SAO climate forecast ^ 1 ^

irt 5

T-T 105

Was this article helpful?

0 0

Post a comment