Drought Monitoring Methods

To study agricultural droughts in Poland, a special index was constructed which was found to be a better indicator of drought than the precipitation alone. The index is defined as EP/Pveg, where EP is potential evapotranspiration and Pveg is the average sum of rainfall during vegetation period (April-

Figure 13.2 Average yield of the main cereals in Poland.

September). In Poland, the index ranges from 1.6 for areas heavily prone to drought to 1.0 for areas where sum of summer precipitation exceeds potential evapotranspiration. Such areas exist in mountainous regions in southern Poland and are characterized by high rainfall in summer months. In addition, field-level assessment of potential yield of crops is done several times throughout the growing season by experts in the Central Statistical Office. A wealth of information about actual crop conditions is available from agricultural correspondents posted in different regions of the country (www.stat.gov.pl). The results of these predictions are published in official bulletins and delivered to the Ministry of Agriculture and other governmental agencies throughout Poland.

Until about 1990, drought monitoring was limited to assessing drought conditions as described above as well as by studying agrometeorological data collected from 60 meteorological stations across Poland. Drought is considered to have occurred if some threshold conditions are met. Agrom-eteorological variables and the threshold conditions (in parentheses) are (1) precipitation deficit from April 1 to September 1 (<50% of multiyear mean); (2) precipitation deficit during the 3 dekads (1 dekad = 10-day period) preceding the drought (<25% of multiyear mean); (3) the number of successive nonrainy days (>17); (4) the difference between total precipitation and evapotranspiration from June 1 to September 1 (< 75 mm); (5) number of days for which the mean daily soil temperature at 5-cm depth

Figure 13.3 Map of drought-prone areas in Poland.

in each dekad after June 1 was higher than 25° C (>2); (6) if soil moisture was extremely low; and (7) if shortage of water was reported by farmers for at least 25% of farm wells.

Every indicator described above is assigned 1 point if the threshold condition is met and 0 point if the condition is not met. The total sum of the assigned points is denoted as the drought index (DI), which is used to assess drought conditions. A drought is classified as no drought if DI = 0; drought if DI = 1 or 2; heavy drought if DI = 3, 4, or 5; and very heavy drought if DI = 6 or 7.

The Institute of Meteorology and Water Management (www.imgw.pl) developed the above criterion of drought assessment based on point measurement (Slota et al., 1992). Figure 13.4 shows classification of drought during 1992 using the above criterion. But this criterion could not be used because it was too time consuming and labor intensive. Therefore, another approach based on satellite data was developed during the 1990s at the Remote Sensing Center of the Institute of Geodesy and Cartography, Warsaw (www.igik.edu.pl) for monitoring crop conditions and drought.

Remote Sensing-Based Crop Condition Assessment System

Using the National Atmospheric Oceanic Administration (NOAA) satellite's Advanced Very High Resolution Radiometer (AVHRR) data, three

Figure 13.4 Map of distribution of drought conditions at the end of June 1992.

indices, vegetation condition index (VCI), normalized difference vegetation index (NDVI), and temperature condition index (TCI), were used to monitor crop conditions and droughts in Poland. Chapter 6 provides a detailed description of derivation of these indices. Indices were computed for every week for 1985-98. During this period of 14 years, crop yields varied significantly and so did the indices. The following yield models were developed by regressing the deviations in yields and indices.

Analysis of TCI and VCI in different years revealed that the most important for crop yield assessment are the TCI values in weeks 16 and 22 and the VCI values in week 25. Based on these findings the model was developed using 1985-97 data, which relates yield deviation from mean (Y/Ymean) with TCI and VCI values as described below:

The above model was used to make yield predictions for 1998 for Poland. The yield predictions were issued four weeks before harvest. The results were compared with yield estimates produced with the use of conventional methods by the Central Statistical Office. The mean error of cereal yield assessment for 49 administrative units was about 4%. The results of cereal yield estimation were accepted by agricultural experts and statisticians, creating a basis for operational crop condition assessment in Poland with the use of remotely sensed data. In addition, using TCI, the 1992 drought was also detected effectively, as shown in figure 13.5 (Dabrowska-Zielinska et al., 1998, 2002). It can be seen from figure 13.5 that western part of Poland was affected by drought already in May.

Two important phases of crop development had the highest correlation with crop yield. The increase of water demand by plants responds to the increase of sensitivity of TCI during these periods. The period of significant correlation during early summer (Julian week 22-25) is critical because cereals pass through reproductive phases, when cooler weather is favorable for crop development and yield formation. The second important period was spring (week 14-16); the negative correlation for this period indicates the low temperature during spring yields of cereals in Poland. It was also found that TCI could be used to interpret soil moisture conditions over large areas.

The data used in the above system were based on using 4-km GAC (global area coverage) NOAA/AVHRR data (www.saa.noaa.gov/cocoon/ nsaa/products/) for global monitoring of crop and drought conditions. However, the diversified cropping and small land holdings that characterize Polish agriculture make 1-km LAC (local area coverage) NOAA-AVHRR data more useful for monitoring drought conditions. The research for this purpose began in cooperation with the Canada Centre of Remote Sensing, Ottawa (www.ccrs.nrcan.gc.ca), in 1996. The global land data (1km resolution) collected under International Geosphere Biosphere Program for 1992-95 were used for this research. A Geographic Information System database was developed and used for making regression analyses and for determining relationships between indices derived from remotely sensed data and parameters characterizing agricultural production (Walker, 1988; Wood, 1993; Yang et al., 1997; Boken and Shaykewich, 2002).

Figure 13.5 Development of drought conditions in Poland in 1992, characterized by temperature condition index.

Crop conditions were monitored during 1997-2001 vegetation periods by comparing dekadal NDVI values with the previous year's values and the long-term mean values for each administrative unit in Poland. Administrative units were grouped into regions to study regional differences in NDVI levels. All of the above mentioned materials were used for analyzing crop development in Poland. For example, figure 13.6 illustrates crop growth conditions in Poland in 2001 based on analysis of NDVI data.

These NDVI-based outputs were delivered to the Central Statistical Office for comparative analysis. In addition, research conducted at the Institute of Geodesy and Cartography found the following indices useful for monitoring vegetation and drought conditions in Poland:

1. Crop growth index (CGI), the ratio of NDVI to radiation temperature (NDVI/Ts). This index more precisely reflects the deficits in soil water content than the NDVI or temperature alone. The lower the CGI, the worse the vegetation conditions.

2. Soil-adjusted vegetation index, a function of NDVI, taking into account reflection from soil background (Huete, 1988; chapter 5).

3. Ratio of sensible heat to latent heat (H/LE); the higher the index, the worse the crop conditions. Figure 13.7 shows spatial variation of H/LE index for 1998 and 2000.

4. Water deficit index, the ratio of actual to potential latent heat, LE/LEp (Moran et al., 1994).

Was this article helpful?

0 0

Post a comment