Kenya and Tanzania 2005 Crop Tour Summary

A joint mission between the USDA Foreign Agricultural Service (FAS) and the USAID FEWS NET Project conducted a crop tour in Kenya and northeastern Tanzania from June 23-July 5, 2005. The purpose of the crop tour was to assess current crop conditions, collect ancillary data from local industry sources, and validate several operational crop models currently used by FEWS NET and USDA/FAS analysts. Digital photos and spatial model images from this year's crop tour are presented at “WRSI Relative Yield Model Compared to Ground Conditions” and "2005 Kenya-Tanzania Crop Tour Pictures". In addition, recent Kenya and Tanzania corn production estimates derived from the 2005 crop tour are described in the PECAD articles titled, "Record Corn Area Increases Production in Tanzania" and “Kenya’s Corn Production Greater than Last Year”.

The 2005 crop tour transected most major crop areas within Kenya's corn belt and key agricultural regions in northeastern Tanzania. All digital photos were linked to GPS route data to georeference the photos and document field conditions. These georeferenced photos help to determine whether spatial models properly reflect general crop conditions year-to-year and will be archived for future comparisons with spatial product results. Field comparisons between ground conditions and spatial model outputs are visual and qualitative in nature, but general results from the crop tour indicate the operational spatial models from FEWS NET were adequately reflecting ground conditions.

Georeferenced digital photos from the crop tour can be viewed at "2005 Kenya-Tanzania Crop Tour Pictures". A new GPS-photo link software package was utilized that linked the time stamps from digital photos to the time log from GPS track routes. The linking process takes about 45-minutes which includes downloading the digital photos; downloading the GPS track route; and linking the digital photos to GPS track routes so that the digital photos are georeferenced with latitude/longitude coordinates. In addition, the software creates two ESRI shape files (GPS route and georeferenced photos); provides two extensions for hotlink display of photos in ArcView or ArcGIS; adds latitude/longitude watermarks to the photos; creates a web page with metadata for viewing the photos; and links the pictures to Mapquest or other web mapping utilities.

It is envisaged that future crop tours conducted by FEWS NET and USDA/FAS will georeference all digital photos for continuous crop model validation from year-to-year. USDA/FAS has been conducting similar studies in South Africa (2004 and 2005) and these crop surveys have shown that, in most cases, the operational FEWS NET spatial models closely reflect ground conditions.

Start of Season (SoS) and WRSI Relative-Yield Models

Africa Data Dissemination Service (ADDS) provides operational crop model products for Start of Season (SoS), Season Progress, and Water Requirement Satisfaction Index (WRSI). These spatial products were validated by making field observations; conducting interviews with government personnel, agronomists, traders, and farmers; and utilizing GPS units and digital cameras.

The SoS algorithm is based on 45-millimeter rainfall accumulation over a one-month period, and initializes the Season Progress and WRSI relative-yield models. In addition, a SoS anomaly image is produced by subtracting the current SoS images from the average SoS image derived from long-term average rainfall data. The SoS anomaly image helps to depicts whether the current season arrived earlier or later than average SoS values. However, it should be noted that the SoS model does not always model the exact planting dates as different agroclimatic conditions and local practices may require farmers to plant seeds in dry soil several weeks before rains are scheduled to arrive while other farmers may plant several weeks after the rains have arrived and soil moisture conditions are very moist.

For Kenya and Tanzania, the SoS and SoS anomaly images accurately reflected ground conditions except in the highlands of southwestern Kenya where rainfall is abundant and the growing seasons is longer than six months due to slow plant growth in cool regions. In these high-rainfall regions of southwestern Kenya, local farmers practice “relay planting" in which different fields have staggered planting dates and one finds fields with mature crops grown next to fields recently planted.

In Tanzania, the SoS product had difficulty near the highland regions near Arusha where the unimodal (one rainy period during the year) and bimodal (two rainy periods per year) rainfall patterns merged together and the dividing line between the two patterns tends to be a function of elevation. Reports from local experts confirmed a GIS rainfall analysis that shows highland regions near Arusha are unimodal (with one long growing season caused by slow plant growth from cooler temperatures), and lowland regions tend to be bimodal (with two shorter growing seasons and rapid plant growth from warmer temperatures).

Digital photos comparing the WRSI relative-yield model to crop conditions on the ground can be viewed from “WRSI Relative Yield Model Compared to Ground Conditions” .

Season Progress Model

The Season Progress model from ADDS is largely dependent on SoS and Length of Growing Period (LGP) data layers. The LGP database for East Africa was revised before the crop tour as the previous LGP data layer did not account for longer growing seasons greater than 180-days and it mistakenly represented Kenya's major grain basket in the Uasin Gishu (Eldoret) and Trans Nzoia (Kitale) districts as having only 120-day seasons. The crop tour mission validated the crop stage in the revised season progress model and the survey team felt the final season progress model closely reflects ground conditions.

In Kenya, the crop stage for the long-rains season (March-December) varied from vegetative to maturity with most of the crop ranging from vegetative to tasseling stages in the mid- to higher-elevation regions, and tasseling to maturity stages in the lowland regions. In northeastern Tanzania, the crop stage was more advanced than in Kenya and ranged from tasseling to harvest stages with most of the crop in the grain-filling stages.

One important field technique was to estimate "average" crop stages across a valley as several different crop stages may be recorded within the same general vicinity. In view of this procedure, the mission team found a comprehensive crop tour in late June or early July was an ideal time for crop assessment tours in Kenya as most of the corn crop was in advance stages in the lowland regions and beginning to tassel in the highland regions.

Landcover Data Layers

Kenya’s Department of Remote Sensing and Resources Survey (DRSRS) developed corn density maps derived from aerial photos and radiometer measurements. The mission team found that DRSRS digital corn density maps were in close agreement to observed ground conditions and to FAO’s Africover database for Kenya. Therefore, the mission team recommended that WRSI model utilizes a crop mask, as demonstrated for WRSI data for Tanzania, from FAO’s Africover project so that non-crop areas are not displayed with the WRSI product.

Conclusions

The main conclusions from the Kenya-Tanzania 2005 crop tour include:

The SoS, Season Progress, and WRSI relative-yield products available from ADDS properly represented ground conditions in both lowland and highland regions visited in Kenya and Tanzania. However, one region of concern is the Arusha-Moshi region where Rainfall Estimates (RFE) utilized in the WRSI model appear to underestimate rainfall received in this highland region.
The Season Progress model was greatly improved by adjusting the LGP data layer for the East African highlands. Further LGP validation is required in other parts of East Africa, especially in the Ethiopia highlands.
The SoS model appears to objectively identify planting opportunities and provide an objective method to determine if planting conditions start early or late (in comparison to long-term averages), given the wide variety of planting conditions and opportunities in East Africa.
The software for geoferencing all digital photos taken during the tour (more than 1700 photos total) was very useful. A software that links digital photos to GPS routes is recommended for future crop tours so that digital photos taken in the field can be easily georeferenced and archived for future spatial crop model validation exercises.
Both FEWS NET regional and national offices expressed the desire to be equipped with GPS receivers, digital cameras, and GPS-Photo Link software, which would help to document general field conditions and be very useful for future spatial crop model validation.

Additional WRSI model improvements recommended include:

SoS, Season Progress, and WRSI products from ADDS should be made available to users in ESRI GIS format for facilitating future field validation exercises.
The WRSI relative-yield model for East Africa requires a crop mask so that non-crop areas and national parks and forests are not included as crop regions. FAO’s Africover database should serve is suitable crop mask for East Africa, with the exception for Ethiopia which did not develop a FAO Africover product.
The WRSI product requires additional Rift Valley lakes to be added which will assist users to locate and benchmark key agricultural regions along East Africa's Rift Valley system.
The WRSI product results did not agree with ground conditions in the northern Tanzania highlands near the Arusha-Moshi region. Further investigations are currently being conducted to rectify this error. The Arusha-Moshi region is difficult to model because it is where unimodal and bimodal rainfall patterns transition and RFE data from NOAA are often underestimated in highland regions.